diff --git a/internal/admin/handler.go b/internal/admin/handler.go
index 058ba2ce94..d0835febdd 100644
--- a/internal/admin/handler.go
+++ b/internal/admin/handler.go
@@ -92,6 +92,7 @@ func (h *Handler) Ping(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /admin/login [post]
 func (h *Handler) Login(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req service.EmailLoginRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, common.CodeBadRequest, nil, err.Error())
@@ -100,7 +101,7 @@ func (h *Handler) Login(c *gin.Context) {
 
 	// Use userService.LoginByEmail with adminLogin=true
 	// This allows default admin account to log in admin system
-	user, code, err := h.userService.LoginByEmail(&req)
+	user, code, err := h.userService.LoginByEmail(ctx, &req)
 	if err != nil {
 		common.ErrorWithCode(c, code, err.Error())
 		return
@@ -816,6 +817,7 @@ func (h *Handler) TestSandboxConnection(c *gin.Context) {
 // Validates that the user is authenticated and is a superuser (admin)
 func (h *Handler) AuthMiddleware() gin.HandlerFunc {
 	return func(c *gin.Context) {
+		ctx := c.Request.Context()
 		token := c.GetHeader("Authorization")
 		if token == "" {
 			common.ErrorWithCode(c, common.CodeUnauthorized, "Missing authorization header")
@@ -824,7 +826,7 @@ func (h *Handler) AuthMiddleware() gin.HandlerFunc {
 		}
 
 		// Get user by access token
-		user, code, err := h.userService.GetUserByToken(token)
+		user, code, err := h.userService.GetUserByToken(ctx, token)
 		if err != nil {
 			common.ResponseWithHttpCodeData(c, http.StatusUnauthorized, code, nil, "Invalid access token")
 			c.Abort()
diff --git a/internal/agent/audio/tts_dispatch.go b/internal/agent/audio/tts_dispatch.go
index 272fdaf435..9952e67aa9 100644
--- a/internal/agent/audio/tts_dispatch.go
+++ b/internal/agent/audio/tts_dispatch.go
@@ -61,6 +61,7 @@ import (
 // consumer.
 type TTSDispatcher interface {
 	AudioSpeech(
+		ctx context.Context,
 		providerName, instanceName, modelName, modelID *string,
 		userID string,
 		audioContent *string,
@@ -123,6 +124,7 @@ func NewTTSDispatchFunc(d TTSDispatcher) ModelProviderFunc {
 
 		text := req.Text
 		resp, code, err := d.AudioSpeech(
+			ctx,
 			nil, // providerName — let the dispatcher resolve by name
 			nil, // instanceName — same
 			modelName,
diff --git a/internal/agent/audio/tts_dispatch_test.go b/internal/agent/audio/tts_dispatch_test.go
index 43e17dc929..22e092b245 100644
--- a/internal/agent/audio/tts_dispatch_test.go
+++ b/internal/agent/audio/tts_dispatch_test.go
@@ -51,6 +51,7 @@ type fakeTTSDispatcher struct {
 }
 
 func (f *fakeTTSDispatcher) AudioSpeech(
+	ctx context.Context,
 	providerName, instanceName, modelName, modelID *string,
 	userID string,
 	audioContent *string,
diff --git a/internal/cli/user_command.go b/internal/cli/user_command.go
index 779b54c40e..12ed05d9a7 100644
--- a/internal/cli/user_command.go
+++ b/internal/cli/user_command.go
@@ -18,6 +18,7 @@ package cli
 
 import (
 	"bufio"
+	"context"
 	"encoding/base64"
 	"encoding/json"
 	"errors"
@@ -3558,7 +3559,8 @@ func (c *CLI) APIParseLocalFileCommand(cmd *Command) (ResponseIf, error) {
 		return nil, fmt.Errorf("failed to read dsl file: %w", err)
 	}
 
-	parseResult := fileParser.ParseWithResult(filename, fileContent)
+	ctx := context.Background()
+	parseResult := fileParser.ParseWithResult(ctx, filename, fileContent)
 	if parseResult.Err != nil {
 		return nil, formatRequestError("parse local file", parseResult.Err)
 	}
diff --git a/internal/dao/user.go b/internal/dao/user.go
index 6ecbabc1ea..2aedbc50b3 100644
--- a/internal/dao/user.go
+++ b/internal/dao/user.go
@@ -36,7 +36,7 @@ func (dao *UserDAO) Create(user *entity.User) error {
 }
 
 // GetByID get user by ID
-func (dao *UserDAO) GetByID(id uint) (*entity.User, error) {
+func (dao *UserDAO) GetByID(ctx context.Context, id uint) (*entity.User, error) {
 	var user entity.User
 	err := DB.First(&user, id).Error
 	if err != nil {
diff --git a/internal/entity/models/302ai.go b/internal/entity/models/302ai.go
index 11e4ef31a8..960e1a17e4 100644
--- a/internal/entity/models/302ai.go
+++ b/internal/entity/models/302ai.go
@@ -71,7 +71,7 @@ func validateAI302DocumentURL(rawURL string) (string, error) {
 	return documentURL, nil
 }
 
-func (a *AI302Model) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AI302Model) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -147,7 +147,7 @@ func (a *AI302Model) ChatWithMessages(modelName string, messages []Message, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -219,7 +219,7 @@ func (a *AI302Model) ChatWithMessages(modelName string, messages []Message, apiC
 	return chatResponse, nil
 }
 
-func (a *AI302Model) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AI302Model) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -302,7 +302,7 @@ func (a *AI302Model) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -365,7 +365,7 @@ func (a *AI302Model) ChatStreamlyWithSender(modelName string, messages []Message
 	return sender(&endOfStream, nil)
 }
 
-func (a *AI302Model) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AI302Model) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -394,7 +394,7 @@ func (a *AI302Model) Embed(modelName *string, texts []string, apiConfig *APIConf
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -446,7 +446,7 @@ func (a *AI302Model) Embed(modelName *string, texts []string, apiConfig *APIConf
 	return embeddings, nil
 }
 
-func (a *AI302Model) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AI302Model) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -486,7 +486,7 @@ func (a *AI302Model) Rerank(modelName *string, query string, documents []string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -535,7 +535,7 @@ func (a *AI302Model) Rerank(modelName *string, query string, documents []string,
 	return &rerankResponse, nil
 }
 
-func (a *AI302Model) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AI302Model) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -623,7 +623,7 @@ func (a *AI302Model) TranscribeAudio(modelName *string, file *string, apiConfig
 	}
 
 	// build request
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -663,20 +663,20 @@ func (a *AI302Model) TranscribeAudio(modelName *string, file *string, apiConfig
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (a *AI302Model) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AI302Model) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", a.Name())
 }
 
-func (a *AI302Model) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AI302Model) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	// TODO https://302ai-en.apifox.cn/225254060e0
 	return nil, fmt.Errorf("%s no such method", a.Name())
 }
 
-func (a *AI302Model) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AI302Model) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", a.Name())
 }
 
-func (a *AI302Model) OCRFile(modelName *string, content []byte, urls *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AI302Model) OCRFile(ctx context.Context, modelName *string, content []byte, urls *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -721,7 +721,7 @@ func (a *AI302Model) OCRFile(modelName *string, content []byte, urls *string, ap
 		return nil, fmt.Errorf("failed to marshal json payload: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -771,7 +771,7 @@ func (a *AI302Model) OCRFile(modelName *string, content []byte, urls *string, ap
 	}, nil
 }
 
-func (a *AI302Model) ParseFile(modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AI302Model) ParseFile(ctx context.Context, modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -803,7 +803,7 @@ func (a *AI302Model) ParseFile(modelName *string, content []byte, documentURL *s
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", apiURL, bytes.NewBuffer(jsonData))
@@ -843,7 +843,7 @@ func (a *AI302Model) ParseFile(modelName *string, content []byte, documentURL *s
 	}, nil
 }
 
-func (a *AI302Model) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AI302Model) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -855,7 +855,7 @@ func (a *AI302Model) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, a.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -895,20 +895,20 @@ func (a *AI302Model) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return models, nil
 }
 
-func (a *AI302Model) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AI302Model) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", a.Name())
 }
 
-func (a *AI302Model) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AI302Model) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (a *AI302Model) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AI302Model) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", a.Name())
 }
 
-func (a *AI302Model) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AI302Model) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -924,7 +924,7 @@ func (a *AI302Model) ShowTask(taskID string, apiConfig *APIConfig) (*TaskRespons
 	}
 	apiURL := fmt.Sprintf("%s/%s/%s", resolvedBaseURL, a.baseModel.URLSuffix.DocumentParse, url.PathEscape(strings.TrimSpace(taskID)))
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", apiURL, nil)
diff --git a/internal/entity/models/302ai_test.go b/internal/entity/models/302ai_test.go
index 6052529d15..1c7eace058 100644
--- a/internal/entity/models/302ai_test.go
+++ b/internal/entity/models/302ai_test.go
@@ -89,11 +89,13 @@ func TestAI302ChatForcesNonStreaming(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
 	stream := true
 	thinking := true
 	resp, err := newAI302ForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"gpt-5",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -131,11 +133,13 @@ func TestAI302StreamForcesStreaming(t *testing.T) {
 		}, "\n"))
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
 	stream := false
 	var content, reasoning []string
 	err := newAI302ForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"gpt-5",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -178,9 +182,10 @@ func TestAI302ListModelsHappyPath(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
-	models, err := newAI302ForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newAI302ForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -196,12 +201,13 @@ func TestAI302ListModelsRejectsMalformedResponse(t *testing.T) {
 		"empty id":     map[string]interface{}{"data": []map[string]string{{"id": ""}}},
 	} {
 		t.Run(name, func(t *testing.T) {
+			ctx := t.Context()
 			srv := newAI302Server(t, func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 				_ = json.NewEncoder(w).Encode(response)
 			})
 			defer srv.Close()
 
-			if _, err := newAI302ForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey}); err == nil {
+			if _, err := newAI302ForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 				t.Fatal("expected malformed response error")
 			}
 		})
@@ -226,9 +232,10 @@ func TestAI302ShowTaskEscapesTaskID(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
-	resp, err := newAI302ForTest(srv.URL).ShowTask(" task/with?query#fragment ", &APIConfig{ApiKey: &apiKey})
+	resp, err := newAI302ForTest(srv.URL).ShowTask(ctx, " task/with?query#fragment ", &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ShowTask: %v", err)
 	}
@@ -246,7 +253,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 	docURL := "https://example.com/doc.pdf"
 	invalidURL := "ftp://example.com/doc.pdf"
 	send := func(*string, *string) error { return nil }
-
+	ctx := t.Context()
 	tests := []struct {
 		name string
 		run  func() error
@@ -255,7 +262,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "chat api key",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, &ChatConfig{}, &common.ModelUsage{})
+				_, err := newAI302ForTest("http://unused").ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, &ChatConfig{}, &common.ModelUsage{})
 				return err
 			},
 			want: "api key is required",
@@ -263,7 +270,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "chat model",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ChatWithMessages("  ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, &ChatConfig{}, &common.ModelUsage{})
+				_, err := newAI302ForTest("http://unused").ChatWithMessages(ctx, "  ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, &ChatConfig{}, &common.ModelUsage{})
 				return err
 			},
 			want: "model name is required",
@@ -271,21 +278,21 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "stream api key",
 			run: func() error {
-				return newAI302ForTest("http://unused").ChatStreamlyWithSender("gpt-5", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
+				return newAI302ForTest("http://unused").ChatStreamlyWithSender(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
 			},
 			want: "api key is required",
 		},
 		{
 			name: "stream sender",
 			run: func() error {
-				return newAI302ForTest("http://unused").ChatStreamlyWithSender("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
+				return newAI302ForTest("http://unused").ChatStreamlyWithSender(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 			},
 			want: "sender is required",
 		},
 		{
 			name: "embed model",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").Embed(nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").Embed(ctx, nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -293,7 +300,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "rerank api key",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").Rerank(&model, "q", []string{"doc"}, nil, nil, nil)
+				_, err := newAI302ForTest("http://unused").Rerank(ctx, &model, "q", []string{"doc"}, nil, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -301,7 +308,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "rerank query",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").Rerank(&model, "  ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").Rerank(ctx, &model, "  ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "query is required",
@@ -309,7 +316,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "asr model",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").TranscribeAudio(nil, &docURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").TranscribeAudio(ctx, nil, &docURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -317,7 +324,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "asr file",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").TranscribeAudio(&model, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").TranscribeAudio(ctx, &model, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "file is missing",
@@ -325,7 +332,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "ocr api key",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").OCRFile(&model, nil, &docURL, nil, nil, nil)
+				_, err := newAI302ForTest("http://unused").OCRFile(ctx, &model, nil, &docURL, nil, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -333,7 +340,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "ocr input",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").OCRFile(&model, nil, &blankURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").OCRFile(ctx, &model, nil, &blankURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "file url or content is required",
@@ -341,7 +348,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "ocr invalid url",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").OCRFile(&model, nil, &invalidURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").OCRFile(ctx, &model, nil, &invalidURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "invalid document URL",
@@ -349,7 +356,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "parse file url",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ParseFile(&model, nil, &blankURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").ParseFile(ctx, &model, nil, &blankURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "valid public document URL",
@@ -357,7 +364,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "parse file invalid url",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ParseFile(&model, nil, &invalidURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newAI302ForTest("http://unused").ParseFile(ctx, &model, nil, &invalidURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "invalid document URL",
@@ -365,7 +372,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "models api key",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ListModels(&APIConfig{})
+				_, err := newAI302ForTest("http://unused").ListModels(ctx, &APIConfig{})
 				return err
 			},
 			want: "api key is required",
@@ -373,7 +380,7 @@ func TestAI302ValidatesInputs(t *testing.T) {
 		{
 			name: "show task id",
 			run: func() error {
-				_, err := newAI302ForTest("http://unused").ShowTask("  ", &APIConfig{ApiKey: &apiKey})
+				_, err := newAI302ForTest("http://unused").ShowTask(ctx, "  ", &APIConfig{ApiKey: &apiKey})
 				return err
 			},
 			want: "task id is required",
diff --git a/internal/entity/models/aliyun.go b/internal/entity/models/aliyun.go
index 0d4acdcbfd..80292015ed 100644
--- a/internal/entity/models/aliyun.go
+++ b/internal/entity/models/aliyun.go
@@ -52,7 +52,7 @@ func (a *AliyunModel) Name() string {
 	return "Tongyi-Qianwen"
 }
 
-func (a *AliyunModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AliyunModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -117,7 +117,7 @@ func (a *AliyunModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -192,7 +192,7 @@ func (a *AliyunModel) ChatWithMessages(modelName string, messages []Message, api
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (a *AliyunModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AliyunModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -258,7 +258,7 @@ func (a *AliyunModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -381,7 +381,7 @@ type aliyunUsage struct {
 }
 
 // Embed embeds a list of texts into embeddings
-func (a *AliyunModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AliyunModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -412,7 +412,7 @@ func (a *AliyunModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -469,7 +469,7 @@ type aliyunRerankResponse struct {
 	} `json:"results"`
 }
 
-func (a *AliyunModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AliyunModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -507,7 +507,7 @@ func (a *AliyunModel) Rerank(modelName *string, query string, documents []string
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -542,30 +542,30 @@ func (a *AliyunModel) Rerank(modelName *string, query string, documents []string
 }
 
 // TranscribeAudio transcribe audio
-func (a *AliyunModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AliyunModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AliyunModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AliyunModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
 // AudioSpeech convert text to audio
-func (a *AliyunModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AliyunModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AliyunModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AliyunModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
 // OCRFile OCR file
-func (a *AliyunModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AliyunModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
 // ParseFile parse file
-func (a *AliyunModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AliyunModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
@@ -586,7 +586,7 @@ type AliyunModelList struct {
 	Output    AliyunModelOutput `json:"output"`
 }
 
-func (a *AliyunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AliyunModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -607,7 +607,7 @@ func (a *AliyunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -641,22 +641,22 @@ func (a *AliyunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	return ParseListModel(modelList), nil
 }
 
-func (a *AliyunModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AliyunModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AliyunModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AliyunModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
 	return nil
 }
 
-func (a *AliyunModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AliyunModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AliyunModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AliyunModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
diff --git a/internal/entity/models/aliyun_test.go b/internal/entity/models/aliyun_test.go
index dad23a9252..4d5d86c569 100644
--- a/internal/entity/models/aliyun_test.go
+++ b/internal/entity/models/aliyun_test.go
@@ -39,6 +39,7 @@ func TestAliyunChatWithMessagesSupportsToolCalls(t *testing.T) {
 		_, _ = w.Write([]byte(`{"choices":[{"message":{"content":null,"tool_calls":[{"id":"call-1","type":"function","function":{"name":"retrieval","arguments":"{\"query\":\"ragflow\"}"}}]}}]}`))
 	}))
 	defer server.Close()
+	ctx := t.Context()
 
 	model := NewAliyunModel(
 		map[string]string{"default": server.URL},
@@ -56,6 +57,7 @@ func TestAliyunChatWithMessagesSupportsToolCalls(t *testing.T) {
 	messages := []Message{{Role: "user", Content: "find ragflow"}}
 
 	response, err := model.ChatWithMessages(
+		ctx,
 		"qwen-flash",
 		messages,
 		&APIConfig{ApiKey: &apiKey},
@@ -85,6 +87,7 @@ func TestAliyunChatWithMessagesSupportsToolCalls(t *testing.T) {
 }
 
 func TestAliyunChatWithMessagesStopsQwenFlashAfterToolResult(t *testing.T) {
+	ctx := t.Context()
 	requestBody := make(chan map[string]interface{}, 1)
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		var body map[string]interface{}
@@ -121,6 +124,7 @@ func TestAliyunChatWithMessagesStopsQwenFlashAfterToolResult(t *testing.T) {
 		{Role: "tool", Content: "retrieved text", ToolCallID: "previous-call"},
 	}
 	response, err := model.ChatWithMessages(
+		ctx,
 		"qwen-flash",
 		messages,
 		&APIConfig{ApiKey: &apiKey},
@@ -184,6 +188,7 @@ func TestAliyunToolChoiceStopsOnlyQwenFlashAfterToolResult(t *testing.T) {
 }
 
 func TestAliyunChatStreamlyWithSenderSupportsToolCalls(t *testing.T) {
+	ctx := t.Context()
 	requestBody := make(chan map[string]interface{}, 1)
 	requestPath := make(chan string, 1)
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -221,6 +226,7 @@ data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ra
 	}
 	var streamed []string
 	err := model.ChatStreamlyWithSender(
+		ctx,
 		"qwen-flash",
 		[]Message{{Role: "user", Content: "find ragflow"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -276,6 +282,7 @@ data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ra
 }
 
 func TestAliyunChatStreamlyWithSenderRejectsStreamFalse(t *testing.T) {
+	ctx := t.Context()
 	model := NewAliyunModel(
 		map[string]string{"default": "https://dashscope.example"},
 		URLSuffix{Chat: "compatible-mode/v1/chat/completions"},
@@ -283,6 +290,7 @@ func TestAliyunChatStreamlyWithSenderRejectsStreamFalse(t *testing.T) {
 	apiKey := "test-key"
 	stream := false
 	err := model.ChatStreamlyWithSender(
+		ctx,
 		"qwen-flash",
 		[]Message{{Role: "user", Content: "hello"}},
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/anthropic.go b/internal/entity/models/anthropic.go
index 19159ebbf9..b754ccd6cb 100644
--- a/internal/entity/models/anthropic.go
+++ b/internal/entity/models/anthropic.go
@@ -61,7 +61,7 @@ func (a *AnthropicModel) region(apiConfig *APIConfig) string {
 	return "default"
 }
 
-func (a *AnthropicModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AnthropicModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -102,7 +102,7 @@ func (a *AnthropicModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -368,7 +368,7 @@ func parseAnthropicChatResponse(body []byte) (string, string, error) {
 	return answer.String(), reasoning.String(), nil
 }
 
-func (a *AnthropicModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AnthropicModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -386,7 +386,7 @@ func (a *AnthropicModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = strings.TrimSpace(strings.TrimSuffix(baseURL, "/"))
 	url := fmt.Sprintf("%s/%s", baseURL, strings.TrimLeft(a.baseModel.URLSuffix.Models, "/"))
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -428,55 +428,55 @@ func (a *AnthropicModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return models, nil
 }
 
-func (a *AnthropicModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AnthropicModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (a *AnthropicModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AnthropicModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AnthropicModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AnthropicModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AnthropicModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AnthropicModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, asrConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AnthropicModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, asrConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AnthropicModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AnthropicModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AnthropicModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AnthropicModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AnthropicModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AnthropicModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AnthropicModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
diff --git a/internal/entity/models/anthropic_test.go b/internal/entity/models/anthropic_test.go
index 3a4a2ff533..d26629319e 100644
--- a/internal/entity/models/anthropic_test.go
+++ b/internal/entity/models/anthropic_test.go
@@ -86,9 +86,11 @@ func TestAnthropicChatHappyPath(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
 	resp, err := newAnthropicForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"claude-sonnet-4-5-20250929",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -157,6 +159,7 @@ func TestAnthropicChatMapsSystemConfigAndImages(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
 	maxTokens := 64
@@ -164,6 +167,7 @@ func TestAnthropicChatMapsSystemConfigAndImages(t *testing.T) {
 	topP := 0.8
 	stop := []string{"END"}
 	_, err := newAnthropicForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"claude-opus-4-5-20251101",
 		[]Message{
 			{Role: "system", Content: "be concise"},
@@ -216,9 +220,11 @@ func TestAnthropicChatMapsDataImageURL(t *testing.T) {
 		})
 	})
 	defer srv.Close()
+	ctx := t.Context()
 
 	apiKey := "test-key"
 	_, err := newAnthropicForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"claude-sonnet-4-5-20250929",
 		[]Message{{Role: "user", Content: []interface{}{
 			map[string]interface{}{"type": "image_url", "image_url": map[string]interface{}{"url": "data:image/png;base64,aGVsbG8="}},
@@ -233,29 +239,31 @@ func TestAnthropicChatMapsDataImageURL(t *testing.T) {
 }
 
 func TestAnthropicChatValidationErrors(t *testing.T) {
+	ctx := t.Context()
 	m := newAnthropicForTest("http://unused")
 	apiKey := "test-key"
-	if _, err := m.ChatWithMessages("claude", []Message{{Role: "user", Content: "x"}}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := m.ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: "x"}}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("nil api config: got %v", err)
 	}
-	if _, err := m.ChatWithMessages("claude", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "messages is empty") {
+	if _, err := m.ChatWithMessages(ctx, "claude", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("empty messages: got %v", err)
 	}
-	if _, err := m.ChatWithMessages("claude", []Message{{Role: "tool", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported message role") {
+	if _, err := m.ChatWithMessages(ctx, "claude", []Message{{Role: "tool", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported message role") {
 		t.Errorf("bad role: got %v", err)
 	}
-	if _, err := m.ChatWithMessages("claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "video_url"}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported content block type") {
+	if _, err := m.ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "video_url"}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported content block type") {
 		t.Errorf("bad block: got %v", err)
 	}
-	if _, err := m.ChatWithMessages("claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "text", "text": 42}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "invalid text field") {
+	if _, err := m.ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "text", "text": 42}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "invalid text field") {
 		t.Errorf("bad text block: got %v", err)
 	}
-	if _, err := m.ChatWithMessages("claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "image"}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "image block missing source") {
+	if _, err := m.ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: []interface{}{map[string]interface{}{"type": "image"}}}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "image block missing source") {
 		t.Errorf("bad image block: got %v", err)
 	}
 }
 
 func TestAnthropicChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newAnthropicServer(t, "/v1/messages", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":{"message":"bad key"}}`))
@@ -263,26 +271,28 @@ func TestAnthropicChatRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newAnthropicForTest(srv.URL).ChatWithMessages("claude", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newAnthropicForTest(srv.URL).ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") || !strings.Contains(err.Error(), "bad key") {
 		t.Errorf("expected provider error, got %v", err)
 	}
 }
 
 func TestAnthropicChatRejectsMalformedResponse(t *testing.T) {
+	ctx := t.Context()
 	srv := newAnthropicServer(t, "/v1/messages", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"content": []map[string]interface{}{{"type": "tool_use"}}})
 	})
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newAnthropicForTest(srv.URL).ChatWithMessages("claude", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newAnthropicForTest(srv.URL).ChatWithMessages(ctx, "claude", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no text content") {
 		t.Errorf("expected no-text error, got %v", err)
 	}
 }
 
 func TestAnthropicListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	var calls int
 	srv := newAnthropicServer(t, "/v1/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		calls++
@@ -297,14 +307,14 @@ func TestAnthropicListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	m := newAnthropicForTest(srv.URL)
-	models, err := m.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "claude-sonnet-4-5-20250929,claude-haiku-4-5-20251001" {
 		t.Errorf("models=%v", models)
 	}
-	if err := m.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err = m.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 	if calls != 2 {
@@ -313,6 +323,7 @@ func TestAnthropicListModelsAndCheckConnection(t *testing.T) {
 }
 
 func TestAnthropicListModelsRejectsProviderError(t *testing.T) {
+	ctx := t.Context()
 	srv := newAnthropicServer(t, "/v1/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusForbidden)
 		_, _ = w.Write([]byte(`{"error":"forbidden"}`))
@@ -320,7 +331,7 @@ func TestAnthropicListModelsRejectsProviderError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newAnthropicForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := newAnthropicForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "403") {
 		t.Errorf("expected 403 error, got %v", err)
 	}
@@ -337,6 +348,7 @@ func TestAnthropicFactoryRegistration(t *testing.T) {
 }
 
 func TestAnthropicUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newAnthropicForTest("http://unused")
 	apiKey := "test-key"
 	modelName := "claude"
@@ -344,44 +356,44 @@ func TestAnthropicUnsupportedMethods(t *testing.T) {
 		name string
 		err  error
 	}{
-		{"stream", m.ChatStreamlyWithSender(modelName, []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })},
+		{"stream", m.ChatStreamlyWithSender(ctx, modelName, []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })},
 	}
 	for _, check := range checks {
 		if check.err == nil || !strings.Contains(check.err.Error(), "no such method") {
 			t.Errorf("%s: want no such method, got %v", check.name, check.err)
 		}
 	}
-	if _, err := m.Embed(&modelName, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Embed(ctx, &modelName, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: got %v", err)
 	}
-	if _, err := m.Rerank(&modelName, "q", []string{"d"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, &modelName, "q", []string{"d"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: got %v", err)
 	}
-	if _, err := m.Balance(&APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: got %v", err)
 	}
-	if _, err := m.TranscribeAudio(&modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: got %v", err)
 	}
-	if err := m.TranscribeAudioWithSender(&modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.TranscribeAudioWithSender(ctx, &modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudioWithSender: got %v", err)
 	}
-	if _, err := m.AudioSpeech(&modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: got %v", err)
 	}
-	if err := m.AudioSpeechWithSender(&modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.AudioSpeechWithSender(ctx, &modelName, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeechWithSender: got %v", err)
 	}
-	if _, err := m.OCRFile(&modelName, nil, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &modelName, nil, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: got %v", err)
 	}
-	if _, err := m.ParseFile(&modelName, nil, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ParseFile(ctx, &modelName, nil, &modelName, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ParseFile: got %v", err)
 	}
-	if _, err := m.ListTasks(&APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ListTasks(ctx, &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ListTasks: got %v", err)
 	}
-	if _, err := m.ShowTask("task-id", &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ShowTask(ctx, "task-id", &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ShowTask: got %v", err)
 	}
 }
diff --git a/internal/entity/models/astraflow.go b/internal/entity/models/astraflow.go
index 3b4d23db35..bc28e5ab15 100644
--- a/internal/entity/models/astraflow.go
+++ b/internal/entity/models/astraflow.go
@@ -68,7 +68,7 @@ func (a *AstraflowModel) Name() string {
 }
 
 // ChatWithMessages sends a non-streaming chat request
-func (a *AstraflowModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AstraflowModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -117,7 +117,7 @@ func (a *AstraflowModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -178,7 +178,7 @@ func (a *AstraflowModel) ChatWithMessages(modelName string, messages []Message,
 }
 
 // ChatStreamlyWithSender opens the SSE chat-completions
-func (a *AstraflowModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AstraflowModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -236,7 +236,7 @@ func (a *AstraflowModel) ChatStreamlyWithSender(modelName string, messages []Mes
 
 	// SSE is long-lived; rely on the transport's ResponseHeaderTimeout
 	// to cap connection-establishment instead of a hard deadline.
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -301,7 +301,7 @@ func (a *AstraflowModel) ChatStreamlyWithSender(modelName string, messages []Mes
 	return nil
 }
 
-func (a *AstraflowModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AstraflowModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -313,7 +313,7 @@ func (a *AstraflowModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, a.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -348,12 +348,12 @@ func (a *AstraflowModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(modelList), nil
 }
 
-func (a *AstraflowModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AstraflowModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (a *AstraflowModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AstraflowModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -430,7 +430,7 @@ func (a *AstraflowModel) Embed(modelName *string, texts []string, apiConfig *API
 	return embeddings, nil
 }
 
-func (a *AstraflowModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AstraflowModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -508,19 +508,19 @@ func (a *AstraflowModel) Rerank(modelName *string, query string, documents []str
 	return &rerankResponse, nil
 }
 
-func (a *AstraflowModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AstraflowModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AstraflowModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AstraflowModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AstraflowModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -580,22 +580,22 @@ func (a *AstraflowModel) AudioSpeech(modelName *string, audioContent *string, ap
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (a *AstraflowModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AstraflowModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AstraflowModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AstraflowModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AstraflowModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AstraflowModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AstraflowModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
diff --git a/internal/entity/models/astraflow_test.go b/internal/entity/models/astraflow_test.go
index 056a430283..3a41ac73f4 100644
--- a/internal/entity/models/astraflow_test.go
+++ b/internal/entity/models/astraflow_test.go
@@ -92,6 +92,7 @@ func TestAstraflowFactory(t *testing.T) {
 }
 
 func TestAstraflowChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "claude-opus-4-7" {
 			t.Errorf("model=%v", body["model"])
@@ -120,6 +121,7 @@ func TestAstraflowChatHappyPath(t *testing.T) {
 	mt := 64
 	temp := 0.3
 	resp, err := newAstraflowForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -138,6 +140,7 @@ func TestAstraflowChatHappyPath(t *testing.T) {
 }
 
 func TestAstraflowChatNoReasoning(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -149,6 +152,7 @@ func TestAstraflowChatNoReasoning(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newAstraflowForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -164,7 +168,9 @@ func TestAstraflowChatNoReasoning(t *testing.T) {
 }
 
 func TestAstraflowChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newAstraflowForTest("http://unused").ChatWithMessages(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil)
@@ -174,15 +180,17 @@ func TestAstraflowChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestAstraflowChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	_, err := newAstraflowForTest("http://unused").ChatWithMessages(
-		"claude-opus-4-7", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+		ctx, "claude-opus-4-7", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
 }
 
 func TestAstraflowChatPropagatesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"bad key"}`))
@@ -191,6 +199,7 @@ func TestAstraflowChatPropagatesHTTPError(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newAstraflowForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -200,6 +209,7 @@ func TestAstraflowChatPropagatesHTTPError(t *testing.T) {
 }
 
 func TestAstraflowStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowSSEServer(t, "/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"content":"Hello"}}]}`+"\n"+
@@ -212,6 +222,7 @@ func TestAstraflowStreamHappyPath(t *testing.T) {
 	var chunks []string
 	var sawDone bool
 	err := newAstraflowForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -238,6 +249,7 @@ func TestAstraflowStreamHappyPath(t *testing.T) {
 }
 
 func TestAstraflowStreamSplitsReasoning(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowSSEServer(t, "/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"reasoning_content":"step 1. "}}]}`+"\n"+
@@ -250,6 +262,7 @@ func TestAstraflowStreamSplitsReasoning(t *testing.T) {
 	apiKey := "test-key"
 	var content, reasoning []string
 	err := newAstraflowForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"kimi-k2.6",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -277,9 +290,11 @@ func TestAstraflowStreamSplitsReasoning(t *testing.T) {
 }
 
 func TestAstraflowStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
 	err := newAstraflowForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -292,8 +307,10 @@ func TestAstraflowStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestAstraflowStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newAstraflowForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
@@ -303,6 +320,7 @@ func TestAstraflowStreamRequiresSender(t *testing.T) {
 }
 
 func TestAstraflowStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowSSEServer(t, "/chat/completions",
 		`data: {"choices":[{"delta":{"content":"half"}}]}`+"\n",
 	)
@@ -310,6 +328,7 @@ func TestAstraflowStreamFailsWithoutTerminal(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newAstraflowForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -320,6 +339,7 @@ func TestAstraflowStreamFailsWithoutTerminal(t *testing.T) {
 }
 
 func TestAstraflowStreamRejectsMalformedFrame(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowSSEServer(t, "/chat/completions",
 		`data: {"choices":[{"delta":{"content":"ok"}}]}`+"\n"+
 			`data: {oops not json}`+"\n",
@@ -328,6 +348,7 @@ func TestAstraflowStreamRejectsMalformedFrame(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newAstraflowForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -338,6 +359,7 @@ func TestAstraflowStreamRejectsMalformedFrame(t *testing.T) {
 }
 
 func TestAstraflowStreamSurfacesUpstreamError(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowSSEServer(t, "/chat/completions",
 		`data: {"choices":[{"delta":{"content":"partial "}}]}`+"\n"+
 			`data: {"error":{"message":"rate limit","type":"rate_limit_error"}}`+"\n",
@@ -346,6 +368,7 @@ func TestAstraflowStreamSurfacesUpstreamError(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newAstraflowForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"claude-opus-4-7",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -359,6 +382,7 @@ func TestAstraflowStreamSurfacesUpstreamError(t *testing.T) {
 }
 
 func TestAstraflowListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -371,7 +395,7 @@ func TestAstraflowListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newAstraflowForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newAstraflowForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -382,13 +406,15 @@ func TestAstraflowListModelsHappyPath(t *testing.T) {
 }
 
 func TestAstraflowListModelsRequiresAPIKey(t *testing.T) {
-	_, err := newAstraflowForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newAstraflowForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestAstraflowCheckConnectionDelegatesToListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{{"id": "claude-opus-4-7"}},
@@ -397,12 +423,13 @@ func TestAstraflowCheckConnectionDelegatesToListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newAstraflowForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newAstraflowForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 }
 
 func TestAstraflowCheckConnectionPropagatesError(t *testing.T) {
+	ctx := t.Context()
 	srv := newAstraflowServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"bad key"}`))
@@ -410,47 +437,50 @@ func TestAstraflowCheckConnectionPropagatesError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	err := newAstraflowForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey})
+	err := newAstraflowForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestAstraflowBaseURLForRegionUnknown(t *testing.T) {
+	ctx := t.Context()
 	m := newAstraflowForTest("http://unused")
 	apiKey := "test-key"
 	region := "missing"
-	_, err := m.ListModels(&APIConfig{ApiKey: &apiKey, Region: &region})
+	_, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: &region})
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured") {
 		t.Errorf("expected base-URL error, got %v", err)
 	}
 }
 
 func TestAstraflowEmbedReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	// Embed IS implemented (not a stub). It should NOT be blocked by APIConfigCheck.
 	// With empty input texts it short-circuits to empty result (no error).
 	apiKey := "test-key"
-	embeddings, err := newAstraflowForTest("http://unused").Embed(nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	embeddings, err := newAstraflowForTest("http://unused").Embed(ctx, nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil || len(embeddings) != 0 {
 		t.Errorf("Embed: want empty result (no error), got embeddings=%v err=%v", embeddings, err)
 	}
 }
 
 func TestAstraflowAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newAstraflowForTest("http://unused")
 	model := "x"
 	apiKey := "test-key"
 	// TranscribeAudio is a stub → "no such method"
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
 	// AudioSpeech IS implemented; pass nil content to hit input validation,
 	// not api-key check (which would mean APIConfigCheck still blocks it).
-	if _, err := m.AudioSpeech(&model, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
+	if _, err := m.AudioSpeech(ctx, &model, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("AudioSpeech: expected non-api-key error, got %v", err)
 	}
 	// OCRFile is a stub → "no such method"
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
diff --git a/internal/entity/models/avian.go b/internal/entity/models/avian.go
index 282669fe8b..4818965d45 100644
--- a/internal/entity/models/avian.go
+++ b/internal/entity/models/avian.go
@@ -114,7 +114,7 @@ type avianChatResponse struct {
 	FinishReason string            `json:"finish_reason"`
 }
 
-func (a *AvianModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AvianModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -135,7 +135,7 @@ func (a *AvianModel) ChatWithMessages(modelName string, messages []Message, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -181,7 +181,7 @@ func (a *AvianModel) ChatWithMessages(modelName string, messages []Message, apiC
 	}, nil
 }
 
-func (a *AvianModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AvianModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -209,7 +209,7 @@ func (a *AvianModel) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -273,7 +273,7 @@ type avianModelInfo struct {
 	ID string `json:"id"`
 }
 
-func (a *AvianModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AvianModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -285,7 +285,7 @@ func (a *AvianModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, a.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -317,51 +317,51 @@ func (a *AvianModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return ParseListModel(modelList), nil
 }
 
-func (a *AvianModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AvianModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (a *AvianModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AvianModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AvianModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AvianModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AvianModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AvianModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AvianModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AvianModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AvianModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AvianModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AvianModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AvianModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AvianModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
diff --git a/internal/entity/models/avian_test.go b/internal/entity/models/avian_test.go
index 9322102b6d..cc1302a17a 100644
--- a/internal/entity/models/avian_test.go
+++ b/internal/entity/models/avian_test.go
@@ -79,6 +79,7 @@ func TestAvianFactory(t *testing.T) {
 }
 
 func TestAvianChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newAvianServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/chat/completions" {
 			t.Errorf("path=%s, want /v1/chat/completions", r.URL.Path)
@@ -118,6 +119,7 @@ func TestAvianChatHappyPath(t *testing.T) {
 	topP := 0.9
 	stop := []string{"END"}
 	resp, err := newAvianForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -136,6 +138,7 @@ func TestAvianChatHappyPath(t *testing.T) {
 }
 
 func TestAvianChatFallsBackToReasoningField(t *testing.T) {
+	ctx := t.Context()
 	srv := newAvianServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -150,6 +153,7 @@ func TestAvianChatFallsBackToReasoningField(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newAvianForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"moonshotai/kimi-k2.5",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -165,7 +169,9 @@ func TestAvianChatFallsBackToReasoningField(t *testing.T) {
 }
 
 func TestAvianChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newAvianForTest("http://unused").ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{},
@@ -178,8 +184,10 @@ func TestAvianChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestAvianChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	_, err := newAvianForTest("http://unused").ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{},
 		&APIConfig{ApiKey: &apiKey},
@@ -192,6 +200,7 @@ func TestAvianChatRequiresMessages(t *testing.T) {
 }
 
 func TestAvianChatPropagatesUpstreamErrorStatus(t *testing.T) {
+	ctx := t.Context()
 	srv := newAvianServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = io.WriteString(w, `{"error":"invalid key"}`)
@@ -200,6 +209,7 @@ func TestAvianChatPropagatesUpstreamErrorStatus(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newAvianForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -212,6 +222,7 @@ func TestAvianChatPropagatesUpstreamErrorStatus(t *testing.T) {
 }
 
 func TestAvianStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newAvianServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -233,6 +244,7 @@ func TestAvianStreamHappyPath(t *testing.T) {
 	var content, reasoning []string
 	var sawDone bool
 	err := newAvianForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -265,9 +277,11 @@ func TestAvianStreamHappyPath(t *testing.T) {
 }
 
 func TestAvianStreamRejectsFalseStreamConfig(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
 	err := newAvianForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -281,8 +295,10 @@ func TestAvianStreamRejectsFalseStreamConfig(t *testing.T) {
 }
 
 func TestAvianStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newAvianForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -296,6 +312,7 @@ func TestAvianStreamRequiresSender(t *testing.T) {
 }
 
 func TestAvianListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newAvianServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/models" {
 			t.Errorf("path=%s, want /v1/models", r.URL.Path)
@@ -306,22 +323,24 @@ func TestAvianListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	cfg := &APIConfig{ApiKey: &apiKey}
-	models, err := newAvianForTest(srv.URL).ListModels(cfg)
+	models, err := newAvianForTest(srv.URL).ListModels(ctx, cfg)
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "deepseek/deepseek-v3.2,moonshotai/kimi-k2.5" {
 		t.Errorf("models=%v", models)
 	}
-	if err := newAvianForTest(srv.URL).CheckConnection(cfg); err != nil {
+	if err = newAvianForTest(srv.URL).CheckConnection(ctx, cfg); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestAvianMissingBaseURLFailsClearly(t *testing.T) {
+	ctx := t.Context()
 	a := NewAvianModel(map[string]string{}, URLSuffix{Chat: "v1/chat/completions"})
 	apiKey := "test-key"
 	_, err := a.ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -334,25 +353,26 @@ func TestAvianMissingBaseURLFailsClearly(t *testing.T) {
 }
 
 func TestAvianUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	a := newAvianForTest("http://unused")
 	model := "deepseek/deepseek-v3.2"
 
-	if _, err := a.Embed(&model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: expected no such method, got %v", err)
 	}
-	if _, err := a.Rerank(&model, "q", []string{"d"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.Rerank(ctx, &model, "q", []string{"d"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: expected no such method, got %v", err)
 	}
-	if _, err := a.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected no such method, got %v", err)
 	}
-	if _, err := a.TranscribeAudio(&model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.TranscribeAudio(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: expected no such method, got %v", err)
 	}
-	if _, err := a.AudioSpeech(&model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.AudioSpeech(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: expected no such method, got %v", err)
 	}
-	if _, err := a.OCRFile(&model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := a.OCRFile(ctx, &model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: expected no such method, got %v", err)
 	}
 }
diff --git a/internal/entity/models/azure_openai.go b/internal/entity/models/azure_openai.go
index 79d277411f..55c5febcb5 100644
--- a/internal/entity/models/azure_openai.go
+++ b/internal/entity/models/azure_openai.go
@@ -61,7 +61,7 @@ func (a *AzureOpenAIModel) deploymentURL(baseURL, deployment, op string) string
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (a *AzureOpenAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (a *AzureOpenAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -115,7 +115,7 @@ func (a *AzureOpenAIModel) ChatWithMessages(modelName string, messages []Message
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -182,7 +182,7 @@ func (a *AzureOpenAIModel) ChatWithMessages(modelName string, messages []Message
 
 // ChatStreamlyWithSender sends messages and streams the response via the
 // sender function. Used for streaming chat responses with no extra channel.
-func (a *AzureOpenAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AzureOpenAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -241,7 +241,7 @@ func (a *AzureOpenAIModel) ChatStreamlyWithSender(modelName string, messages []M
 	if err != nil {
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -313,7 +313,7 @@ type azureEmbeddingResponse struct {
 }
 
 // Embed turns a list of texts into embedding vectors
-func (a *AzureOpenAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (a *AzureOpenAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -346,7 +346,7 @@ func (a *AzureOpenAIModel) Embed(modelName *string, texts []string, apiConfig *A
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -391,7 +391,7 @@ func (a *AzureOpenAIModel) Embed(modelName *string, texts []string, apiConfig *A
 }
 
 // ListModels returns the deployment names visible to the configured API key.
-func (a *AzureOpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (a *AzureOpenAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := a.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -404,7 +404,7 @@ func (a *AzureOpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 	url := fmt.Sprintf("%s/%s?api-version=%s",
 		strings.TrimRight(baseURL, "/"), a.baseModel.URLSuffix.Models, azureAPIVersion)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -442,49 +442,49 @@ func (a *AzureOpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the endpoint
-func (a *AzureOpenAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := a.ListModels(apiConfig)
+func (a *AzureOpenAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := a.ListModels(ctx, apiConfig)
 	return err
 }
 
 // Balance is not exposed by the Azure OpenAI API.
-func (a *AzureOpenAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (a *AzureOpenAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // Rerank is not exposed by the Azure OpenAI API.
-func (a *AzureOpenAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (a *AzureOpenAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (a *AzureOpenAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AzureOpenAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (a *AzureOpenAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (a *AzureOpenAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (a *AzureOpenAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (a *AzureOpenAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (a *AzureOpenAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
 
-func (a *AzureOpenAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (a *AzureOpenAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", a.Name())
 }
diff --git a/internal/entity/models/baichuan.go b/internal/entity/models/baichuan.go
index fed00bf8af..f85c1d50fe 100644
--- a/internal/entity/models/baichuan.go
+++ b/internal/entity/models/baichuan.go
@@ -49,7 +49,7 @@ func (b *BaichuanModel) Name() string {
 	return "BaiChuan"
 }
 
-func (b *BaichuanModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (b *BaichuanModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -103,7 +103,7 @@ func (b *BaichuanModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -165,7 +165,7 @@ func (b *BaichuanModel) ChatWithMessages(modelName string, messages []Message, a
 	return chatResponse, nil
 }
 
-func (b *BaichuanModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaichuanModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -223,7 +223,7 @@ func (b *BaichuanModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -290,7 +290,7 @@ func (b *BaichuanModel) ChatStreamlyWithSender(modelName string, messages []Mess
 	return sender(&endOfStream, nil)
 }
 
-func (b *BaichuanModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (b *BaichuanModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -315,7 +315,7 @@ func (b *BaichuanModel) Embed(modelName *string, texts []string, apiConfig *APIC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -367,54 +367,54 @@ func (b *BaichuanModel) Embed(modelName *string, texts []string, apiConfig *APIC
 	return embeddings, nil
 }
 
-func (b *BaichuanModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (b *BaichuanModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (b *BaichuanModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (b *BaichuanModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaichuanModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaichuanModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
 // AudioSpeech convert text to audio
-func (b *BaichuanModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (b *BaichuanModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaichuanModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaichuanModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
 // OCRFile OCR file
-func (b *BaichuanModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (b *BaichuanModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // ParseFile parse file
-func (b *BaichuanModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (b *BaichuanModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaichuanModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (b *BaichuanModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (b *BaichuanModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (b *BaichuanModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (b *BaichuanModel) CheckConnection(apiConfig *APIConfig) error {
+func (b *BaichuanModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("no such method")
 }
 
-func (b *BaichuanModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (b *BaichuanModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaichuanModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (b *BaichuanModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
diff --git a/internal/entity/models/baidu.go b/internal/entity/models/baidu.go
index 6f9da4bc63..1db8a39457 100644
--- a/internal/entity/models/baidu.go
+++ b/internal/entity/models/baidu.go
@@ -50,7 +50,7 @@ func (b *BaiduModel) Name() string {
 	return "BaiduYiyan"
 }
 
-func (b *BaiduModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (b *BaiduModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -146,7 +146,7 @@ func (b *BaiduModel) ChatWithMessages(modelName string, messages []Message, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -233,7 +233,7 @@ func (b *BaiduModel) ChatWithMessages(modelName string, messages []Message, apiC
 	return chatResponse, nil
 }
 
-func (b *BaiduModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaiduModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -334,7 +334,7 @@ func (b *BaiduModel) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -438,7 +438,7 @@ type baiduUsage struct {
 	TotalTokens  int `json:"total_tokens"`
 }
 
-func (b *BaiduModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (b *BaiduModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -468,7 +468,7 @@ func (b *BaiduModel) Embed(modelName *string, texts []string, apiConfig *APIConf
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -535,7 +535,7 @@ func (b *BaiduModel) Embed(modelName *string, texts []string, apiConfig *APIConf
 	return embeddings, nil
 }
 
-func (b *BaiduModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (b *BaiduModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -567,7 +567,7 @@ func (b *BaiduModel) Rerank(modelName *string, query string, documents []string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -617,20 +617,20 @@ func (b *BaiduModel) Rerank(modelName *string, query string, documents []string,
 }
 
 // TranscribeAudio transcribe audio
-func (b *BaiduModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (b *BaiduModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaiduModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaiduModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
 // AudioSpeech convert text to audio
-func (b *BaiduModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (b *BaiduModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaiduModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BaiduModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
@@ -646,7 +646,7 @@ type qianfanOCRResponse struct {
 	} `json:"result"`
 }
 
-func (b *BaiduModel) OCRFile(modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (b *BaiduModel) OCRFile(ctx context.Context, modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -691,7 +691,7 @@ func (b *BaiduModel) OCRFile(modelName *string, content []byte, fileURL *string,
 		return nil, fmt.Errorf("failed to marshal json payload: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -736,7 +736,7 @@ func (b *BaiduModel) OCRFile(modelName *string, content []byte, fileURL *string,
 	return &ocrResponse, nil
 }
 
-func (b *BaiduModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (b *BaiduModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -754,7 +754,7 @@ func (b *BaiduModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -789,23 +789,23 @@ func (b *BaiduModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return ParseListModel(modelList), nil
 }
 
-func (b *BaiduModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (b *BaiduModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaiduModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := b.ListModels(apiConfig)
+func (b *BaiduModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := b.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (b *BaiduModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (b *BaiduModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaiduModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (b *BaiduModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BaiduModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (b *BaiduModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
diff --git a/internal/entity/models/bedrock.go b/internal/entity/models/bedrock.go
index fc2371102a..6b42401368 100644
--- a/internal/entity/models/bedrock.go
+++ b/internal/entity/models/bedrock.go
@@ -505,7 +505,7 @@ func signBedrockRequest(ctx context.Context, req *http.Request, body []byte, cre
 // the joined assistant answer. ReasonContent is always non-nil per the
 // driver contract; Bedrock surfaces no reasoning channel today, so it
 // is left empty rather than nil.
-func (b *BedrockModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (b *BedrockModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -531,7 +531,7 @@ func (b *BedrockModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("bedrock: marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	creds, err := resolveBedrockCredentials(ctx, key, region)
@@ -589,7 +589,7 @@ func (b *BedrockModel) ChatWithMessages(modelName string, messages []Message, ap
 // payload. For chat we only need messageStart, contentBlockDelta,
 // messageStop, and (for error propagation) exception frames; other
 // events are ignored.
-func (b *BedrockModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BedrockModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -625,7 +625,6 @@ func (b *BedrockModel) ChatStreamlyWithSender(modelName string, messages []Messa
 	// Background context: event streams are long-lived so we attach
 	// no overall deadline. ResponseHeaderTimeout (set in the
 	// constructor) still caps connection setup.
-	ctx := context.Background()
 	creds, err := resolveBedrockCredentials(ctx, key, region)
 	if err != nil {
 		return err
@@ -767,7 +766,7 @@ type bedrockListModelsResponse struct {
 // configured credentials. The control plane lives at
 // bedrock.{region}.amazonaws.com (not bedrock-runtime), signs against
 // the "bedrock" service, and is GET-only.
-func (b *BedrockModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (b *BedrockModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := b.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -781,7 +780,7 @@ func (b *BedrockModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	creds, err := resolveBedrockCredentials(ctx, key, region)
@@ -836,8 +835,8 @@ func (b *BedrockModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 // CheckConnection delegates to ListModels: a successful catalog query
 // proves credentials, region, and network reachability in one round
 // trip without burning a chat completion.
-func (b *BedrockModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := b.ListModels(apiConfig)
+func (b *BedrockModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := b.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -864,7 +863,7 @@ type bedrockCohereEmbeddingResponse struct {
 // InvokeModel. Titan's embedding API accepts one inputText per call,
 // while Cohere accepts a texts batch and returns vectors in input
 // order.
-func (b *BedrockModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (b *BedrockModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if len(texts) == 0 {
 		return []EmbeddingData{}, nil
 	}
@@ -885,7 +884,7 @@ func (b *BedrockModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	creds, err := resolveBedrockCredentials(ctx, key, region)
@@ -1026,52 +1025,52 @@ func decodeCohereEmbeddingVectors(raw json.RawMessage) ([][]float64, error) {
 }
 
 // Rerank is not exposed by Bedrock.
-func (b *BedrockModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (b *BedrockModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // Balance is not exposed by Bedrock.
-func (b *BedrockModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (b *BedrockModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // TranscribeAudio is not exposed by Bedrock. Speech-to-text on AWS
 // lives in Amazon Transcribe, a separate service.
-func (b *BedrockModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (b *BedrockModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BedrockModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BedrockModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
 // AudioSpeech is not exposed by Bedrock. Text-to-speech on AWS lives
 // in Amazon Polly, a separate service.
-func (b *BedrockModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (b *BedrockModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
-func (b *BedrockModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BedrockModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", b.Name())
 }
 
 // OCRFile is not exposed by Bedrock. OCR on AWS lives in Amazon
 // Textract, a separate service.
-func (b *BedrockModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (b *BedrockModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // ParseFile is not exposed by Bedrock.
-func (b *BedrockModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (b *BedrockModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // ListTasks is not exposed by Bedrock through the Converse API.
-func (b *BedrockModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (b *BedrockModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
 
 // ShowTask is not exposed by Bedrock through the Converse API.
-func (b *BedrockModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (b *BedrockModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", b.Name())
 }
diff --git a/internal/entity/models/bedrock_test.go b/internal/entity/models/bedrock_test.go
index 156d41ec22..9687e1dec0 100644
--- a/internal/entity/models/bedrock_test.go
+++ b/internal/entity/models/bedrock_test.go
@@ -327,6 +327,7 @@ func newBedrockServer(t *testing.T, wantMethod, wantPath string, handler http.Ha
 }
 
 func TestBedrockChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/anthropic.claude-3-haiku-20240307-v1:0/converse",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -350,7 +351,7 @@ func TestBedrockChatHappyPath(t *testing.T) {
 
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	resp, err := m.ChatWithMessages("anthropic.claude-3-haiku-20240307-v1:0",
+	resp, err := m.ChatWithMessages(ctx, "anthropic.claude-3-haiku-20240307-v1:0",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &key}, nil, nil)
 	if err != nil {
@@ -368,17 +369,19 @@ func TestBedrockChatHappyPath(t *testing.T) {
 }
 
 func TestBedrockChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
-	_, err := m.ChatWithMessages("m", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "m", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("want api-key error, got %v", err)
 	}
 }
 
 func TestBedrockChatRequiresModelID(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	key := validBedrockKey()
-	_, err := m.ChatWithMessages("", []Message{{Role: "user", Content: "x"}},
+	_, err := m.ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model id is required") {
 		t.Errorf("want model-required error, got %v", err)
@@ -386,6 +389,7 @@ func TestBedrockChatRequiresModelID(t *testing.T) {
 }
 
 func TestBedrockChatPropagatesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/m/converse",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -396,7 +400,7 @@ func TestBedrockChatPropagatesHTTPError(t *testing.T) {
 
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	_, err := m.ChatWithMessages("m",
+	_, err := m.ChatWithMessages(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
@@ -408,6 +412,7 @@ func TestBedrockChatPropagatesHTTPError(t *testing.T) {
 }
 
 func TestBedrockListModelsParsesCatalog(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodGet,
 		"/foundation-models",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -424,7 +429,7 @@ func TestBedrockListModelsParsesCatalog(t *testing.T) {
 
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	got, err := m.ListModels(&APIConfig{ApiKey: &key})
+	got, err := m.ListModels(ctx, &APIConfig{ApiKey: &key})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -445,6 +450,7 @@ func TestBedrockListModelsParsesCatalog(t *testing.T) {
 }
 
 func TestBedrockCheckConnectionDelegates(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodGet,
 		"/foundation-models",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -453,7 +459,7 @@ func TestBedrockCheckConnectionDelegates(t *testing.T) {
 	defer srv.Close()
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	if err := m.CheckConnection(&APIConfig{ApiKey: &key}); err == nil || !strings.Contains(err.Error(), "403") {
+	if err := m.CheckConnection(ctx, &APIConfig{ApiKey: &key}); err == nil || !strings.Contains(err.Error(), "403") {
 		t.Errorf("want 403 surfaced via ListModels, got %v", err)
 	}
 }
@@ -497,6 +503,7 @@ func encodeBedrockEventFrames(t *testing.T, events []struct {
 }
 
 func TestBedrockStreamDecodesContentDeltas(t *testing.T) {
+	ctx := t.Context()
 	frames := encodeBedrockEventFrames(t, []struct {
 		eventType   string
 		messageType string
@@ -520,7 +527,7 @@ func TestBedrockStreamDecodesContentDeltas(t *testing.T) {
 	key := validBedrockKey()
 	var chunks []string
 	sawDone := false
-	err := m.ChatStreamlyWithSender("m",
+	err := m.ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &key}, nil, nil,
 		func(c *string, _ *string) error {
@@ -546,6 +553,7 @@ func TestBedrockStreamDecodesContentDeltas(t *testing.T) {
 }
 
 func TestBedrockStreamSurfacesException(t *testing.T) {
+	ctx := t.Context()
 	frames := encodeBedrockEventFrames(t, []struct {
 		eventType   string
 		messageType string
@@ -568,7 +576,7 @@ func TestBedrockStreamSurfacesException(t *testing.T) {
 
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	err := m.ChatStreamlyWithSender("m",
+	err := m.ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -578,6 +586,7 @@ func TestBedrockStreamSurfacesException(t *testing.T) {
 }
 
 func TestBedrockStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	// Connection closed cleanly after a delta but before messageStop.
 	// This used to be silently treated as success, masking truncated
 	// answers; the driver must now surface a "stream ended before
@@ -599,7 +608,7 @@ func TestBedrockStreamFailsWithoutTerminal(t *testing.T) {
 
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
-	err := m.ChatStreamlyWithSender("m",
+	err := m.ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -609,10 +618,11 @@ func TestBedrockStreamFailsWithoutTerminal(t *testing.T) {
 }
 
 func TestBedrockStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	key := validBedrockKey()
 	stream := false
-	err := m.ChatStreamlyWithSender("m",
+	err := m.ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key},
 		&ChatConfig{Stream: &stream},
@@ -624,9 +634,10 @@ func TestBedrockStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestBedrockStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	key := validBedrockKey()
-	err := m.ChatStreamlyWithSender("m",
+	err := m.ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -648,6 +659,7 @@ func TestLookupBedrockEventHeader(t *testing.T) {
 }
 
 func TestBedrockTitanEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	var seenInputs []string
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/amazon.titan-embed-text-v2:0/invoke",
@@ -674,7 +686,7 @@ func TestBedrockTitanEmbedHappyPath(t *testing.T) {
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
 	model := "amazon.titan-embed-text-v2:0"
-	got, err := m.Embed(&model, []string{"alpha", "beta"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 256}, nil)
+	got, err := m.Embed(ctx, &model, []string{"alpha", "beta"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 256}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -690,6 +702,7 @@ func TestBedrockTitanEmbedHappyPath(t *testing.T) {
 }
 
 func TestBedrockTitanV1OmitsDimension(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/amazon.titan-embed-text-v1/invoke",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -704,12 +717,13 @@ func TestBedrockTitanV1OmitsDimension(t *testing.T) {
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
 	model := "amazon.titan-embed-text-v1"
-	if _, err := m.Embed(&model, []string{"alpha"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 256}, nil); err != nil {
+	if _, err := m.Embed(ctx, &model, []string{"alpha"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 256}, nil); err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 }
 
 func TestBedrockCohereEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/cohere.embed-english-v3/invoke",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -736,7 +750,7 @@ func TestBedrockCohereEmbedHappyPath(t *testing.T) {
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
 	model := "cohere.embed-english-v3"
-	got, err := m.Embed(&model, []string{"first", "second"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 128}, nil)
+	got, err := m.Embed(ctx, &model, []string{"first", "second"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 128}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -746,6 +760,7 @@ func TestBedrockCohereEmbedHappyPath(t *testing.T) {
 }
 
 func TestBedrockCohereV4ForwardsDimensionAndParsesTypedResponse(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/cohere.embed-v4:0/invoke",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -765,7 +780,7 @@ func TestBedrockCohereV4ForwardsDimensionAndParsesTypedResponse(t *testing.T) {
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
 	model := "cohere.embed-v4:0"
-	got, err := m.Embed(&model, []string{"first"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 512}, nil)
+	got, err := m.Embed(ctx, &model, []string{"first"}, &APIConfig{ApiKey: &key}, &EmbeddingConfig{Dimension: 512}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -775,8 +790,9 @@ func TestBedrockCohereV4ForwardsDimensionAndParsesTypedResponse(t *testing.T) {
 }
 
 func TestBedrockEmbedShortCircuitsEmptyInput(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
-	got, err := m.Embed(nil, nil, nil, nil, nil)
+	got, err := m.Embed(ctx, nil, nil, nil, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed empty: %v", err)
 	}
@@ -786,28 +802,31 @@ func TestBedrockEmbedShortCircuitsEmptyInput(t *testing.T) {
 }
 
 func TestBedrockEmbedRequiresAPIKeyAndModel(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	model := "x"
-	if _, err := m.Embed(&model, []string{"a"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("Embed: want api-key error, got %v", err)
 	}
 	key := validBedrockKey()
 	blank := " "
-	if _, err := m.Embed(&blank, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := m.Embed(ctx, &blank, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("Embed: want model-required error, got %v", err)
 	}
 }
 
 func TestBedrockEmbedRejectsUnsupportedModel(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	key := validBedrockKey()
 	model := "anthropic.claude-3-haiku-20240307-v1:0"
-	if _, err := m.Embed(&model, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported embedding model") {
+	if _, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "unsupported embedding model") {
 		t.Errorf("Embed: want unsupported-model error, got %v", err)
 	}
 }
 
 func TestBedrockEmbedPropagatesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newBedrockServer(t, http.MethodPost,
 		"/model/amazon.titan-embed-text-v2:0/invoke",
 		func(w http.ResponseWriter, r *http.Request) {
@@ -819,45 +838,48 @@ func TestBedrockEmbedPropagatesHTTPError(t *testing.T) {
 	m := newBedrockForTest(srv.URL)
 	key := validBedrockKey()
 	model := "amazon.titan-embed-text-v2:0"
-	if _, err := m.Embed(&model, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "400") || !strings.Contains(err.Error(), "bad input") {
+	if _, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &key}, nil, nil); err == nil || !strings.Contains(err.Error(), "400") || !strings.Contains(err.Error(), "bad input") {
 		t.Errorf("Embed: want HTTP error with body, got %v", err)
 	}
 }
 
 func TestBedrockRerankReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	model := "x"
-	if _, err := m.Rerank(&model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: want no-such-method, got %v", err)
 	}
 }
 
 func TestBedrockBalanceReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
-	if _, err := m.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: want no-such-method, got %v", err)
 	}
 }
 
 func TestBedrockAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newBedrockForTest("http://unused")
 	model := "x"
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: want no-such-method, got %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: want no-such-method, got %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: want no-such-method, got %v", err)
 	}
-	if _, err := m.ParseFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ParseFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ParseFile: want no-such-method, got %v", err)
 	}
-	if _, err := m.ListTasks(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ListTasks(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ListTasks: want no-such-method, got %v", err)
 	}
-	if _, err := m.ShowTask("t", &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ShowTask(ctx, "t", &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ShowTask: want no-such-method, got %v", err)
 	}
 }
diff --git a/internal/entity/models/builtin.go b/internal/entity/models/builtin.go
index bb4c8fcf20..cdefc00794 100644
--- a/internal/entity/models/builtin.go
+++ b/internal/entity/models/builtin.go
@@ -2,6 +2,7 @@ package models
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -47,16 +48,16 @@ func (b *BuiltinModel) NewInstance(baseURL map[string]string) ModelDriver {
 	}
 }
 
-func (b *BuiltinModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (b *BuiltinModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support chat")
 }
 
-func (b *BuiltinModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BuiltinModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("builtin model does not support chat")
 }
 
 // Embed sends texts to a TEI (Text Embeddings Inference) server and returns embeddings
-func (b *BuiltinModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (b *BuiltinModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if len(texts) == 0 {
 		return []EmbeddingData{}, nil
 	}
@@ -122,35 +123,35 @@ func (b *BuiltinModel) Embed(modelName *string, texts []string, apiConfig *APICo
 	return result, nil
 }
 
-func (b *BuiltinModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (b *BuiltinModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support rerank")
 }
 
-func (b *BuiltinModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (b *BuiltinModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support transcription")
 }
 
-func (b *BuiltinModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BuiltinModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("builtin model does not support transcription")
 }
 
-func (b *BuiltinModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (b *BuiltinModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support TTS")
 }
 
-func (b *BuiltinModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (b *BuiltinModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("builtin model does not support TTS")
 }
 
-func (b *BuiltinModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (b *BuiltinModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support OCR")
 }
 
-func (b *BuiltinModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (b *BuiltinModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support parse file")
 }
 
-func (b *BuiltinModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (b *BuiltinModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return []ListModelResponse{
 		{
 			Name: b.model,
@@ -158,21 +159,21 @@ func (b *BuiltinModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	}, nil
 }
 
-func (b *BuiltinModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (b *BuiltinModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("builtin model does not support balance")
 }
 
-func (b *BuiltinModel) CheckConnection(apiConfig *APIConfig) error {
+func (b *BuiltinModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	// Try to get model info to verify connection
-	_, err := b.Embed(nil, []string{"test"}, apiConfig, nil, nil)
+	_, err := b.Embed(ctx, nil, []string{"test"}, apiConfig, nil, nil)
 	return err
 }
 
-func (b *BuiltinModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (b *BuiltinModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("builtin model does not support tasks")
 }
 
-func (b *BuiltinModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (b *BuiltinModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("builtin model does not support tasks")
 }
 
diff --git a/internal/entity/models/chat_tools.go b/internal/entity/models/chat_tools.go
index b49af5b2e0..933cf3f806 100644
--- a/internal/entity/models/chat_tools.go
+++ b/internal/entity/models/chat_tools.go
@@ -129,7 +129,7 @@ func runToolLoop(ctx context.Context, cm *ChatModel, history []Message, toolsLis
 		tcChoice := "auto"
 		cfg.ToolChoice = &tcChoice
 
-		resp, err := cm.ModelDriver.ChatWithMessages(*cm.ModelName, history, cm.APIConfig, &cfg, nil)
+		resp, err := cm.ModelDriver.ChatWithMessages(ctx, *cm.ModelName, history, cm.APIConfig, &cfg, nil)
 		if err != nil {
 			return "", totalTokens, fmt.Errorf("round %d: %w", round, err)
 		}
@@ -163,7 +163,7 @@ func runToolLoop(ctx context.Context, cm *ChatModel, history []Message, toolsLis
 		Content: fmt.Sprintf("Exceed max rounds: %d", maxRounds),
 	})
 	cfg := *chatCfg
-	resp, err := cm.ModelDriver.ChatWithMessages(*cm.ModelName, history, cm.APIConfig, &cfg, nil)
+	resp, err := cm.ModelDriver.ChatWithMessages(ctx, *cm.ModelName, history, cm.APIConfig, &cfg, nil)
 	if err != nil {
 		return "", totalTokens, fmt.Errorf("final call: %w", err)
 	}
@@ -278,7 +278,7 @@ func runStreamToolLoop(ctx context.Context, cm *ChatModel, history []Message, to
 		var pendingThinkClose bool
 		var roundTokens int
 
-		err := cm.ModelDriver.ChatStreamlyWithSender(*cm.ModelName, history, cm.APIConfig, &cfg, nil, func(delta *string, reason *string) error {
+		err := cm.ModelDriver.ChatStreamlyWithSender(ctx, *cm.ModelName, history, cm.APIConfig, &cfg, nil, func(delta *string, reason *string) error {
 			if reason != nil && *reason != "" {
 				if !reasoningStarted {
 					reasoningStarted = true
@@ -351,7 +351,7 @@ func runStreamToolLoop(ctx context.Context, cm *ChatModel, history []Message, to
 	var exceedUsage TokenUsage
 	cfg.UsageResult = &exceedUsage
 	var exceedTokens int
-	err := cm.ModelDriver.ChatStreamlyWithSender(*cm.ModelName, history, cm.APIConfig, &cfg, nil, func(delta *string, reason *string) error {
+	err := cm.ModelDriver.ChatStreamlyWithSender(ctx, *cm.ModelName, history, cm.APIConfig, &cfg, nil, func(delta *string, reason *string) error {
 		if delta != nil && *delta != "" && *delta != "[DONE]" {
 			exceedTokens += tokenizer.NumTokensFromString(*delta)
 		}
diff --git a/internal/entity/models/cohere.go b/internal/entity/models/cohere.go
index 5f4877ebe1..5b963285d4 100644
--- a/internal/entity/models/cohere.go
+++ b/internal/entity/models/cohere.go
@@ -53,7 +53,7 @@ func (c *CoHereModel) Name() string {
 	return "Cohere"
 }
 
-func (c *CoHereModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (c *CoHereModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -119,7 +119,7 @@ func (c *CoHereModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -186,7 +186,7 @@ func (c *CoHereModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (c *CoHereModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CoHereModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -263,7 +263,7 @@ func (c *CoHereModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -337,7 +337,7 @@ func (c *CoHereModel) ChatStreamlyWithSender(modelName string, messages []Messag
 	return sender(&endOfStream, nil)
 }
 
-func (c *CoHereModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (c *CoHereModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -370,7 +370,7 @@ func (c *CoHereModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -421,7 +421,7 @@ func (c *CoHereModel) Embed(modelName *string, texts []string, apiConfig *APICon
 	return embeddings, nil
 }
 
-func (c *CoHereModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (c *CoHereModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -455,7 +455,7 @@ func (c *CoHereModel) Rerank(modelName *string, query string, documents []string
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -506,7 +506,7 @@ func (c *CoHereModel) Rerank(modelName *string, query string, documents []string
 }
 
 // TranscribeAudio transcribe audio
-func (c *CoHereModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (c *CoHereModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -584,7 +584,7 @@ func (c *CoHereModel) TranscribeAudio(modelName *string, file *string, apiConfig
 	}
 
 	// build request
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -621,30 +621,30 @@ func (c *CoHereModel) TranscribeAudio(modelName *string, file *string, apiConfig
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (c *CoHereModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CoHereModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", c.Name())
 }
 
 // AudioSpeech convert text to audio
-func (c *CoHereModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (c *CoHereModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CoHereModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CoHereModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", c.Name())
 }
 
 // OCRFile OCR file
-func (c *CoHereModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (c *CoHereModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
 // ParseFile parse file
-func (c *CoHereModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (c *CoHereModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CoHereModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (c *CoHereModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -655,7 +655,7 @@ func (c *CoHereModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, c.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -704,19 +704,19 @@ func (c *CoHereModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	return ParseListModel(ModelList{Models: models}), nil
 }
 
-func (c *CoHereModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (c *CoHereModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CoHereModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := c.ListModels(apiConfig)
+func (c *CoHereModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := c.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (c *CoHereModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (c *CoHereModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CoHereModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (c *CoHereModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
diff --git a/internal/entity/models/cometapi.go b/internal/entity/models/cometapi.go
index 5eccfa1169..992bc5e7ae 100644
--- a/internal/entity/models/cometapi.go
+++ b/internal/entity/models/cometapi.go
@@ -235,7 +235,7 @@ type cometapiModelCatalogItem struct {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (c *CometAPIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (c *CometAPIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -255,7 +255,7 @@ func (c *CometAPIModel) ChatWithMessages(modelName string, messages []Message, a
 
 	reqBody := buildCometAPIChatRequest(modelName, messages, false, chatModelConfig)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newCometAPIJSONRequest(ctx, "POST", url, reqBody, apiKey)
@@ -274,7 +274,7 @@ func (c *CometAPIModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (c *CometAPIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CometAPIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -309,7 +309,7 @@ func (c *CometAPIModel) ChatStreamlyWithSender(modelName string, messages []Mess
 	}
 	reqBody := buildCometAPIChatRequest(modelName, messages, true, chatModelConfig)
 
-	req, err := newCometAPIJSONRequest(context.Background(), "POST", url, reqBody, apiKey)
+	req, err := newCometAPIJSONRequest(ctx, "POST", url, reqBody, apiKey)
 	if err != nil {
 		return err
 	}
@@ -384,7 +384,7 @@ type cometapiEmbeddingRequest struct {
 }
 
 // Embed turns a list of texts into embedding vectors
-func (c *CometAPIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (c *CometAPIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -412,7 +412,7 @@ func (c *CometAPIModel) Embed(modelName *string, texts []string, apiConfig *APIC
 		reqBody.Dimensions = embeddingConfig.Dimension
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newCometAPIJSONRequest(ctx, "POST", url, reqBody, apiKey)
@@ -459,13 +459,13 @@ func (c *CometAPIModel) Embed(modelName *string, texts []string, apiConfig *APIC
 }
 
 // ListModels returns the public CometAPI model catalog.
-func (c *CometAPIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (c *CometAPIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	url, err := c.endpointURL(cometapiRegion(apiConfig), c.baseModel.URLSuffix.Models)
 	if err != nil {
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -491,7 +491,7 @@ func (c *CometAPIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 }
 
 // Balance queries CometAPI's quota service.
-func (c *CometAPIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (c *CometAPIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -499,7 +499,7 @@ func (c *CometAPIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 		return nil, fmt.Errorf("balance URL is required")
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", c.balanceURL(*apiConfig.ApiKey), nil)
@@ -525,8 +525,8 @@ func (c *CometAPIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 }
 
 // CheckConnection runs a quota query to verify the API key.
-func (c *CometAPIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := c.Balance(apiConfig)
+func (c *CometAPIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := c.Balance(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -534,12 +534,12 @@ func (c *CometAPIModel) CheckConnection(apiConfig *APIConfig) error {
 }
 
 // Rerank calculates similarity scores between query and documents.
-func (c *CometAPIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (c *CometAPIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (c *CometAPIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (c *CometAPIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -654,12 +654,12 @@ func (c *CometAPIModel) TranscribeAudio(modelName *string, file *string, apiConf
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (c *CometAPIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CometAPIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", c.Name())
 }
 
 // AudioSpeech synthesizes speech audio from text.
-func (c *CometAPIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (c *CometAPIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := c.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -719,23 +719,23 @@ func (c *CometAPIModel) AudioSpeech(modelName *string, audioContent *string, api
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (c *CometAPIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (c *CometAPIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", c.Name())
 }
 
 // OCRFile OCR file
-func (c *CometAPIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (c *CometAPIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CometAPIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (c *CometAPIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CometAPIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (c *CometAPIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
 
-func (c *CometAPIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (c *CometAPIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", c.Name())
 }
diff --git a/internal/entity/models/cometapi_test.go b/internal/entity/models/cometapi_test.go
index 24f0f6343c..1a93719b17 100644
--- a/internal/entity/models/cometapi_test.go
+++ b/internal/entity/models/cometapi_test.go
@@ -77,6 +77,7 @@ func TestCometAPIFactoryRoute(t *testing.T) {
 }
 
 func TestCometAPIChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "gpt-5" {
 			t.Errorf("expected model=gpt-5, got %v", body["model"])
@@ -99,7 +100,7 @@ func TestCometAPIChatHappyPath(t *testing.T) {
 
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("gpt-5", []Message{
+	resp, err := m.ChatWithMessages(ctx, "gpt-5", []Message{
 		{Role: "user", Content: "ping"},
 	}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -114,6 +115,7 @@ func TestCometAPIChatHappyPath(t *testing.T) {
 }
 
 func TestCometAPIChatPropagatesConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["max_tokens"] != float64(64) {
 			t.Errorf("max_tokens=%v want 64", body["max_tokens"])
@@ -140,7 +142,7 @@ func TestCometAPIChatPropagatesConfig(t *testing.T) {
 	temp := 0.3
 	topP := 0.9
 	stop := []string{"END"}
-	_, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "ping"}},
+	_, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{MaxTokens: &mt, Temperature: &temp, TopP: &topP, Stop: &stop},
 		nil)
@@ -150,6 +152,7 @@ func TestCometAPIChatPropagatesConfig(t *testing.T) {
 }
 
 func TestCometAPIChatReturnsReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{
@@ -161,7 +164,7 @@ func TestCometAPIChatReturnsReasoningContent(t *testing.T) {
 
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("ChatWithMessages: %v", err)
 	}
@@ -171,41 +174,45 @@ func TestCometAPIChatReturnsReasoningContent(t *testing.T) {
 }
 
 func TestCometAPIChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
-	_, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 	emptyKey := ""
-	_, err = m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+	_, err = m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("empty key: expected api-key error, got %v", err)
 	}
 }
 
 func TestCometAPIChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
-	err = m.ChatStreamlyWithSender(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })
+	err = m.ChatStreamlyWithSender(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("stream: expected model-name error, got %v", err)
 	}
 }
 
 func TestCometAPIChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("gpt-5", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "gpt-5", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
 }
 
 func TestCometAPIChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -214,13 +221,14 @@ func TestCometAPIChatRejectsHTTPError(t *testing.T) {
 
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestCometAPIChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
+	ctx := t.Context()
 	// Empty *Region pointer must fall back to the "default" entry, not
 	// be treated as an explicit "" region (which would miss the lookup).
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -233,7 +241,7 @@ func TestCometAPIChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	emptyRegion := ""
-	_, err := m.ChatWithMessages("gpt-5",
+	_, err := m.ChatWithMessages(ctx, "gpt-5",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &emptyRegion}, nil, nil)
 	if err != nil {
@@ -242,6 +250,7 @@ func TestCometAPIChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 }
 
 func TestCometAPIListModelsFallsBackToDefaultOnEmptyRegion(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/api/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"data": []map[string]interface{}{{"id": "x"}}})
 	})
@@ -250,15 +259,16 @@ func TestCometAPIListModelsFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	emptyRegion := ""
-	if _, err := m.ListModels(&APIConfig{ApiKey: &apiKey, Region: &emptyRegion}); err != nil {
+	if _, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: &emptyRegion}); err != nil {
 		t.Errorf("empty Region: expected fallback to default, got %v", err)
 	}
 }
 
 func TestCometAPIStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("gpt-5",
+	err := m.ChatStreamlyWithSender(ctx, "gpt-5",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -267,10 +277,11 @@ func TestCometAPIStreamRequiresSender(t *testing.T) {
 }
 
 func TestCometAPIChatRejectsUnknownRegion(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
 	region := "eu"
-	_, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}},
+	_, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &region}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured for region") {
 		t.Errorf("expected region error, got %v", err)
@@ -278,6 +289,7 @@ func TestCometAPIChatRejectsUnknownRegion(t *testing.T) {
 }
 
 func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
+	ctx := t.Context()
 	tests := []struct {
 		name string
 		path string
@@ -287,7 +299,7 @@ func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
 			name: "Chat",
 			path: "/v1/chat/completions",
 			run: func(m *CometAPIModel, apiConfig *APIConfig) error {
-				_, err := m.ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, apiConfig, nil, nil)
+				_, err := m.ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, apiConfig, nil, nil)
 				return err
 			},
 		},
@@ -295,7 +307,7 @@ func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
 			name: "Stream",
 			path: "/v1/chat/completions",
 			run: func(m *CometAPIModel, apiConfig *APIConfig) error {
-				return m.ChatStreamlyWithSender("gpt-5", []Message{{Role: "user", Content: "x"}}, apiConfig, nil, nil, func(*string, *string) error { return nil })
+				return m.ChatStreamlyWithSender(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, apiConfig, nil, nil, func(*string, *string) error { return nil })
 			},
 		},
 		{
@@ -303,7 +315,7 @@ func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
 			path: "/v1/embeddings",
 			run: func(m *CometAPIModel, apiConfig *APIConfig) error {
 				model := "text-embedding-3-small"
-				_, err := m.Embed(&model, []string{"x"}, apiConfig, nil, nil)
+				_, err := m.Embed(ctx, &model, []string{"x"}, apiConfig, nil, nil)
 				return err
 			},
 		},
@@ -311,7 +323,7 @@ func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
 			name: "ListModels",
 			path: "/api/models",
 			run: func(m *CometAPIModel, apiConfig *APIConfig) error {
-				_, err := m.ListModels(apiConfig)
+				_, err := m.ListModels(ctx, apiConfig)
 				return err
 			},
 		},
@@ -347,6 +359,7 @@ func TestCometAPIBaseURLNormalizesSlashes(t *testing.T) {
 }
 
 func TestCometAPIStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["stream"] != true {
 			t.Errorf("expected stream=true, got %v", body["stream"])
@@ -367,7 +380,7 @@ func TestCometAPIStreamHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone int32
-	err := m.ChatStreamlyWithSender("gpt-5",
+	err := m.ChatStreamlyWithSender(ctx, "gpt-5",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(content *string, _ *string) error {
@@ -394,10 +407,11 @@ func TestCometAPIStreamHappyPath(t *testing.T) {
 }
 
 func TestCometAPIStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
 	stream := false
-	err := m.ChatStreamlyWithSender("gpt-5",
+	err := m.ChatStreamlyWithSender(ctx, "gpt-5",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Stream: &stream},
@@ -410,6 +424,7 @@ func TestCometAPIStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestCometAPIStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	// Body closes before [DONE] or a finish_reason -> driver must complain
 	// instead of pretending the stream finished cleanly.
 	srv := newCometAPIServer(t, "/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
@@ -420,7 +435,7 @@ func TestCometAPIStreamFailsWithoutTerminal(t *testing.T) {
 
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("gpt-5",
+	err := m.ChatStreamlyWithSender(ctx, "gpt-5",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(*string, *string) error { return nil },
@@ -431,6 +446,7 @@ func TestCometAPIStreamFailsWithoutTerminal(t *testing.T) {
 }
 
 func TestCometAPIListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/api/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -444,7 +460,7 @@ func TestCometAPIListModelsHappyPath(t *testing.T) {
 
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
-	ids, err := m.ListModels(&APIConfig{ApiKey: &apiKey})
+	ids, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -454,13 +470,14 @@ func TestCometAPIListModelsHappyPath(t *testing.T) {
 }
 
 func TestCometAPIListModelsAllowsNilAPIConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/api/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"data": []map[string]interface{}{{"id": "gpt-5"}}})
 	})
 	defer srv.Close()
 
 	m := newCometAPIForTest(srv.URL)
-	ids, err := m.ListModels(nil)
+	ids, err := m.ListModels(ctx, nil)
 	if err != nil {
 		t.Fatalf("ListModels(nil): %v", err)
 	}
@@ -470,6 +487,7 @@ func TestCometAPIListModelsAllowsNilAPIConfig(t *testing.T) {
 }
 
 func TestCometAPICheckConnectionDelegatesToBalance(t *testing.T) {
+	ctx := t.Context()
 	// 200 -> CheckConnection succeeds; 401 -> CheckConnection propagates.
 	okSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/user/quota" {
@@ -489,17 +507,18 @@ func TestCometAPICheckConnectionDelegatesToBalance(t *testing.T) {
 	apiKey := "test-key"
 	mOK := newCometAPIForTest(okSrv.URL)
 	mOK.baseModel.URLSuffix.Balance = okSrv.URL + "/user/quota"
-	if err := mOK.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := mOK.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection(ok): %v", err)
 	}
 	mFail := newCometAPIForTest(failSrv.URL)
 	mFail.baseModel.URLSuffix.Balance = failSrv.URL + "/user/quota"
-	if err := mFail.CheckConnection(&APIConfig{ApiKey: &apiKey}); err == nil {
+	if err := mFail.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 		t.Error("CheckConnection(fail): expected error, got nil")
 	}
 }
 
 func TestCometAPIBalanceHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/user/quota" {
 			t.Errorf("path=%s want /user/quota", r.URL.Path)
@@ -522,7 +541,7 @@ func TestCometAPIBalanceHappyPath(t *testing.T) {
 	m := newCometAPIForTest("http://unused")
 	m.baseModel.URLSuffix.Balance = srv.URL + "/user/quota"
 	apiKey := "test-key"
-	balance, err := m.Balance(&APIConfig{ApiKey: &apiKey})
+	balance, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("Balance: %v", err)
 	}
@@ -532,33 +551,37 @@ func TestCometAPIBalanceHappyPath(t *testing.T) {
 }
 
 func TestCometAPIBalanceRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
-	_, err := m.Balance(&APIConfig{})
+	_, err := m.Balance(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("Balance: expected api-key error, got %v", err)
 	}
 }
 
 func TestCometAPIBalanceRequiresConfiguredURL(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	m.baseModel.URLSuffix.Balance = ""
 	apiKey := "test-key"
-	_, err := m.Balance(&APIConfig{ApiKey: &apiKey})
+	_, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "balance URL is required") {
 		t.Errorf("Balance: expected balance URL error, got %v", err)
 	}
 }
 
 func TestCometAPIRerankReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	q := "gpt-5"
-	_, err := m.Rerank(&q, "what is rag?", []string{"a", "b"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
+	_, err := m.Rerank(ctx, &q, "what is rag?", []string{"a", "b"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: expected 'no such method', got %v", err)
 	}
 }
 
 func TestCometAPIEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "text-embedding-3-small" {
 			t.Errorf("model=%v want text-embedding-3-small", body["model"])
@@ -583,7 +606,7 @@ func TestCometAPIEmbedHappyPath(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	vecs, err := m.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, &EmbeddingConfig{Dimension: 256}, nil)
+	vecs, err := m.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, &EmbeddingConfig{Dimension: 256}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -596,6 +619,7 @@ func TestCometAPIEmbedHappyPath(t *testing.T) {
 }
 
 func TestCometAPIEmbedReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	// Upstream returns the three vectors in shuffled order. The driver
 	// must reorder them so the slot at position i corresponds to input i.
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -612,7 +636,7 @@ func TestCometAPIEmbedReordersByIndex(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	vecs, err := m.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := m.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -624,6 +648,7 @@ func TestCometAPIEmbedReordersByIndex(t *testing.T) {
 }
 
 func TestCometAPIEmbedEmptyInputShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	// Empty input must NOT make an HTTP call; the test fails the request
 	// rather than the assertion if it does.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
@@ -635,7 +660,7 @@ func TestCometAPIEmbedEmptyInputShortCircuits(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	vecs, err := m.Embed(&model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := m.Embed(ctx, &model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed([]): %v", err)
 	}
@@ -645,29 +670,32 @@ func TestCometAPIEmbedEmptyInputShortCircuits(t *testing.T) {
 }
 
 func TestCometAPIEmbedRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	model := "text-embedding-3-small"
-	_, err := m.Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestCometAPIEmbedRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	m := newCometAPIForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.Embed(nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 	empty := ""
-	_, err = m.Embed(&empty, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err = m.Embed(ctx, &empty, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("empty model: expected model-name error, got %v", err)
 	}
 }
 
 func TestCometAPIEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	// A malformed upstream that repeats data[*].index would silently
 	// overwrite the earlier vector; the driver must fail loudly instead.
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -683,13 +711,14 @@ func TestCometAPIEmbedRejectsDuplicateIndex(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index 0") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
 }
 
 func TestCometAPIEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -702,13 +731,14 @@ func TestCometAPIEmbedRejectsOutOfRangeIndex(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestCometAPIEmbedRejectsMissingSlot(t *testing.T) {
+	ctx := t.Context()
 	// Upstream returns only one of the two requested embeddings.
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
@@ -722,13 +752,14 @@ func TestCometAPIEmbedRejectsMissingSlot(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding for input index 1") {
 		t.Errorf("expected missing-embedding error for slot 1, got %v", err)
 	}
 }
 
 func TestCometAPIEmbedRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newCometAPIServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -738,7 +769,7 @@ func TestCometAPIEmbedRejectsHTTPError(t *testing.T) {
 	m := newCometAPIForTest(srv.URL)
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	_, err := m.Embed(&model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "CometAPI embeddings API error") {
 		t.Errorf("expected CometAPI embeddings API error, got %v", err)
 	}
diff --git a/internal/entity/models/deepinfra.go b/internal/entity/models/deepinfra.go
index 60c827f7d3..fb5554b735 100644
--- a/internal/entity/models/deepinfra.go
+++ b/internal/entity/models/deepinfra.go
@@ -55,7 +55,7 @@ func (d *DeepInfraModel) Name() string {
 	return "deepinfra"
 }
 
-func (d *DeepInfraModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (d *DeepInfraModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -125,7 +125,7 @@ func (d *DeepInfraModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -188,7 +188,7 @@ func (d *DeepInfraModel) ChatWithMessages(modelName string, messages []Message,
 	return chatResponse, nil
 }
 
-func (d *DeepInfraModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepInfraModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -265,7 +265,7 @@ func (d *DeepInfraModel) ChatStreamlyWithSender(modelName string, messages []Mes
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -338,7 +338,7 @@ func (d *DeepInfraModel) ChatStreamlyWithSender(modelName string, messages []Mes
 	return sender(&endOfStream, nil)
 }
 
-func (d *DeepInfraModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (d *DeepInfraModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -367,7 +367,7 @@ func (d *DeepInfraModel) Embed(modelName *string, texts []string, apiConfig *API
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -421,7 +421,7 @@ type deepinfraRerankResponse struct {
 }
 
 // Rerank scores documents against a query using DeepInfra's inference endpoint.
-func (d *DeepInfraModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (d *DeepInfraModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -452,7 +452,7 @@ func (d *DeepInfraModel) Rerank(modelName *string, query string, documents []str
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -510,7 +510,7 @@ func (d *DeepInfraModel) Rerank(modelName *string, query string, documents []str
 	return &RerankResponse{Data: results}, nil
 }
 
-func (d *DeepInfraModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (d *DeepInfraModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -586,7 +586,7 @@ func (d *DeepInfraModel) TranscribeAudio(modelName *string, file *string, apiCon
 		return nil, fmt.Errorf("failed to close multipart writer: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -627,11 +627,11 @@ func (d *DeepInfraModel) TranscribeAudio(modelName *string, file *string, apiCon
 	}, nil
 }
 
-func (d *DeepInfraModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepInfraModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", d.Name())
 }
 
-func (d *DeepInfraModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (d *DeepInfraModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -679,7 +679,7 @@ func (d *DeepInfraModel) AudioSpeech(modelName *string, audioContent *string, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -708,7 +708,7 @@ func (d *DeepInfraModel) AudioSpeech(modelName *string, audioContent *string, ap
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (d *DeepInfraModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepInfraModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -757,7 +757,7 @@ func (d *DeepInfraModel) AudioSpeechWithSender(modelName *string, audioContent *
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -806,15 +806,15 @@ func (d *DeepInfraModel) AudioSpeechWithSender(modelName *string, audioContent *
 	return nil
 }
 
-func (d *DeepInfraModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (d *DeepInfraModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", d.Name())
 }
 
-func (d *DeepInfraModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (d *DeepInfraModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", d.Name())
 }
 
-func (d *DeepInfraModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (d *DeepInfraModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 
 	resolvedBaseURL, err := d.baseModel.GetBaseURL(apiConfig)
 	if err != nil {
@@ -829,7 +829,7 @@ func (d *DeepInfraModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -875,7 +875,7 @@ func (d *DeepInfraModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(ModelList{Models: models}), nil
 }
 
-func (d *DeepInfraModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (d *DeepInfraModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -886,7 +886,7 @@ func (d *DeepInfraModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 	}
 	url := fmt.Sprintf("%s/%s", baseURL, d.baseModel.URLSuffix.Balance)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -926,15 +926,15 @@ func (d *DeepInfraModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 	}, nil
 }
 
-func (d *DeepInfraModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := d.ListModels(apiConfig)
+func (d *DeepInfraModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := d.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (d *DeepInfraModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (d *DeepInfraModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", d.Name())
 }
 
-func (d *DeepInfraModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (d *DeepInfraModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", d.Name())
 }
diff --git a/internal/entity/models/deepinfra_test.go b/internal/entity/models/deepinfra_test.go
index 1c70c04450..354d53443c 100644
--- a/internal/entity/models/deepinfra_test.go
+++ b/internal/entity/models/deepinfra_test.go
@@ -23,6 +23,7 @@ func newDeepInfraForTest(baseURL string) *DeepInfraModel {
 
 // TestDeepInfraRerankHappyPath verifies request shape and score mapping.
 func TestDeepInfraRerankHappyPath(t *testing.T) {
+	ctx := t.Context()
 	const modelPath = "/v1/inference/Qwen/Qwen3-Reranker-4B"
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != modelPath {
@@ -59,6 +60,7 @@ func TestDeepInfraRerankHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	model := "Qwen/Qwen3-Reranker-4B"
 	resp, err := newDeepInfraForTest(srv.URL).Rerank(
+		ctx,
 		&model,
 		"capital of France?",
 		[]string{"Paris is the capital.", "Berlin is the capital."},
@@ -77,6 +79,7 @@ func TestDeepInfraRerankHappyPath(t *testing.T) {
 
 // TestDeepInfraRerankNoTopNLimit returns every scored document when TopN is unset.
 func TestDeepInfraRerankNoTopNLimit(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"scores": []float64{0.9, 0.1},
@@ -87,6 +90,7 @@ func TestDeepInfraRerankNoTopNLimit(t *testing.T) {
 	apiKey := "test-key"
 	model := "Qwen/Qwen3-Reranker-4B"
 	resp, err := newDeepInfraForTest(srv.URL).Rerank(
+		ctx,
 		&model,
 		"capital of France?",
 		[]string{"Paris is the capital.", "Berlin is the capital."},
@@ -106,9 +110,10 @@ func TestDeepInfraRerankNoTopNLimit(t *testing.T) {
 
 // TestDeepInfraRerankEmptyDocuments returns an empty result without calling the API.
 func TestDeepInfraRerankEmptyDocuments(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := "Qwen/Qwen3-Reranker-4B"
-	resp, err := newDeepInfraForTest("http://unused").Rerank(&model, "q", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newDeepInfraForTest("http://unused").Rerank(ctx, &model, "q", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
@@ -119,8 +124,9 @@ func TestDeepInfraRerankEmptyDocuments(t *testing.T) {
 
 // TestDeepInfraRerankRequiresAPIKey rejects requests without an API key.
 func TestDeepInfraRerankRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := "Qwen/Qwen3-Reranker-4B"
-	_, err := newDeepInfraForTest("http://unused").Rerank(&model, "q", []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := newDeepInfraForTest("http://unused").Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
@@ -128,6 +134,7 @@ func TestDeepInfraRerankRequiresAPIKey(t *testing.T) {
 
 // TestDeepInfraRerankRejectsScoreCountMismatch errors when scores length mismatches documents.
 func TestDeepInfraRerankRejectsScoreCountMismatch(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"scores": []float64{0.5}})
 	}))
@@ -136,7 +143,7 @@ func TestDeepInfraRerankRejectsScoreCountMismatch(t *testing.T) {
 	apiKey := "test-key"
 	model := "cross-encoder/ms-marco-MiniLM-L-12-v2"
 	_, err := newDeepInfraForTest(srv.URL).Rerank(
-		&model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+		ctx, &model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "expected 2 scores") {
 		t.Errorf("expected score-count error, got %v", err)
 	}
diff --git a/internal/entity/models/deepseek.go b/internal/entity/models/deepseek.go
index 7d54adcefa..3ec8e61ed8 100644
--- a/internal/entity/models/deepseek.go
+++ b/internal/entity/models/deepseek.go
@@ -52,7 +52,7 @@ func (d *DeepSeekModel) Name() string {
 	return "deepseek"
 }
 
-func (d *DeepSeekModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (d *DeepSeekModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -149,7 +149,7 @@ func (d *DeepSeekModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -234,7 +234,7 @@ func (d *DeepSeekModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (d *DeepSeekModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepSeekModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -336,7 +336,7 @@ func (d *DeepSeekModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -417,11 +417,11 @@ func (d *DeepSeekModel) ChatStreamlyWithSender(modelName string, messages []Mess
 }
 
 // Embed embeds a list of texts into embeddings
-func (d *DeepSeekModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (d *DeepSeekModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DeepSeekModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (d *DeepSeekModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -440,7 +440,7 @@ func (d *DeepSeekModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -492,7 +492,7 @@ type deepseekBalanceResponse struct {
 // calling GET /user/balance with the configured Bearer token.
 // The result map matches the shape used by the Moonshot driver,
 // so the UI can render it without provider-specific code.
-func (d *DeepSeekModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (d *DeepSeekModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := d.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -504,7 +504,7 @@ func (d *DeepSeekModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), d.baseModel.URLSuffix.Balance)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -552,8 +552,8 @@ func (d *DeepSeekModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 	}, nil
 }
 
-func (d *DeepSeekModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := d.ListModels(apiConfig)
+func (d *DeepSeekModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := d.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -561,42 +561,42 @@ func (d *DeepSeekModel) CheckConnection(apiConfig *APIConfig) error {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (d *DeepSeekModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (d *DeepSeekModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", d.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (d *DeepSeekModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (d *DeepSeekModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DeepSeekModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepSeekModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", d.Name())
 }
 
 // AudioSpeech convert text to audio
-func (d *DeepSeekModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (d *DeepSeekModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DeepSeekModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DeepSeekModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", d.Name())
 }
 
 // OCRFile OCR file
-func (d *DeepSeekModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (d *DeepSeekModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
 // ParseFile parse file
-func (d *DeepSeekModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (d *DeepSeekModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DeepSeekModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (d *DeepSeekModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DeepSeekModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (d *DeepSeekModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
diff --git a/internal/entity/models/deepseek_test.go b/internal/entity/models/deepseek_test.go
index aab13f72c6..bab79a2ea7 100644
--- a/internal/entity/models/deepseek_test.go
+++ b/internal/entity/models/deepseek_test.go
@@ -18,6 +18,7 @@ func newDeepSeekForTest(baseURL string) *DeepSeekModel {
 }
 
 func TestDeepSeekChatWithMessagesSupportsToolCalls(t *testing.T) {
+	ctx := t.Context()
 	var requestBody map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/chat/completions" {
@@ -62,6 +63,7 @@ func TestDeepSeekChatWithMessagesSupportsToolCalls(t *testing.T) {
 	apiKey := "test-key"
 	toolChoice := "auto"
 	resp, err := newDeepSeekForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek-chat",
 		[]Message{
 			{Role: "user", Content: "what is marigold"},
@@ -106,6 +108,7 @@ func TestDeepSeekChatWithMessagesSupportsToolCalls(t *testing.T) {
 }
 
 func TestDeepSeekChatWithMessagesForwardsToolHistory(t *testing.T) {
+	ctx := t.Context()
 	var requestBody map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if err := json.NewDecoder(r.Body).Decode(&requestBody); err != nil {
@@ -123,6 +126,7 @@ func TestDeepSeekChatWithMessagesForwardsToolHistory(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newDeepSeekForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek-chat",
 		[]Message{
 			{
diff --git a/internal/entity/models/dummy.go b/internal/entity/models/dummy.go
index 86a123ae20..4cc08b57ea 100644
--- a/internal/entity/models/dummy.go
+++ b/internal/entity/models/dummy.go
@@ -17,6 +17,7 @@
 package models
 
 import (
+	"context"
 	"fmt"
 	"ragflow/internal/common"
 )
@@ -45,69 +46,69 @@ func (d *DummyModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (d *DummyModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (d *DummyModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (d *DummyModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DummyModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("not implemented")
 }
 
 // Embed embeds a list of texts into embeddings
-func (d *DummyModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (d *DummyModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
-func (d *DummyModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (d *DummyModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
-func (d *DummyModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (d *DummyModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (d *DummyModel) CheckConnection(apiConfig *APIConfig) error {
+func (d *DummyModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("no such method")
 }
 
 // Rerank calculates similarity scores between query and documents
-func (d *DummyModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (d *DummyModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", d.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (d *DummyModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (d *DummyModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DummyModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DummyModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", d.Name())
 }
 
 // AudioSpeech convert text to audio
-func (d *DummyModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (d *DummyModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DummyModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *DummyModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", d.Name())
 }
 
 // OCRFile OCR file
-func (d *DummyModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (d *DummyModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
 // ParseFile parse file
-func (d *DummyModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (d *DummyModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DummyModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (d *DummyModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
 
-func (d *DummyModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (d *DummyModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", d.Name())
 }
diff --git a/internal/entity/models/fishaudio.go b/internal/entity/models/fishaudio.go
index df05c7c1ec..3710da496d 100644
--- a/internal/entity/models/fishaudio.go
+++ b/internal/entity/models/fishaudio.go
@@ -54,24 +54,24 @@ func (f *FishAudioModel) Name() string {
 	return "Fish Audio"
 }
 
-func (f *FishAudioModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (f *FishAudioModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FishAudioModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FishAudioModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FishAudioModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (f *FishAudioModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (f *FishAudioModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (f *FishAudioModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (f *FishAudioModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (f *FishAudioModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -137,7 +137,7 @@ func (f *FishAudioModel) TranscribeAudio(modelName *string, file *string, apiCon
 	}
 
 	// request
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -177,12 +177,12 @@ func (f *FishAudioModel) TranscribeAudio(modelName *string, file *string, apiCon
 	}, nil
 }
 
-func (f *FishAudioModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FishAudioModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", f.Name())
 }
 
 // AudioSpeech convert text to audio
-func (f *FishAudioModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (f *FishAudioModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -215,7 +215,7 @@ func (f *FishAudioModel) AudioSpeech(modelName *string, audioContent *string, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -245,7 +245,7 @@ func (f *FishAudioModel) AudioSpeech(modelName *string, audioContent *string, ap
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (f *FishAudioModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FishAudioModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -279,7 +279,7 @@ func (f *FishAudioModel) AudioSpeechWithSender(modelName *string, audioContent *
 	}
 
 	// Build Request
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -326,16 +326,16 @@ func (f *FishAudioModel) AudioSpeechWithSender(modelName *string, audioContent *
 }
 
 // OCRFile OCR file
-func (f *FishAudioModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (f *FishAudioModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // ParseFile parse file
-func (f *FishAudioModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (f *FishAudioModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FishAudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (f *FishAudioModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -346,7 +346,7 @@ func (f *FishAudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, f.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -397,7 +397,7 @@ func (f *FishAudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(modelList), nil
 }
 
-func (f *FishAudioModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (f *FishAudioModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -409,7 +409,7 @@ func (f *FishAudioModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 
 	url := fmt.Sprintf("%s/wallet/self/api-credit", strings.TrimSuffix(baseURL, "/"))
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -442,15 +442,15 @@ func (f *FishAudioModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 	return result, nil
 }
 
-func (f *FishAudioModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := f.ListModels(apiConfig)
+func (f *FishAudioModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := f.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (f *FishAudioModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (f *FishAudioModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FishAudioModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (f *FishAudioModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
diff --git a/internal/entity/models/fishaudio_test.go b/internal/entity/models/fishaudio_test.go
index 123f2d303f..18763327b7 100644
--- a/internal/entity/models/fishaudio_test.go
+++ b/internal/entity/models/fishaudio_test.go
@@ -28,6 +28,7 @@ func newFishAudioForListModelsTest(baseURL string) *FishAudioModel {
 }
 
 func TestFishAudioListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -44,7 +45,7 @@ func TestFishAudioListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newFishAudioForListModelsTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newFishAudioForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -61,12 +62,14 @@ func TestFishAudioListModels(t *testing.T) {
 }
 
 func TestFishAudioListModelsRequiresAPIKey(t *testing.T) {
-	if _, err := newFishAudioForListModelsTest("http://unused").ListModels(&APIConfig{}); err == nil {
+	ctx := t.Context()
+	if _, err := newFishAudioForListModelsTest("http://unused").ListModels(ctx, &APIConfig{}); err == nil {
 		t.Fatal("ListModels: expected error for missing api key, got nil")
 	}
 }
 
 func TestFishAudioListModelsRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = io.WriteString(w, `{"error":"unauthorized"}`)
@@ -74,7 +77,7 @@ func TestFishAudioListModelsRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "bad-key"
-	if _, err := newFishAudioForListModelsTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey}); err == nil {
+	if _, err := newFishAudioForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 		t.Fatal("ListModels: expected error for HTTP 401, got nil")
 	}
 }
diff --git a/internal/entity/models/funasr.go b/internal/entity/models/funasr.go
index be81620b01..e6c7e805c7 100644
--- a/internal/entity/models/funasr.go
+++ b/internal/entity/models/funasr.go
@@ -53,23 +53,23 @@ func (f *FunASR) Name() string {
 	return "funasr"
 }
 
-func (f *FunASR) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (f *FunASR) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FunASR) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (f *FunASR) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (f *FunASR) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (f *FunASR) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -147,7 +147,7 @@ func (f *FunASR) TranscribeAudio(modelName *string, file *string, apiConfig *API
 	}
 
 	// build request
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -186,27 +186,27 @@ func (f *FunASR) TranscribeAudio(modelName *string, file *string, apiConfig *API
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (f *FunASR) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FunASR) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (f *FunASR) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FunASR) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (f *FunASR) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (f *FunASR) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (f *FunASR) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -218,7 +218,7 @@ func (f *FunASR) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, f.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -258,19 +258,19 @@ func (f *FunASR) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return models, nil
 }
 
-func (f *FunASR) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (f *FunASR) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) CheckConnection(apiConfig *APIConfig) error {
-	_, err := f.ListModels(apiConfig)
+func (f *FunASR) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := f.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (f *FunASR) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (f *FunASR) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
 
-func (f *FunASR) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (f *FunASR) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", f.Name())
 }
diff --git a/internal/entity/models/futurmix.go b/internal/entity/models/futurmix.go
index c46b9ade77..5805400bfa 100644
--- a/internal/entity/models/futurmix.go
+++ b/internal/entity/models/futurmix.go
@@ -146,7 +146,7 @@ type futurmixChatResponse struct {
 }
 
 // ChatWithMessages sends a non-streaming chat completion
-func (f *FuturMixModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (f *FuturMixModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -165,7 +165,7 @@ func (f *FuturMixModel) ChatWithMessages(modelName string, messages []Message, a
 
 	reqBody := buildFuturMixChatRequest(modelName, messages, false, chatModelConfig)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newFuturMixJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
@@ -208,7 +208,7 @@ func (f *FuturMixModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender sends a streaming chat completion
-func (f *FuturMixModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FuturMixModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := f.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -235,7 +235,7 @@ func (f *FuturMixModel) ChatStreamlyWithSender(modelName string, messages []Mess
 
 	reqBody := buildFuturMixChatRequest(modelName, messages, true, chatModelConfig)
 
-	req, err := newFuturMixJSONRequest(context.Background(), "POST", endpoint, reqBody, apiKey)
+	req, err := newFuturMixJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
 	if err != nil {
 		return err
 	}
@@ -286,64 +286,64 @@ func (f *FuturMixModel) ChatStreamlyWithSender(modelName string, messages []Mess
 }
 
 // Embed is not exposed by the FuturMix API per the public docs.
-func (f *FuturMixModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (f *FuturMixModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // Rerank is not exposed by the FuturMix API per the public docs.
-func (f *FuturMixModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (f *FuturMixModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // ListModels is not documented as a public endpoint by FuturMix.
-func (f *FuturMixModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (f *FuturMixModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // CheckConnection is not exposed by the FuturMix API.
-func (f *FuturMixModel) CheckConnection(apiConfig *APIConfig) error {
+func (f *FuturMixModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s, no such method", f.Name())
 }
 
 // Balance is not exposed by the FuturMix public API.
-func (f *FuturMixModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (f *FuturMixModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // TranscribeAudio is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (f *FuturMixModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FuturMixModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FuturMixModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", f.Name())
 }
 
 // AudioSpeech is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (f *FuturMixModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
-func (f *FuturMixModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (f *FuturMixModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", f.Name())
 }
 
 // OCRFile is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (f *FuturMixModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // ParseFile is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (f *FuturMixModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // ListTasks is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (f *FuturMixModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
 
 // ShowTask is not exposed by the FuturMix API per the docs.
-func (f *FuturMixModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (f *FuturMixModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", f.Name())
 }
diff --git a/internal/entity/models/gitee.go b/internal/entity/models/gitee.go
index 58fa8d588b..c9c5400e8a 100644
--- a/internal/entity/models/gitee.go
+++ b/internal/entity/models/gitee.go
@@ -55,7 +55,7 @@ func (g *GiteeModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (g *GiteeModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (g *GiteeModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -140,7 +140,7 @@ func (g *GiteeModel) ChatWithMessages(modelName string, messages []Message, apiC
 
 	common.Info(fmt.Sprintf("GiteeAPI request body: %s", string(jsonData)))
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -232,7 +232,7 @@ func (g *GiteeModel) ChatWithMessages(modelName string, messages []Message, apiC
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (g *GiteeModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GiteeModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -316,7 +316,7 @@ func (g *GiteeModel) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -502,7 +502,7 @@ type giteeUsage struct {
 }
 
 // Embed embeds a list of texts into embeddings
-func (g *GiteeModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (g *GiteeModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -536,7 +536,7 @@ func (g *GiteeModel) Embed(modelName *string, texts []string, apiConfig *APIConf
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -587,7 +587,7 @@ type giteeRerankRequest struct {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (g *GiteeModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (g *GiteeModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -626,7 +626,7 @@ func (g *GiteeModel) Rerank(modelName *string, query string, documents []string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -661,20 +661,20 @@ func (g *GiteeModel) Rerank(modelName *string, query string, documents []string,
 }
 
 // TranscribeAudio transcribe audio
-func (g *GiteeModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (g *GiteeModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GiteeModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GiteeModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
 // AudioSpeech convert text to audio
-func (g *GiteeModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (g *GiteeModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GiteeModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GiteeModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
@@ -684,7 +684,7 @@ type giteeOCRResponse struct {
 }
 
 // OCRFile OCR file
-func (g *GiteeModel) OCRFile(modelName *string, content []byte, imageURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (g *GiteeModel) OCRFile(ctx context.Context, modelName *string, content []byte, imageURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -730,7 +730,7 @@ func (g *GiteeModel) OCRFile(modelName *string, content []byte, imageURL *string
 
 	writer.Close()
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, payload)
@@ -781,7 +781,7 @@ type giteeURLs struct {
 }
 
 // ParseFile parse file
-func (g *GiteeModel) ParseFile(modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (g *GiteeModel) ParseFile(ctx context.Context, modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -827,7 +827,7 @@ func (g *GiteeModel) ParseFile(modelName *string, content []byte, documentURL *s
 
 	writer.Close()
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, payload)
@@ -858,7 +858,7 @@ func (g *GiteeModel) ParseFile(modelName *string, content []byte, documentURL *s
 		return nil, fmt.Errorf("failed to parse response: %w", err)
 	}
 
-	_, err = g.getParseFile(&baseURL, apiConfig.ApiKey, &giteeParseFileResp.TaskID, 5*time.Second, 10)
+	_, err = g.getParseFile(ctx, &baseURL, apiConfig.ApiKey, &giteeParseFileResp.TaskID, 5*time.Second, 10)
 	if err != nil {
 		return nil, err
 	}
@@ -871,7 +871,7 @@ func (g *GiteeModel) ParseFile(modelName *string, content []byte, documentURL *s
 type giteeGetParseFileResponse struct {
 }
 
-func (g *GiteeModel) getParseFile(baseURL *string, apiKey, taskID *string, timeOut time.Duration, count int) (*giteeGetParseFileResponse, error) {
+func (g *GiteeModel) getParseFile(ctx context.Context, baseURL *string, apiKey, taskID *string, timeOut time.Duration, count int) (*giteeGetParseFileResponse, error) {
 	url := fmt.Sprintf("%s/task/%s/status", strings.TrimSuffix(*baseURL, "/"), *taskID)
 
 	reqBody := map[string]interface{}{}
@@ -881,7 +881,7 @@ func (g *GiteeModel) getParseFile(baseURL *string, apiKey, taskID *string, timeO
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	for i := 0; i < count; i++ {
@@ -927,7 +927,7 @@ func (g *GiteeModel) getParseFile(baseURL *string, apiKey, taskID *string, timeO
 	return nil, nil
 }
 
-func (g *GiteeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (g *GiteeModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 
 	resolvedBaseURL, err := g.baseModel.GetBaseURL(apiConfig)
 	if err != nil {
@@ -943,7 +943,7 @@ func (g *GiteeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -977,7 +977,7 @@ func (g *GiteeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return ParseListModel(modelList), nil
 }
 
-func (g *GiteeModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (g *GiteeModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -996,7 +996,7 @@ func (g *GiteeModel) Balance(apiConfig *APIConfig) (map[string]interface{}, erro
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -1038,7 +1038,7 @@ func (g *GiteeModel) Balance(apiConfig *APIConfig) (map[string]interface{}, erro
 	return response, nil
 }
 
-func (g *GiteeModel) CheckConnection(apiConfig *APIConfig) error {
+func (g *GiteeModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -1057,7 +1057,7 @@ func (g *GiteeModel) CheckConnection(apiConfig *APIConfig) error {
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -1116,7 +1116,7 @@ type giteeTaskURLs struct {
 	Cancel string `json:"cancel"`
 }
 
-func (g *GiteeModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (g *GiteeModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -1135,7 +1135,7 @@ func (g *GiteeModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -1177,7 +1177,7 @@ func (g *GiteeModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return taskListResp, nil
 }
 
-func (g *GiteeModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (g *GiteeModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -1196,7 +1196,7 @@ func (g *GiteeModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskRespons
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
diff --git a/internal/entity/models/gitee_test.go b/internal/entity/models/gitee_test.go
index e8e96acfbf..6dfeb62620 100644
--- a/internal/entity/models/gitee_test.go
+++ b/internal/entity/models/gitee_test.go
@@ -76,6 +76,7 @@ func deepSeekAliasModelsForTest(t *testing.T) map[string]Model {
 }
 
 func TestGiteeListModelsMapsAllDeepSeekAliasesToModelMetadata(t *testing.T) {
+	ctx := t.Context()
 	initProviderManagerWithGiteeForTest(t)
 	aliasModels := deepSeekAliasModelsForTest(t)
 	aliases := make([]string, 0, len(aliasModels))
@@ -111,7 +112,7 @@ func TestGiteeListModelsMapsAllDeepSeekAliasesToModelMetadata(t *testing.T) {
 	}))
 	defer srv.Close()
 
-	models, err := newGiteeForListModelsTest(srv.URL).ListModels(&APIConfig{})
+	models, err := newGiteeForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -152,6 +153,7 @@ func TestGiteeListModelsMapsAllDeepSeekAliasesToModelMetadata(t *testing.T) {
 }
 
 func TestGiteeListModelsKeepsOwnedBySuffixAfterAliasMetadataLookup(t *testing.T) {
+	ctx := t.Context()
 	initProviderManagerWithGiteeForTest(t)
 
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -159,7 +161,7 @@ func TestGiteeListModelsKeepsOwnedBySuffixAfterAliasMetadataLookup(t *testing.T)
 	}))
 	defer srv.Close()
 
-	models, err := newGiteeForListModelsTest(srv.URL).ListModels(&APIConfig{})
+	models, err := newGiteeForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -182,6 +184,7 @@ func TestGiteeListModelsKeepsOwnedBySuffixAfterAliasMetadataLookup(t *testing.T)
 }
 
 func TestGiteeListModelsIntegration(t *testing.T) {
+	ctx := t.Context()
 	if common.GetEnv(common.EnvGiteeListModelsIntegration) != "1" {
 		t.Skip("set GITEE_LIST_MODELS_INTEGRATION=1 to call the real Gitee models endpoint")
 	}
@@ -197,7 +200,7 @@ func TestGiteeListModelsIntegration(t *testing.T) {
 		apiConfig.ApiKey = &apiKey
 	}
 
-	models, err := newGiteeForListModelsTest(baseURL).ListModels(apiConfig)
+	models, err := newGiteeForListModelsTest(baseURL).ListModels(ctx, apiConfig)
 	if err != nil {
 		t.Fatalf("real Gitee ListModels: %v", err)
 	}
diff --git a/internal/entity/models/google.go b/internal/entity/models/google.go
index 171f096ddf..eb01866175 100644
--- a/internal/entity/models/google.go
+++ b/internal/entity/models/google.go
@@ -119,7 +119,7 @@ func (g *GoogleModel) baseURL(apiConfig *APIConfig) string {
 	return strings.TrimSpace(baseURL)
 }
 
-func (g *GoogleModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (g *GoogleModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -131,7 +131,6 @@ func (g *GoogleModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("messages is empty")
 	}
 
-	ctx := context.Background()
 	client, err := genai.NewClient(ctx, g.clientConfig(strings.TrimSpace(*apiConfig.ApiKey), apiConfig))
 	if err != nil {
 		return nil, err
@@ -193,7 +192,7 @@ func (g *GoogleModel) ChatWithMessages(modelName string, messages []Message, api
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (g *GoogleModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GoogleModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -208,7 +207,6 @@ func (g *GoogleModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("sender is nil")
 	}
 
-	ctx := context.Background()
 	client, err := genai.NewClient(ctx, g.clientConfig(strings.TrimSpace(*apiConfig.ApiKey), apiConfig))
 	if err != nil {
 		return err
@@ -298,7 +296,7 @@ func (g *GoogleModel) ChatStreamlyWithSender(modelName string, messages []Messag
 
 // Embed generates embeddings for a batch of texts using the Gemini embeddings API.
 // The SDK routes to batchEmbedContents internally, so all texts are sent in one request.
-func (g *GoogleModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (g *GoogleModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -309,7 +307,7 @@ func (g *GoogleModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("texts is empty")
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	client, err := genai.NewClient(ctx, g.clientConfig(strings.TrimSpace(*apiConfig.ApiKey), apiConfig))
@@ -352,60 +350,60 @@ func (g *GoogleModel) Embed(modelName *string, texts []string, apiConfig *APICon
 	return result, nil
 }
 
-func (g *GoogleModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (g *GoogleModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
 
-	return googleListModels(context.Background(), g.clientConfig(strings.TrimSpace(*apiConfig.ApiKey), apiConfig))
+	return googleListModels(ctx, g.clientConfig(strings.TrimSpace(*apiConfig.ApiKey), apiConfig))
 }
 
-func (g *GoogleModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (g *GoogleModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (g *GoogleModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := g.ListModels(apiConfig)
+func (g *GoogleModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := g.ListModels(ctx, apiConfig)
 	return err
 }
 
 // Rerank calculates similarity scores between query and documents
-func (g *GoogleModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (g *GoogleModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", g.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (g *GoogleModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (g *GoogleModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GoogleModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GoogleModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
 // AudioSpeech convert text to audio
-func (g *GoogleModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (g *GoogleModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GoogleModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GoogleModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
 // OCRFile OCR file
-func (g *GoogleModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (g *GoogleModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
 // ParseFile parse file
-func (g *GoogleModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (g *GoogleModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GoogleModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (g *GoogleModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GoogleModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (g *GoogleModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
diff --git a/internal/entity/models/google_test.go b/internal/entity/models/google_test.go
index f19cbb1ce1..d441ff0b30 100644
--- a/internal/entity/models/google_test.go
+++ b/internal/entity/models/google_test.go
@@ -26,6 +26,7 @@ func withGoogleListModelsStub(t *testing.T, fn func(context.Context, *genai.Clie
 }
 
 func TestGoogleModelListModelsRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	cases := []struct {
 		name      string
@@ -61,7 +62,7 @@ func TestGoogleModelListModelsRequiresAPIKey(t *testing.T) {
 
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
-			models, err := model.ListModels(tc.apiConfig)
+			models, err := model.ListModels(ctx, tc.apiConfig)
 			if err == nil {
 				t.Fatal("expected an API key error")
 			}
@@ -80,6 +81,7 @@ func TestGoogleModelListModelsRequiresAPIKey(t *testing.T) {
 }
 
 func TestGoogleModelListModelsReturnsModelNames(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	apiKey := "test-api-key"
 	configuredAPIKey := "  " + apiKey + "  "
@@ -92,7 +94,7 @@ func TestGoogleModelListModelsReturnsModelNames(t *testing.T) {
 		return expected, nil
 	})
 
-	models, err := model.ListModels(&APIConfig{ApiKey: &configuredAPIKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &configuredAPIKey})
 	if err != nil {
 		t.Fatalf("expected no error, got %v", err)
 	}
@@ -102,6 +104,7 @@ func TestGoogleModelListModelsReturnsModelNames(t *testing.T) {
 }
 
 func TestGoogleModelCheckConnectionUsesListModels(t *testing.T) {
+	ctx := t.Context()
 	customBaseURL := "https://check-connection.example.test/google"
 	model := NewGoogleModel(map[string]string{"default": customBaseURL}, URLSuffix{})
 	apiKey := "test-api-key"
@@ -118,7 +121,7 @@ func TestGoogleModelCheckConnectionUsesListModels(t *testing.T) {
 		return []ListModelResponse{{Name: "models/gemini-2.5-flash"}}, nil
 	})
 
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("expected no error, got %v", err)
 	}
 	if calls != 1 {
@@ -127,6 +130,7 @@ func TestGoogleModelCheckConnectionUsesListModels(t *testing.T) {
 }
 
 func TestGoogleModelCheckConnectionRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	calls := 0
 
@@ -163,7 +167,7 @@ func TestGoogleModelCheckConnectionRequiresAPIKey(t *testing.T) {
 
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
-			err := model.CheckConnection(tc.apiConfig)
+			err := model.CheckConnection(ctx, tc.apiConfig)
 			if err == nil {
 				t.Fatal("expected an API key error")
 			}
@@ -178,6 +182,7 @@ func TestGoogleModelCheckConnectionRequiresAPIKey(t *testing.T) {
 }
 
 func TestGoogleModelCheckConnectionReturnsListModelsError(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	apiKey := "test-api-key"
 	listErr := errors.New("list models failed")
@@ -186,13 +191,14 @@ func TestGoogleModelCheckConnectionReturnsListModelsError(t *testing.T) {
 		return nil, listErr
 	})
 
-	err := model.CheckConnection(&APIConfig{ApiKey: &apiKey})
+	err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey})
 	if !errors.Is(err, listErr) {
 		t.Fatalf("expected ListModels error %v, got %v", listErr, err)
 	}
 }
 
 func TestGoogleModelChatStreamlyRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	messages := []Message{{Role: "user", Content: "hello"}}
 	cases := []struct {
@@ -207,7 +213,7 @@ func TestGoogleModelChatStreamlyRequiresAPIKey(t *testing.T) {
 
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
-			err := model.ChatStreamlyWithSender("gemini-2.5-flash", messages, tc.apiConfig, nil, nil, func(*string, *string) error {
+			err := model.ChatStreamlyWithSender(ctx, "gemini-2.5-flash", messages, tc.apiConfig, nil, nil, func(*string, *string) error {
 				t.Errorf("sender should not be called without an API key")
 				return nil
 			})
@@ -222,11 +228,12 @@ func TestGoogleModelChatStreamlyRequiresAPIKey(t *testing.T) {
 }
 
 func TestGoogleModelChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	model := &GoogleModel{}
 	apiKey := "test-api-key"
 	messages := []Message{{Role: "user", Content: "hello"}}
 
-	response, err := model.ChatWithMessages("", messages, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	response, err := model.ChatWithMessages(ctx, "", messages, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil {
 		t.Fatal("expected a model name error")
 	}
@@ -237,7 +244,7 @@ func TestGoogleModelChatRequiresModelName(t *testing.T) {
 		t.Fatalf("expected no response, got %v", response)
 	}
 
-	err = model.ChatStreamlyWithSender("", messages, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error {
+	err = model.ChatStreamlyWithSender(ctx, "", messages, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error {
 		t.Errorf("sender should not be called without a model name")
 		return nil
 	})
@@ -248,7 +255,7 @@ func TestGoogleModelChatRequiresModelName(t *testing.T) {
 		t.Fatalf("expected model name error, got %v", err)
 	}
 
-	err = model.ChatStreamlyWithSender("gemini-2.5-flash", messages, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
+	err = model.ChatStreamlyWithSender(ctx, "gemini-2.5-flash", messages, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 	if err == nil {
 		t.Fatal("expected a sender error")
 	}
@@ -312,6 +319,7 @@ func TestGoogleModelListModelsPassesBaseURL(t *testing.T) {
 
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
+			ctx := t.Context()
 			model := NewGoogleModel(tc.baseURL, URLSuffix{})
 			withGoogleListModelsStub(t, func(_ context.Context, config *genai.ClientConfig) ([]ListModelResponse, error) {
 				if config.HTTPOptions.BaseURL != tc.expectedBaseURL {
@@ -320,7 +328,7 @@ func TestGoogleModelListModelsPassesBaseURL(t *testing.T) {
 				return []ListModelResponse{{Name: "models/gemini-2.5-flash"}}, nil
 			})
 
-			if _, err := model.ListModels(&APIConfig{ApiKey: &apiKey, Region: tc.region}); err != nil {
+			if _, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: tc.region}); err != nil {
 				t.Fatalf("expected no error, got %v", err)
 			}
 		})
diff --git a/internal/entity/models/gpustack.go b/internal/entity/models/gpustack.go
index d3dd772100..14edd508f7 100644
--- a/internal/entity/models/gpustack.go
+++ b/internal/entity/models/gpustack.go
@@ -52,7 +52,7 @@ func (g *GPUStackModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages and returns the response.
-func (g *GPUStackModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (g *GPUStackModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -104,7 +104,7 @@ func (g *GPUStackModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -164,7 +164,7 @@ func (g *GPUStackModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender streams the response via the sender.
-func (g *GPUStackModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GPUStackModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -223,7 +223,7 @@ func (g *GPUStackModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), http.MethodPost, url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -304,7 +304,7 @@ type gpustackModelsResponse struct {
 	Data []ModelListItem `json:"data"`
 }
 
-func (g *GPUStackModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (g *GPUStackModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -316,7 +316,7 @@ func (g *GPUStackModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, g.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -350,8 +350,8 @@ func (g *GPUStackModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	return ParseListModel(ModelList{Models: parsed.Data}), nil
 }
 
-func (g *GPUStackModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := g.ListModels(apiConfig)
+func (g *GPUStackModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := g.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -368,6 +368,7 @@ type gpustackEmbeddingResponse struct {
 
 // Embed requests embedding vectors via GPUStack's v1-openai/embeddings endpoint.
 func (g *GPUStackModel) Embed(
+	ctx context.Context,
 	modelName *string,
 	texts []string,
 	apiConfig *APIConfig,
@@ -408,7 +409,7 @@ func (g *GPUStackModel) Embed(
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -469,42 +470,42 @@ func (g *GPUStackModel) Embed(
 	return embeddings, nil
 }
 
-func (g *GPUStackModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (g *GPUStackModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (g *GPUStackModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (g *GPUStackModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GPUStackModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (g *GPUStackModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GPUStackModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (g *GPUStackModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (g *GPUStackModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (g *GPUStackModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GPUStackModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (g *GPUStackModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
diff --git a/internal/entity/models/gpustack_test.go b/internal/entity/models/gpustack_test.go
index cbbfbeeb17..4c3675a5f6 100644
--- a/internal/entity/models/gpustack_test.go
+++ b/internal/entity/models/gpustack_test.go
@@ -112,7 +112,9 @@ func TestGPUStackChatHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	resp, err := newGPUStackForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -143,7 +145,9 @@ func TestGPUStackChatExtractsReasoningContent(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	resp, err := newGPUStackForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -179,7 +183,9 @@ func TestGPUStackChatForwardsDocumentedFields(t *testing.T) {
 	temp := 0.5
 	topP := 0.95
 	stop := []string{"END"}
+	ctx := t.Context()
 	_, err := newGPUStackForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -192,7 +198,9 @@ func TestGPUStackChatForwardsDocumentedFields(t *testing.T) {
 }
 
 func TestGPUStackChatAllowsEmptyAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newGPUStackForTest("http://unused").ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
@@ -204,7 +212,9 @@ func TestGPUStackChatAllowsEmptyAPIKey(t *testing.T) {
 
 func TestGPUStackChatRequiresModelName(t *testing.T) {
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := newGPUStackForTest("http://unused").ChatWithMessages(
+		ctx,
 		"",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -216,7 +226,9 @@ func TestGPUStackChatRequiresModelName(t *testing.T) {
 
 func TestGPUStackChatRequiresMessages(t *testing.T) {
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := newGPUStackForTest("http://unused").ChatWithMessages(
+		ctx,
 		"qwen3-8b", nil, &APIConfig{ApiKey: &apiKey}, nil, nil,
 	)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
@@ -232,7 +244,9 @@ func TestGPUStackChatRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := newGPUStackForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -245,7 +259,9 @@ func TestGPUStackChatRejectsHTTPError(t *testing.T) {
 func TestGPUStackChatRequiresBaseURL(t *testing.T) {
 	model := NewGPUStackModel(map[string]string{}, URLSuffix{Chat: "v1/chat/completions"})
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := model.ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -267,7 +283,9 @@ func TestGPUStackStreamHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone bool
+	ctx := t.Context()
 	err := newGPUStackForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -305,7 +323,9 @@ func TestGPUStackStreamExtractsReasoningContent(t *testing.T) {
 
 	apiKey := "test-key"
 	var content, reasoning []string
+	ctx := t.Context()
 	err := newGPUStackForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -333,7 +353,9 @@ func TestGPUStackStreamExtractsReasoningContent(t *testing.T) {
 func TestGPUStackStreamRejectsExplicitFalse(t *testing.T) {
 	apiKey := "test-key"
 	stream := false
+	ctx := t.Context()
 	err := newGPUStackForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -348,7 +370,9 @@ func TestGPUStackStreamRejectsExplicitFalse(t *testing.T) {
 
 func TestGPUStackStreamRequiresSender(t *testing.T) {
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGPUStackForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil,
@@ -365,7 +389,9 @@ func TestGPUStackStreamFailsWithoutTerminal(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGPUStackForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -384,7 +410,9 @@ func TestGPUStackStreamRejectsMalformedFrame(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGPUStackForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -403,7 +431,9 @@ func TestGPUStackStreamSurfacesUpstreamError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGPUStackForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -415,6 +445,7 @@ func TestGPUStackStreamSurfacesUpstreamError(t *testing.T) {
 }
 
 func TestGPUStackListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s want GET", r.Method)
@@ -436,20 +467,21 @@ func TestGPUStackListModelsHappyPath(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newGPUStackForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "qwen3-8b,qwen3-32b" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestGPUStackListModelsAllowsEmptyAPIKey(t *testing.T) {
-	_, err := newGPUStackForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newGPUStackForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("self-hosted model should not require api key, got %v", err)
 	}
@@ -479,8 +511,9 @@ func TestGPUStackEmbedHappyPath(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
+	ctx := t.Context()
 	vecs, err := newGPUStackForTest(srv.URL).Embed(
-		&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &EmbeddingConfig{Dimension: 512}, nil)
+		ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &EmbeddingConfig{Dimension: 512}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -507,8 +540,9 @@ func TestGPUStackEmbedReordersByIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
+	ctx := t.Context()
 	vecs, err := newGPUStackForTest(srv.URL).Embed(
-		&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+		ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -529,7 +563,9 @@ func TestGPUStackEmbedEmptyInputShortCircuits(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	vecs, err := newGPUStackForTest(srv.URL).Embed(&model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	vecs, err := newGPUStackForTest(srv.URL).Embed(
+		ctx, &model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed([]): %v", err)
 	}
@@ -540,8 +576,9 @@ func TestGPUStackEmbedEmptyInputShortCircuits(t *testing.T) {
 
 // TestGPUStackEmbedRequiresAPIKey rejects requests without an API key.
 func TestGPUStackEmbedAllowsEmptyAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := "bge-m3"
-	_, err := newGPUStackForTest("http://unused").Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := newGPUStackForTest("http://unused").Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("self-hosted model should not require api key, got %v", err)
 	}
@@ -561,7 +598,8 @@ func TestGPUStackEmbedRejectsDuplicateIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	_, err := newGPUStackForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	_, err := newGPUStackForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
@@ -580,7 +618,8 @@ func TestGPUStackEmbedRejectsOutOfRangeIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	_, err := newGPUStackForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	_, err := newGPUStackForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
@@ -599,7 +638,8 @@ func TestGPUStackEmbedRejectsMissingIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	_, err := newGPUStackForTest(srv.URL).Embed(&model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	_, err := newGPUStackForTest(srv.URL).Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding index") {
 		t.Errorf("expected missing-index error, got %v", err)
 	}
@@ -618,7 +658,8 @@ func TestGPUStackEmbedRejectsEmptyVector(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	_, err := newGPUStackForTest(srv.URL).Embed(&model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	_, err := newGPUStackForTest(srv.URL).Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "empty embedding vector") {
 		t.Errorf("expected empty-vector error, got %v", err)
 	}
@@ -637,28 +678,30 @@ func TestGPUStackEmbedRejectsMissingSlot(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "bge-m3"
-	_, err := newGPUStackForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	ctx := t.Context()
+	_, err := newGPUStackForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding for input index") {
 		t.Errorf("expected missing-slot error, got %v", err)
 	}
 }
 
 func TestGPUStackUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newGPUStackForTest("http://unused")
 	model := "x"
-	if _, err := m.Rerank(&model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: %v", err)
 	}
-	if _, err := m.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: %v", err)
 	}
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
diff --git a/internal/entity/models/groq.go b/internal/entity/models/groq.go
index 8b034681b6..1052dba2c4 100644
--- a/internal/entity/models/groq.go
+++ b/internal/entity/models/groq.go
@@ -125,7 +125,7 @@ type groqChatResponse struct {
 	FinishReason string           `json:"finish_reason"`
 }
 
-func (g *GroqModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (g *GroqModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -146,7 +146,7 @@ func (g *GroqModel) ChatWithMessages(modelName string, messages []Message, apiCo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -193,7 +193,7 @@ func (g *GroqModel) ChatWithMessages(modelName string, messages []Message, apiCo
 	}, nil
 }
 
-func (g *GroqModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GroqModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -221,7 +221,7 @@ func (g *GroqModel) ChatStreamlyWithSender(modelName string, messages []Message,
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -292,7 +292,7 @@ type groqListModelsResponse struct {
 	Error interface{}     `json:"error"`
 }
 
-func (g *GroqModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (g *GroqModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -302,7 +302,7 @@ func (g *GroqModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -337,24 +337,24 @@ func (g *GroqModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (g *GroqModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := g.ListModels(apiConfig)
+func (g *GroqModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := g.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (g *GroqModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (g *GroqModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (g *GroqModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (g *GroqModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (g *GroqModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -471,11 +471,11 @@ func (g *GroqModel) TranscribeAudio(modelName *string, file *string, apiConfig *
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (g *GroqModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GroqModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (g *GroqModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := g.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -535,22 +535,22 @@ func (g *GroqModel) AudioSpeech(modelName *string, audioContent *string, apiConf
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (g *GroqModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (g *GroqModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (g *GroqModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (g *GroqModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (g *GroqModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
 
-func (g *GroqModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (g *GroqModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", g.Name())
 }
diff --git a/internal/entity/models/groq_test.go b/internal/entity/models/groq_test.go
index 3811a35883..b6cc65d1f4 100644
--- a/internal/entity/models/groq_test.go
+++ b/internal/entity/models/groq_test.go
@@ -125,7 +125,9 @@ func TestGroqChatHappyPath(t *testing.T) {
 	topP := 0.9
 	stop := []string{"END"}
 	effort := "high"
+	ctx := t.Context()
 	resp, err := newGroqForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -144,16 +146,18 @@ func TestGroqChatHappyPath(t *testing.T) {
 }
 
 func TestGroqChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newGroqForTest("http://unused").ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newGroqForTest("http://unused").ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestGroqChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newGroqForTest("http://unused").ChatWithMessages("llama-3.3-70b-versatile", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newGroqForTest("http://unused").ChatWithMessages(ctx, "llama-3.3-70b-versatile", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
@@ -167,7 +171,9 @@ func TestGroqChatRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := newGroqForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -203,7 +209,9 @@ func TestGroqStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	var sawDone bool
+	ctx := t.Context()
 	err := newGroqForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -239,7 +247,9 @@ func TestGroqStreamHappyPath(t *testing.T) {
 func TestGroqStreamRejectsExplicitFalse(t *testing.T) {
 	apiKey := "test-key"
 	stream := false
+	ctx := t.Context()
 	err := newGroqForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -254,7 +264,9 @@ func TestGroqStreamRejectsExplicitFalse(t *testing.T) {
 
 func TestGroqStreamRequiresSender(t *testing.T) {
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGroqForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -275,7 +287,9 @@ func TestGroqStreamRejectsUnterminatedStream(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newGroqForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -289,6 +303,7 @@ func TestGroqStreamRejectsUnterminatedStream(t *testing.T) {
 }
 
 func TestGroqListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newGroqServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s", r.Method)
@@ -307,19 +322,20 @@ func TestGroqListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newGroqForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "llama-3.3-70b-versatile,openai/gpt-oss-120b" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestGroqBaseURLTrimsTrailingSlash(t *testing.T) {
+	ctx := t.Context()
 	srv := newGroqServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -334,6 +350,7 @@ func TestGroqBaseURLTrimsTrailingSlash(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newGroqForTest(srv.URL+"/").ChatWithMessages(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -346,6 +363,7 @@ func TestGroqBaseURLTrimsTrailingSlash(t *testing.T) {
 }
 
 func TestGroqUsesEmptyRegionCustomBaseURL(t *testing.T) {
+	ctx := t.Context()
 	srv := newGroqServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -365,6 +383,7 @@ func TestGroqUsesEmptyRegionCustomBaseURL(t *testing.T) {
 		URLSuffix{Chat: "chat/completions", Models: "models"},
 	)
 	_, err := model.ChatWithMessages(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &region},
@@ -377,20 +396,21 @@ func TestGroqUsesEmptyRegionCustomBaseURL(t *testing.T) {
 }
 
 func TestGroqUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newGroqForTest("http://unused")
-	if _, err := m.Embed(nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Embed(ctx, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed error=%v", err)
 	}
-	if _, err := m.Rerank(nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank error=%v", err)
 	}
-	if _, err := m.Balance(nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance error=%v", err)
 	}
-	if _, err := m.ParseFile(nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ParseFile(ctx, nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ParseFile error=%v", err)
 	}
-	if _, err := m.ListTasks(nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ListTasks(ctx, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ListTasks error=%v", err)
 	}
 }
diff --git a/internal/entity/models/huaweicloud.go b/internal/entity/models/huaweicloud.go
index 8cf75c9eb2..54eef1ce4d 100644
--- a/internal/entity/models/huaweicloud.go
+++ b/internal/entity/models/huaweicloud.go
@@ -163,7 +163,7 @@ func huaweiCloudApplyChatConfig(req map[string]any, modelName string, chatModelC
 	}
 }
 
-func (h *HuaweiCloudModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (h *HuaweiCloudModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -200,7 +200,7 @@ func (h *HuaweiCloudModel) ChatWithMessages(modelName string, messages []Message
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -264,7 +264,7 @@ func (h *HuaweiCloudModel) ChatWithMessages(modelName string, messages []Message
 	}, nil
 }
 
-func (h *HuaweiCloudModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuaweiCloudModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -310,7 +310,7 @@ func (h *HuaweiCloudModel) ChatStreamlyWithSender(modelName string, messages []M
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -391,7 +391,7 @@ type huaweiCloudEmbeddingResponse struct {
 	} `json:"data"`
 }
 
-func (h *HuaweiCloudModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (h *HuaweiCloudModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -425,7 +425,7 @@ func (h *HuaweiCloudModel) Embed(modelName *string, texts []string, apiConfig *A
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -489,7 +489,7 @@ func (h *HuaweiCloudModel) Embed(modelName *string, texts []string, apiConfig *A
 	return embeddings, nil
 }
 
-func (h *HuaweiCloudModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (h *HuaweiCloudModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -526,7 +526,7 @@ func (h *HuaweiCloudModel) Rerank(modelName *string, query string, documents []s
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -582,31 +582,31 @@ func (h *HuaweiCloudModel) Rerank(modelName *string, query string, documents []s
 	return result, nil
 }
 
-func (h *HuaweiCloudModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (h *HuaweiCloudModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuaweiCloudModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (h *HuaweiCloudModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuaweiCloudModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (h *HuaweiCloudModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (h *HuaweiCloudModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (h *HuaweiCloudModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -623,7 +623,7 @@ func (h *HuaweiCloudModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, strings.TrimPrefix(h.baseModel.URLSuffix.Models, "/"))
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -659,19 +659,19 @@ func (h *HuaweiCloudModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 	return ParseListModel(ModelList{Models: parsed.Data}), nil
 }
 
-func (h *HuaweiCloudModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (h *HuaweiCloudModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := h.ListModels(apiConfig)
+func (h *HuaweiCloudModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := h.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (h *HuaweiCloudModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (h *HuaweiCloudModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuaweiCloudModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (h *HuaweiCloudModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
diff --git a/internal/entity/models/huggingface.go b/internal/entity/models/huggingface.go
index 4c7e49dd04..1cce23edf7 100644
--- a/internal/entity/models/huggingface.go
+++ b/internal/entity/models/huggingface.go
@@ -49,7 +49,7 @@ func (h *HuggingFaceModel) Name() string {
 	return "huggingface"
 }
 
-func (h *HuggingFaceModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (h *HuggingFaceModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -120,7 +120,7 @@ func (h *HuggingFaceModel) ChatWithMessages(modelName string, messages []Message
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -192,7 +192,7 @@ func (h *HuggingFaceModel) ChatWithMessages(modelName string, messages []Message
 	return chatResponse, nil
 }
 
-func (h *HuggingFaceModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuggingFaceModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -264,7 +264,7 @@ func (h *HuggingFaceModel) ChatStreamlyWithSender(modelName string, messages []M
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -328,7 +328,7 @@ func (h *HuggingFaceModel) ChatStreamlyWithSender(modelName string, messages []M
 	return sender(&endOfStream, nil)
 }
 
-func (h *HuggingFaceModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (h *HuggingFaceModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -356,7 +356,7 @@ func (h *HuggingFaceModel) Embed(modelName *string, texts []string, apiConfig *A
 	}
 	url := fmt.Sprintf("%s/%s/%s", resolvedBaseURL, h.baseModel.URLSuffix.Embedding, *modelName)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -398,39 +398,39 @@ func (h *HuggingFaceModel) Embed(modelName *string, texts []string, apiConfig *A
 	return embeddings, nil
 }
 
-func (h *HuggingFaceModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (h *HuggingFaceModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (h *HuggingFaceModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (h *HuggingFaceModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuggingFaceModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuggingFaceModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
 // AudioSpeech convert text to audio
-func (h *HuggingFaceModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (h *HuggingFaceModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuggingFaceModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HuggingFaceModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
 // OCRFile OCR file
-func (h *HuggingFaceModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (h *HuggingFaceModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
 // ParseFile parse file
-func (h *HuggingFaceModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (h *HuggingFaceModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuggingFaceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (h *HuggingFaceModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -449,7 +449,7 @@ func (h *HuggingFaceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -488,19 +488,19 @@ func (h *HuggingFaceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 	return ParseListModel(modelList), nil
 }
 
-func (h *HuggingFaceModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (h *HuggingFaceModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (h *HuggingFaceModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := h.ListModels(apiConfig)
+func (h *HuggingFaceModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := h.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (h *HuggingFaceModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (h *HuggingFaceModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HuggingFaceModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (h *HuggingFaceModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
diff --git a/internal/entity/models/hunyuan.go b/internal/entity/models/hunyuan.go
index c777706711..a1c8f7caca 100644
--- a/internal/entity/models/hunyuan.go
+++ b/internal/entity/models/hunyuan.go
@@ -51,7 +51,7 @@ func (h *HunyuanModel) Name() string {
 	return "Tencent Hunyuan"
 }
 
-func (h *HunyuanModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (h *HunyuanModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -100,7 +100,7 @@ func (h *HunyuanModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -161,7 +161,7 @@ func (h *HunyuanModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender opens the SSE chat-completions
-func (h *HunyuanModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HunyuanModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -217,7 +217,7 @@ func (h *HunyuanModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -282,7 +282,7 @@ func (h *HunyuanModel) ChatStreamlyWithSender(modelName string, messages []Messa
 	return nil
 }
 
-func (h *HunyuanModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (h *HunyuanModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -294,7 +294,7 @@ func (h *HunyuanModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, h.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -329,12 +329,12 @@ func (h *HunyuanModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	return ParseListModel(modelList), nil
 }
 
-func (h *HunyuanModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := h.ListModels(apiConfig)
+func (h *HunyuanModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := h.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (h *HunyuanModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (h *HunyuanModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := h.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -363,7 +363,7 @@ func (h *HunyuanModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -415,42 +415,42 @@ func (h *HunyuanModel) Embed(modelName *string, texts []string, apiConfig *APICo
 	return embeddings, nil
 }
 
-func (h *HunyuanModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (h *HunyuanModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (h *HunyuanModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (h *HunyuanModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HunyuanModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (h *HunyuanModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (h *HunyuanModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (h *HunyuanModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (h *HunyuanModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (h *HunyuanModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
 
-func (h *HunyuanModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (h *HunyuanModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", h.Name())
 }
diff --git a/internal/entity/models/hunyuan_test.go b/internal/entity/models/hunyuan_test.go
index fb31ff97c0..1c521c42c6 100644
--- a/internal/entity/models/hunyuan_test.go
+++ b/internal/entity/models/hunyuan_test.go
@@ -123,7 +123,9 @@ func TestHunyuanChatHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	mt := 64
 	temp := 0.3
+	ctx := t.Context()
 	resp, err := newHunyuanForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -152,7 +154,9 @@ func TestHunyuanChatNoReasoning(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	resp, err := newHunyuanForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"hunyuan-lite",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -169,7 +173,9 @@ func TestHunyuanChatNoReasoning(t *testing.T) {
 }
 
 func TestHunyuanChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newHunyuanForTest("http://unused").ChatWithMessages(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
@@ -180,8 +186,10 @@ func TestHunyuanChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestHunyuanChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	_, err := newHunyuanForTest("http://unused").ChatWithMessages(
+		ctx,
 		"hunyuan-pro", nil, &APIConfig{ApiKey: &apiKey}, nil, nil,
 	)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
@@ -197,7 +205,9 @@ func TestHunyuanChatPropagatesHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	_, err := newHunyuanForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -219,7 +229,9 @@ func TestHunyuanStreamHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone bool
+	ctx := t.Context()
 	err := newHunyuanForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -257,7 +269,9 @@ func TestHunyuanStreamSplitsReasoning(t *testing.T) {
 
 	apiKey := "test-key"
 	var content, reasoning []string
+	ctx := t.Context()
 	err := newHunyuanForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-standard-256K",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -287,7 +301,9 @@ func TestHunyuanStreamSplitsReasoning(t *testing.T) {
 func TestHunyuanStreamRejectsExplicitFalse(t *testing.T) {
 	apiKey := "test-key"
 	stream := false
+	ctx := t.Context()
 	err := newHunyuanForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -300,8 +316,10 @@ func TestHunyuanStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestHunyuanStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newHunyuanForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
@@ -317,7 +335,9 @@ func TestHunyuanStreamFailsWithoutTerminal(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newHunyuanForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -335,7 +355,9 @@ func TestHunyuanStreamRejectsMalformedFrame(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newHunyuanForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -353,7 +375,9 @@ func TestHunyuanStreamSurfacesUpstreamError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
+	ctx := t.Context()
 	err := newHunyuanForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"hunyuan-pro",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -367,6 +391,7 @@ func TestHunyuanStreamSurfacesUpstreamError(t *testing.T) {
 }
 
 func TestHunyuanListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newHunyuanServer(t, http.MethodGet, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -379,7 +404,7 @@ func TestHunyuanListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newHunyuanForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newHunyuanForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -390,13 +415,15 @@ func TestHunyuanListModelsHappyPath(t *testing.T) {
 }
 
 func TestHunyuanListModelsRequiresAPIKey(t *testing.T) {
-	_, err := newHunyuanForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newHunyuanForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestHunyuanCheckConnectionDelegatesToListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := newHunyuanServer(t, http.MethodGet, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{{"id": "hunyuan-pro"}},
@@ -405,12 +432,13 @@ func TestHunyuanCheckConnectionDelegatesToListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newHunyuanForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newHunyuanForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 }
 
 func TestHunyuanCheckConnectionPropagatesError(t *testing.T) {
+	ctx := t.Context()
 	srv := newHunyuanServer(t, http.MethodGet, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"bad key"}`))
@@ -418,23 +446,25 @@ func TestHunyuanCheckConnectionPropagatesError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	err := newHunyuanForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey})
+	err := newHunyuanForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestHunyuanBaseURLForRegionUnknown(t *testing.T) {
+	ctx := t.Context()
 	m := newHunyuanForTest("http://unused")
 	apiKey := "test-key"
 	region := "missing"
-	_, err := m.ListModels(&APIConfig{ApiKey: &apiKey, Region: &region})
+	_, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: &region})
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured") {
 		t.Errorf("expected base-URL error, got %v", err)
 	}
 }
 
 func TestHunyuanEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newHunyuanServer(t, http.MethodPost, "/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "hunyuan-embedding" {
 			t.Errorf("model=%v", body["model"])
@@ -456,7 +486,7 @@ func TestHunyuanEmbedHappyPath(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "hunyuan-embedding"
-	embeddings, err := newHunyuanForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	embeddings, err := newHunyuanForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -466,41 +496,43 @@ func TestHunyuanEmbedHappyPath(t *testing.T) {
 }
 
 func TestHunyuanEmbedValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := "hunyuan-embedding"
 
-	if embeddings, err := newHunyuanForTest("http://unused").Embed(nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil || len(embeddings) != 0 {
+	if embeddings, err := newHunyuanForTest("http://unused").Embed(ctx, nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil || len(embeddings) != 0 {
 		t.Errorf("empty input: embeddings=%+v err=%v", embeddings, err)
 	}
-	if _, err := newHunyuanForTest("http://unused").Embed(nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := newHunyuanForTest("http://unused").Embed(ctx, nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("nil model: %v", err)
 	}
 	emptyModel := ""
-	if _, err := newHunyuanForTest("http://unused").Embed(&emptyModel, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := newHunyuanForTest("http://unused").Embed(ctx, &emptyModel, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("empty model: %v", err)
 	}
-	if _, err := newHunyuanForTest("http://unused").Embed(&model, []string{"x"}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := newHunyuanForTest("http://unused").Embed(ctx, &model, []string{"x"}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("nil api config: %v", err)
 	}
-	if _, err := newHunyuanForTest("http://unused").Embed(&model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := newHunyuanForTest("http://unused").Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("missing api key: %v", err)
 	}
 	emptyKey := ""
-	if _, err := newHunyuanForTest("http://unused").Embed(&model, []string{"x"}, &APIConfig{ApiKey: &emptyKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := newHunyuanForTest("http://unused").Embed(ctx, &model, []string{"x"}, &APIConfig{ApiKey: &emptyKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("empty api key: %v", err)
 	}
 }
 
 func TestHunyuanAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newHunyuanForTest("http://unused")
 	model := "x"
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
diff --git a/internal/entity/models/jiekouai.go b/internal/entity/models/jiekouai.go
index 31b6a19a57..9a9a0cee4d 100644
--- a/internal/entity/models/jiekouai.go
+++ b/internal/entity/models/jiekouai.go
@@ -18,6 +18,7 @@ package models
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -61,7 +62,7 @@ func validateJieKouAIModelName(modelName *string) (string, error) {
 	return strings.TrimSpace(*modelName), nil
 }
 
-func (j *JieKouAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (j *JieKouAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -196,7 +197,7 @@ func (j *JieKouAIModel) ChatWithMessages(modelName string, messages []Message, a
 	return chatResponse, nil
 }
 
-func (j *JieKouAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JieKouAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -332,7 +333,7 @@ func (j *JieKouAIModel) ChatStreamlyWithSender(modelName string, messages []Mess
 	return nil
 }
 
-func (j *JieKouAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (j *JieKouAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -411,7 +412,7 @@ func (j *JieKouAIModel) Embed(modelName *string, texts []string, apiConfig *APIC
 	return embeddings, nil
 }
 
-func (j *JieKouAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (j *JieKouAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -494,31 +495,31 @@ func (j *JieKouAIModel) Rerank(modelName *string, query string, documents []stri
 	return &rerankResponse, nil
 }
 
-func (j *JieKouAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (j *JieKouAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JieKouAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (j *JieKouAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JieKouAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (j *JieKouAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (j *JieKouAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (j *JieKouAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -574,19 +575,19 @@ func (j *JieKouAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (j *JieKouAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (j *JieKouAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JieKouAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := j.ListModels(apiConfig)
+func (j *JieKouAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := j.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (j *JieKouAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (j *JieKouAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", j.Name())
 }
 
-func (j *JieKouAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (j *JieKouAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", j.Name())
 }
diff --git a/internal/entity/models/jiekouai_test.go b/internal/entity/models/jiekouai_test.go
index 588dcfc15d..c87243a474 100644
--- a/internal/entity/models/jiekouai_test.go
+++ b/internal/entity/models/jiekouai_test.go
@@ -89,7 +89,8 @@ func TestJieKouAIChatForcesNonStreaming(t *testing.T) {
 	apiKey := "test-key"
 	stream := true
 	thinking := true
-	resp, err := newJieKouAIForTest(srv.URL).ChatWithMessages(
+	ctx := t.Context()
+	resp, err := newJieKouAIForTest(srv.URL).ChatWithMessages(ctx,
 		" gpt-5 ",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -131,7 +132,8 @@ func TestJieKouAIStreamForcesStreaming(t *testing.T) {
 	apiKey := "test-key"
 	stream := false
 	var content, reasoning []string
-	err := newJieKouAIForTest(srv.URL).ChatStreamlyWithSender(
+	ctx := t.Context()
+	err := newJieKouAIForTest(srv.URL).ChatStreamlyWithSender(ctx,
 		"gpt-5",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -159,6 +161,7 @@ func TestJieKouAIStreamForcesStreaming(t *testing.T) {
 }
 
 func TestJieKouAIListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newJieKouAIServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -176,7 +179,7 @@ func TestJieKouAIListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newJieKouAIForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newJieKouAIForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -186,6 +189,7 @@ func TestJieKouAIListModelsHappyPath(t *testing.T) {
 }
 
 func TestJieKouAIListModelsRejectsMalformedResponse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	for name, response := range map[string]interface{}{
 		"missing data": map[string]interface{}{"object": "list"},
@@ -197,7 +201,7 @@ func TestJieKouAIListModelsRejectsMalformedResponse(t *testing.T) {
 			})
 			defer srv.Close()
 
-			if _, err := newJieKouAIForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey}); err == nil {
+			if _, err := newJieKouAIForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 				t.Fatal("expected malformed response error")
 			}
 		})
@@ -205,6 +209,7 @@ func TestJieKouAIListModelsRejectsMalformedResponse(t *testing.T) {
 }
 
 func TestJieKouAIEmbedSendsValidatedRequest(t *testing.T) {
+	ctx := t.Context()
 	srv := newJieKouAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/openai/v1/embeddings" {
 			t.Errorf("path=%s, want /openai/v1/embeddings", r.URL.Path)
@@ -227,7 +232,7 @@ func TestJieKouAIEmbedSendsValidatedRequest(t *testing.T) {
 
 	apiKey := "test-key"
 	model := " text-embedding-3-large "
-	embeddings, err := newJieKouAIForTest(srv.URL).Embed(&model, []string{"hello"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	embeddings, err := newJieKouAIForTest(srv.URL).Embed(ctx, &model, []string{"hello"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -237,6 +242,7 @@ func TestJieKouAIEmbedSendsValidatedRequest(t *testing.T) {
 }
 
 func TestJieKouAIRerankHandlesNilConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newJieKouAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/openai/v1/rerank" {
 			t.Errorf("path=%s, want /openai/v1/rerank", r.URL.Path)
@@ -261,7 +267,7 @@ func TestJieKouAIRerankHandlesNilConfig(t *testing.T) {
 
 	apiKey := "test-key"
 	model := " baai/bge-reranker-v2-m3 "
-	resp, err := newJieKouAIForTest(srv.URL).Rerank(&model, " question ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newJieKouAIForTest(srv.URL).Rerank(ctx, &model, " question ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
@@ -271,6 +277,7 @@ func TestJieKouAIRerankHandlesNilConfig(t *testing.T) {
 }
 
 func TestJieKouAIValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	emptyKey := "  "
 	model := "gpt-5"
@@ -284,7 +291,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "chat api key",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").ChatWithMessages("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").ChatWithMessages(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -292,7 +299,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "chat model",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").ChatWithMessages("  ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").ChatWithMessages(ctx, "  ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -300,21 +307,21 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "stream api key",
 			run: func() error {
-				return newJieKouAIForTest("http://unused").ChatStreamlyWithSender("gpt-5", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
+				return newJieKouAIForTest("http://unused").ChatStreamlyWithSender(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
 			},
 			want: "api key is required",
 		},
 		{
 			name: "stream sender",
 			run: func() error {
-				return newJieKouAIForTest("http://unused").ChatStreamlyWithSender("gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
+				return newJieKouAIForTest("http://unused").ChatStreamlyWithSender(ctx, "gpt-5", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 			},
 			want: "sender is required",
 		},
 		{
 			name: "embed model",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").Embed(nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").Embed(ctx, nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -322,7 +329,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "embed api key",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").Embed(&model, []string{"x"}, nil, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").Embed(ctx, &model, []string{"x"}, nil, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -330,7 +337,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "rerank model",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").Rerank(nil, "q", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").Rerank(ctx, nil, "q", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -338,7 +345,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "rerank api key",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").Rerank(&model, "q", []string{"doc"}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").Rerank(ctx, &model, "q", []string{"doc"}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -346,7 +353,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "rerank query",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").Rerank(&model, "  ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newJieKouAIForTest("http://unused").Rerank(ctx, &model, "  ", []string{"doc"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "query is required",
@@ -354,7 +361,7 @@ func TestJieKouAIValidatesInputs(t *testing.T) {
 		{
 			name: "models api key",
 			run: func() error {
-				_, err := newJieKouAIForTest("http://unused").ListModels(&APIConfig{})
+				_, err := newJieKouAIForTest("http://unused").ListModels(ctx, &APIConfig{})
 				return err
 			},
 			want: "api key is required",
diff --git a/internal/entity/models/jina.go b/internal/entity/models/jina.go
index d68a0bf595..fe711a7e97 100644
--- a/internal/entity/models/jina.go
+++ b/internal/entity/models/jina.go
@@ -55,7 +55,7 @@ func (j *JinaModel) Name() string {
 	return "jina"
 }
 
-func (j *JinaModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (j *JinaModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -100,7 +100,7 @@ func (j *JinaModel) ChatWithMessages(modelName string, messages []Message, apiCo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -160,12 +160,12 @@ func (j *JinaModel) ChatWithMessages(modelName string, messages []Message, apiCo
 	}, nil
 }
 
-func (j *JinaModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JinaModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	//TODO implement me: https://api.jina.ai/docs#/Search%20Foundation%20Models/chat_completions_v1_chat_completions_post
 	return fmt.Errorf("jina does not implement ChatStreamlyWithSender(not available for now)")
 }
 
-func (j *JinaModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (j *JinaModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -239,7 +239,7 @@ func (j *JinaModel) Embed(modelName *string, texts []string, apiConfig *APIConfi
 	return embeddings, nil
 }
 
-func (j *JinaModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (j *JinaModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := j.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -317,7 +317,7 @@ func (j *JinaModel) Rerank(modelName *string, query string, documents []string,
 	return &rerankResponse, nil
 }
 
-func (j *JinaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (j *JinaModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 
 	resolvedBaseURL, err := j.baseModel.GetBaseURL(apiConfig)
 	if err != nil {
@@ -366,47 +366,47 @@ func (j *JinaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 	return ParseListModel(ModelList{Models: models}), nil
 }
 
-func (j *JinaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (j *JinaModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (j *JinaModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := j.ListModels(apiConfig)
+func (j *JinaModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := j.ListModels(ctx, apiConfig)
 	return err
 }
 
 // TranscribeAudio transcribe audio
-func (j *JinaModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (j *JinaModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JinaModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JinaModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", j.Name())
 }
 
 // AudioSpeech convert text to audio
-func (j *JinaModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (j *JinaModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JinaModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (j *JinaModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", j.Name())
 }
 
 // OCRFile OCR file
-func (j *JinaModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (j *JinaModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
 // ParseFile parse file
-func (j *JinaModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (j *JinaModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JinaModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (j *JinaModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
 
-func (j *JinaModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (j *JinaModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", j.Name())
 }
diff --git a/internal/entity/models/jina_test.go b/internal/entity/models/jina_test.go
index 26389771d8..c34f9a93ed 100644
--- a/internal/entity/models/jina_test.go
+++ b/internal/entity/models/jina_test.go
@@ -51,6 +51,7 @@ func newJinaForTest(baseURL string) *JinaModel {
 }
 
 func TestJinaChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newJinaServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "jina-vlm" {
 			t.Errorf("expected model=jina-vlm, got %v", body["model"])
@@ -77,7 +78,7 @@ func TestJinaChatHappyPath(t *testing.T) {
 
 	j := newJinaForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("ChatWithMessages: %v", err)
 	}
@@ -96,6 +97,7 @@ func TestJinaChatSupportsToolCalls(t *testing.T) {
 }
 
 func TestJinaChatPropagatesConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newJinaServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["max_tokens"] != float64(128) {
 			t.Errorf("max_tokens=%v want 128", body["max_tokens"])
@@ -122,7 +124,7 @@ func TestJinaChatPropagatesConfig(t *testing.T) {
 	temperature := 0.2
 	topP := 0.8
 	stop := []string{"END"}
-	_, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "ping"}},
+	_, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{MaxTokens: &maxTokens, Temperature: &temperature, TopP: &topP, Stop: &stop},
 		nil,
@@ -180,7 +182,8 @@ func TestJinaChatValidation(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			_, err := j.ChatWithMessages(tt.modelName, tt.messages, tt.apiConfig, nil, nil)
+			ctx := t.Context()
+			_, err := j.ChatWithMessages(ctx, tt.modelName, tt.messages, tt.apiConfig, nil, nil)
 			if err == nil || !strings.Contains(err.Error(), tt.want) {
 				t.Fatalf("expected %q error, got %v", tt.want, err)
 			}
@@ -189,6 +192,7 @@ func TestJinaChatValidation(t *testing.T) {
 }
 
 func TestJinaChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newJinaServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"detail":"invalid api key"}`))
@@ -197,13 +201,14 @@ func TestJinaChatRejectsHTTPError(t *testing.T) {
 
 	j := newJinaForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "status 401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestJinaChatRejectsMalformedResponse(t *testing.T) {
+	ctx := t.Context()
 	srv := newJinaServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"choices": []map[string]interface{}{}})
 	})
@@ -211,17 +216,18 @@ func TestJinaChatRejectsMalformedResponse(t *testing.T) {
 
 	j := newJinaForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no choices in response") {
 		t.Errorf("expected malformed-response error, got %v", err)
 	}
 }
 
 func TestJinaChatRejectsUnknownRegion(t *testing.T) {
+	ctx := t.Context()
 	j := newJinaForTest("http://unused")
 	apiKey := "test-key"
 	region := "eu"
-	_, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "x"}},
+	_, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &region}, nil, nil,
 	)
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured for region") {
@@ -230,6 +236,7 @@ func TestJinaChatRejectsUnknownRegion(t *testing.T) {
 }
 
 func TestJinaChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
+	ctx := t.Context()
 	srv := newJinaServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{"message": map[string]interface{}{"content": "ok"}}},
@@ -240,7 +247,7 @@ func TestJinaChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 	j := newJinaForTest(srv.URL)
 	apiKey := "test-key"
 	emptyRegion := ""
-	_, err := j.ChatWithMessages("jina-vlm", []Message{{Role: "user", Content: "x"}},
+	_, err := j.ChatWithMessages(ctx, "jina-vlm", []Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &emptyRegion}, nil, nil,
 	)
 	if err != nil {
diff --git a/internal/entity/models/llm.go b/internal/entity/models/llm.go
index 8194fd54c1..985e3c917c 100644
--- a/internal/entity/models/llm.go
+++ b/internal/entity/models/llm.go
@@ -147,7 +147,7 @@ func (m *EinoChatModel) Generate(ctx context.Context, msgs []*schema.Message, op
 	if err != nil {
 		return nil, err
 	}
-	resp, err := m.inner.ModelDriver.ChatWithMessages(*m.inner.ModelName, internal, m.inner.APIConfig, chatCfg, nil)
+	resp, err := m.inner.ModelDriver.ChatWithMessages(ctx, *m.inner.ModelName, internal, m.inner.APIConfig, chatCfg, nil)
 	if err != nil {
 		return nil, fmt.Errorf("models: EinoChatModel.Generate(%s): %w", *m.inner.ModelName, err)
 	}
@@ -313,7 +313,7 @@ func (m *EinoChatModel) Stream(ctx context.Context, msgs []*schema.Message, opts
 	}
 	go func() {
 		defer sw.Close()
-		if err := m.inner.ModelDriver.ChatStreamlyWithSender(*m.inner.ModelName, internalMessage, m.inner.APIConfig, chatCfg, nil, sender); err != nil {
+		if err := m.inner.ModelDriver.ChatStreamlyWithSender(ctx, *m.inner.ModelName, internalMessage, m.inner.APIConfig, chatCfg, nil, sender); err != nil {
 			_ = sw.Send(nil, err)
 			return
 		}
diff --git a/internal/entity/models/llm_test.go b/internal/entity/models/llm_test.go
index 5bd22d87d4..b94bbfc957 100644
--- a/internal/entity/models/llm_test.go
+++ b/internal/entity/models/llm_test.go
@@ -250,7 +250,7 @@ type streamSentinelDriver struct {
 	*captureToolDriver
 }
 
-func (d *streamSentinelDriver) ChatStreamlyWithSender(_ string, _ []Message, _ *APIConfig, _ *ChatConfig, _ *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *streamSentinelDriver) ChatStreamlyWithSender(ctx context.Context, _ string, _ []Message, _ *APIConfig, _ *ChatConfig, _ *common.ModelUsage, sender func(*string, *string) error) error {
 	answer := "answer"
 	if err := sender(&answer, nil); err != nil {
 		return err
@@ -265,11 +265,11 @@ func (d *streamSentinelDriver) ChatStreamlyWithSender(_ string, _ []Message, _ *
 
 func (d *captureToolDriver) NewInstance(baseURL map[string]string) ModelDriver { return d }
 func (d *captureToolDriver) Name() string                                      { return "capture" }
-func (d *captureToolDriver) ChatWithMessages(_ string, _ []Message, _ *APIConfig, cfg *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (d *captureToolDriver) ChatWithMessages(ctx context.Context, _ string, _ []Message, _ *APIConfig, cfg *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	d.lastConfig = cfg
 	return d.resp, nil
 }
-func (d *captureToolDriver) ChatStreamlyWithSender(_ string, _ []Message, _ *APIConfig, cfg *ChatConfig, _ *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *captureToolDriver) ChatStreamlyWithSender(ctx context.Context, _ string, _ []Message, _ *APIConfig, cfg *ChatConfig, _ *common.ModelUsage, sender func(*string, *string) error) error {
 	d.lastConfig = cfg
 	if d.resp == nil {
 		return nil
@@ -284,40 +284,40 @@ func (d *captureToolDriver) ChatStreamlyWithSender(_ string, _ []Message, _ *API
 	}
 	return nil
 }
-func (d *captureToolDriver) Embed(_ *string, _ []string, _ *APIConfig, _ *EmbeddingConfig, _ *common.ModelUsage) ([]EmbeddingData, error) {
+func (d *captureToolDriver) Embed(ctx context.Context, _ *string, _ []string, _ *APIConfig, _ *EmbeddingConfig, _ *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) Rerank(_ *string, _ string, _ []string, _ *APIConfig, _ *RerankConfig, _ *common.ModelUsage) (*RerankResponse, error) {
+func (d *captureToolDriver) Rerank(ctx context.Context, _ *string, _ string, _ []string, _ *APIConfig, _ *RerankConfig, _ *common.ModelUsage) (*RerankResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) TranscribeAudio(_ *string, _ *string, _ *APIConfig, _ *ASRConfig, _ *common.ModelUsage) (*ASRResponse, error) {
+func (d *captureToolDriver) TranscribeAudio(ctx context.Context, _ *string, _ *string, _ *APIConfig, _ *ASRConfig, _ *common.ModelUsage) (*ASRResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) TranscribeAudioWithSender(_ *string, _ *string, _ *APIConfig, _ *ASRConfig, _ *common.ModelUsage, _ func(*string, *string) error) error {
+func (d *captureToolDriver) TranscribeAudioWithSender(ctx context.Context, _ *string, _ *string, _ *APIConfig, _ *ASRConfig, _ *common.ModelUsage, _ func(*string, *string) error) error {
 	return nil
 }
-func (d *captureToolDriver) AudioSpeech(_ *string, _ *string, _ *APIConfig, _ *TTSConfig, _ *common.ModelUsage) (*TTSResponse, error) {
+func (d *captureToolDriver) AudioSpeech(ctx context.Context, _ *string, _ *string, _ *APIConfig, _ *TTSConfig, _ *common.ModelUsage) (*TTSResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) AudioSpeechWithSender(_ *string, _ *string, _ *APIConfig, _ *TTSConfig, _ *common.ModelUsage, _ func(*string, *string) error) error {
+func (d *captureToolDriver) AudioSpeechWithSender(ctx context.Context, _ *string, _ *string, _ *APIConfig, _ *TTSConfig, _ *common.ModelUsage, _ func(*string, *string) error) error {
 	return nil
 }
-func (d *captureToolDriver) OCRFile(_ *string, _ []byte, _ *string, _ *APIConfig, _ *OCRConfig, _ *common.ModelUsage) (*OCRFileResponse, error) {
+func (d *captureToolDriver) OCRFile(ctx context.Context, _ *string, _ []byte, _ *string, _ *APIConfig, _ *OCRConfig, _ *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) ParseFile(_ *string, _ []byte, _ *string, _ *APIConfig, _ *ParseFileConfig, _ *common.ModelUsage) (*ParseFileResponse, error) {
+func (d *captureToolDriver) ParseFile(ctx context.Context, _ *string, _ []byte, _ *string, _ *APIConfig, _ *ParseFileConfig, _ *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) ListModels(_ *APIConfig) ([]ListModelResponse, error) {
+func (d *captureToolDriver) ListModels(ctx context.Context, _ *APIConfig) ([]ListModelResponse, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) Balance(_ *APIConfig) (map[string]interface{}, error) {
+func (d *captureToolDriver) Balance(ctx context.Context, _ *APIConfig) (map[string]interface{}, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) CheckConnection(_ *APIConfig) error { return nil }
-func (d *captureToolDriver) ListTasks(_ *APIConfig) ([]ListTaskStatus, error) {
+func (d *captureToolDriver) CheckConnection(ctx context.Context, _ *APIConfig) error { return nil }
+func (d *captureToolDriver) ListTasks(ctx context.Context, _ *APIConfig) ([]ListTaskStatus, error) {
 	return nil, nil
 }
-func (d *captureToolDriver) ShowTask(_ string, _ *APIConfig) (*TaskResponse, error) {
+func (d *captureToolDriver) ShowTask(ctx context.Context, _ string, _ *APIConfig) (*TaskResponse, error) {
 	return nil, nil
 }
diff --git a/internal/entity/models/lmstudio.go b/internal/entity/models/lmstudio.go
index 393f605f56..30b3c1bfc7 100644
--- a/internal/entity/models/lmstudio.go
+++ b/internal/entity/models/lmstudio.go
@@ -53,7 +53,7 @@ func (l *LmStudioModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (l *LmStudioModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (l *LmStudioModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -130,7 +130,7 @@ func (l *LmStudioModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -204,7 +204,7 @@ func (l *LmStudioModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (l *LmStudioModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LmStudioModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -280,7 +280,7 @@ func (l *LmStudioModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -351,7 +351,7 @@ func (l *LmStudioModel) ChatStreamlyWithSender(modelName string, messages []Mess
 	return nil
 }
 
-func (l *LmStudioModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (l *LmStudioModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -391,7 +391,7 @@ func (l *LmStudioModel) Embed(modelName *string, texts []string, apiConfig *APIC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -435,40 +435,40 @@ func (l *LmStudioModel) Embed(modelName *string, texts []string, apiConfig *APIC
 	return embeddings, nil
 }
 
-func (l *LmStudioModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (l *LmStudioModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (l *LmStudioModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (l *LmStudioModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LmStudioModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LmStudioModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
 // AudioSpeech convert text to audio
-func (l *LmStudioModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (l *LmStudioModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LmStudioModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LmStudioModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
 // OCRFile OCR file
-func (l *LmStudioModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (l *LmStudioModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // ParseFile parse file
-func (l *LmStudioModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (l *LmStudioModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // ListModels list supported models
-func (l *LmStudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (l *LmStudioModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -494,7 +494,7 @@ func (l *LmStudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -535,20 +535,20 @@ func (l *LmStudioModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	return ParseListModel(modelList), nil
 }
 
-func (l *LmStudioModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (l *LmStudioModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection verifies that the configured LM Studio base URL is reachable
-func (l *LmStudioModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := l.ListModels(apiConfig)
+func (l *LmStudioModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := l.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (l *LmStudioModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (l *LmStudioModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LmStudioModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (l *LmStudioModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
diff --git a/internal/entity/models/localai.go b/internal/entity/models/localai.go
index 3f06c94dc6..ccd8d8c5f8 100644
--- a/internal/entity/models/localai.go
+++ b/internal/entity/models/localai.go
@@ -81,7 +81,7 @@ func addLocalAIReasoningRequestParams(reqBody map[string]interface{}, cfg *ChatC
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (l *LocalAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (l *LocalAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -129,7 +129,7 @@ func (l *LocalAIModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -195,7 +195,7 @@ func (l *LocalAIModel) ChatWithMessages(modelName string, messages []Message, ap
 // ChatStreamlyWithSender sends messages and streams the response via the
 // sender function. The LocalAI SSE stream uses the same shape as OpenAI:
 // "data:" lines carrying JSON events, with a final "[DONE]" line.
-func (l *LocalAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LocalAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -370,7 +370,7 @@ type localAIEmbeddingResponse struct {
 	Object string                 `json:"object"`
 }
 
-func (l *LocalAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (l *LocalAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -401,7 +401,7 @@ func (l *LocalAIModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -472,7 +472,7 @@ type localAIRerankResponse struct {
 	} `json:"results"`
 }
 
-func (l *LocalAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (l *LocalAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -506,7 +506,7 @@ func (l *LocalAIModel) Rerank(modelName *string, query string, documents []strin
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -553,7 +553,7 @@ func (l *LocalAIModel) Rerank(modelName *string, query string, documents []strin
 	return rerankResponse, nil
 }
 
-func (l *LocalAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (l *LocalAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -564,7 +564,7 @@ func (l *LocalAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), l.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -604,48 +604,48 @@ func (l *LocalAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	return ParseListModel(modelList), nil
 }
 
-func (l *LocalAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (l *LocalAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
-func (l *LocalAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := l.ListModels(apiConfig)
+func (l *LocalAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := l.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
 	return nil
 }
 
-func (l *LocalAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (l *LocalAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LocalAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LocalAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
 // AudioSpeech convert text to audio
-func (l *LocalAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (l *LocalAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LocalAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LocalAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LocalAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (l *LocalAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // ParseFile parse file
-func (l *LocalAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (l *LocalAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LocalAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (l *LocalAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LocalAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (l *LocalAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
diff --git a/internal/entity/models/localai_test.go b/internal/entity/models/localai_test.go
index 63db762e2e..cd573d51d0 100644
--- a/internal/entity/models/localai_test.go
+++ b/internal/entity/models/localai_test.go
@@ -43,6 +43,7 @@ func TestLocalAIName(t *testing.T) {
 }
 
 func TestLocalAIStreamCancelsOnIdle(t *testing.T) {
+	ctx := t.Context()
 	// The server emits one valid chunk and then stalls. Without the
 	// watchdog, scanner.Scan() would hang forever. With the watchdog
 	// at 200ms, it must return a clear "stream idle" error in well
@@ -78,7 +79,7 @@ func TestLocalAIStreamCancelsOnIdle(t *testing.T) {
 	l := newLocalAIForTest(srv.URL)
 	var got []string
 	var mu sync.Mutex
-	err := l.ChatStreamlyWithSender("gpt-4",
+	err := l.ChatStreamlyWithSender(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 		func(content *string, _ *string) error {
@@ -106,6 +107,7 @@ func TestLocalAIStreamCancelsOnIdle(t *testing.T) {
 }
 
 func TestLocalAIStreamCompletesWithoutTriggeringWatchdog(t *testing.T) {
+	ctx := t.Context()
 	// Sanity check: a fast, complete stream should not trip the
 	// watchdog even with a moderately tight idle window.
 	withLocalAIIdleTimeout(t, 500*time.Millisecond)
@@ -128,7 +130,7 @@ func TestLocalAIStreamCompletesWithoutTriggeringWatchdog(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	var chunks []string
-	err := l.ChatStreamlyWithSender("gpt-4",
+	err := l.ChatStreamlyWithSender(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 		func(content *string, _ *string) error {
@@ -147,8 +149,9 @@ func TestLocalAIStreamCompletesWithoutTriggeringWatchdog(t *testing.T) {
 }
 
 func TestLocalAIStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	l := newLocalAIForTest("http://unused")
-	err := l.ChatStreamlyWithSender("gpt-4",
+	err := l.ChatStreamlyWithSender(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -157,11 +160,12 @@ func TestLocalAIStreamRequiresSender(t *testing.T) {
 }
 
 func TestLocalAIChatMissingBaseURLFailsClearly(t *testing.T) {
+	ctx := t.Context()
 	// LocalAI has no public default; resolveBaseURL must fail with a
 	// helpful message when neither the requested region nor "default"
 	// is configured.
 	l := NewLocalAIModel(map[string]string{}, URLSuffix{Chat: "chat/completions"})
-	_, err := l.ChatWithMessages("gpt-4",
+	_, err := l.ChatWithMessages(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -171,6 +175,7 @@ func TestLocalAIChatMissingBaseURLFailsClearly(t *testing.T) {
 }
 
 func TestLocalAIChatOmitsAuthHeaderWhenKeyEmpty(t *testing.T) {
+	ctx := t.Context()
 	// Optional-auth contract: LocalAI accepts an empty key, so the
 	// driver must NOT send a "Bearer " header in that case.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -182,7 +187,7 @@ func TestLocalAIChatOmitsAuthHeaderWhenKeyEmpty(t *testing.T) {
 	defer srv.Close()
 
 	l := newLocalAIForTest(srv.URL)
-	resp, err := l.ChatWithMessages("gpt-4",
+	resp, err := l.ChatWithMessages(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -195,6 +200,7 @@ func TestLocalAIChatOmitsAuthHeaderWhenKeyEmpty(t *testing.T) {
 }
 
 func TestLocalAIChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
+	ctx := t.Context()
 	// And conversely: when a tenant has put LocalAI behind an auth
 	// proxy with a token, the driver does send the Bearer header.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -207,7 +213,7 @@ func TestLocalAIChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	key := "secret"
-	_, err := l.ChatWithMessages("gpt-4",
+	_, err := l.ChatWithMessages(ctx, "gpt-4",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil,
 	)
@@ -217,14 +223,16 @@ func TestLocalAIChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 }
 
 func TestLocalAIBalanceReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	l := newLocalAIForTest("http://unused")
-	_, err := l.Balance(&APIConfig{})
+	_, err := l.Balance(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected 'no such method', got %v", err)
 	}
 }
 
 func TestLocalAIEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/embeddings" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -238,7 +246,7 @@ func TestLocalAIEmbedHappyPath(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	model := "text-embedding-ada-002"
-	vecs, err := l.Embed(&model, []string{"a", "b", "c"}, &APIConfig{}, nil, nil)
+	vecs, err := l.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -251,6 +259,7 @@ func TestLocalAIEmbedHappyPath(t *testing.T) {
 }
 
 func TestLocalAIEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	// CodeRabbit caught that a response repeating data[*].index would
 	// silently overwrite the earlier vector. Verify the driver fails
 	// loudly instead.
@@ -263,13 +272,14 @@ func TestLocalAIEmbedRejectsDuplicateIndex(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	model := "text-embedding-ada-002"
-	_, err := l.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := l.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index 0") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
 }
 
 func TestLocalAIEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"data":[{"embedding":[1],"index":7}]}`)
 	}))
@@ -277,13 +287,14 @@ func TestLocalAIEmbedRejectsOutOfRangeIndex(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	model := "text-embedding-ada-002"
-	_, err := l.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := l.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestLocalAIEmbedRejectsMissingSlot(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"data":[{"embedding":[1],"index":0}]}`)
 	}))
@@ -291,13 +302,14 @@ func TestLocalAIEmbedRejectsMissingSlot(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	model := "text-embedding-ada-002"
-	_, err := l.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := l.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding for input index 1") {
 		t.Errorf("expected missing-slot error, got %v", err)
 	}
 }
 
 func TestLocalAIEmbedEmptyInputShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		t.Error("Embed([]) made an unexpected HTTP call")
 	}))
@@ -305,7 +317,7 @@ func TestLocalAIEmbedEmptyInputShortCircuits(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	model := "text-embedding-ada-002"
-	vecs, err := l.Embed(&model, []string{}, &APIConfig{}, nil, nil)
+	vecs, err := l.Embed(ctx, &model, []string{}, &APIConfig{}, nil, nil)
 	if err != nil || len(vecs) != 0 {
 		t.Errorf("Embed([])=(%v,%v) want ([],nil)", vecs, err)
 	}
@@ -360,6 +372,7 @@ func TestExtractLocalAIReasoning(t *testing.T) {
 // when proxied through OpenAI-shim). The driver must surface it on
 // ChatResponse.ReasonContent.
 func TestLocalAIChatExtractsReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"choices":[{"message":{
 			"role":"assistant",
@@ -370,7 +383,7 @@ func TestLocalAIChatExtractsReasoningContent(t *testing.T) {
 	defer srv.Close()
 
 	l := newLocalAIForTest(srv.URL)
-	resp, err := l.ChatWithMessages("kimi-k2.6",
+	resp, err := l.ChatWithMessages(ctx, "kimi-k2.6",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -388,6 +401,7 @@ func TestLocalAIChatExtractsReasoningContent(t *testing.T) {
 // Non-streaming chat that uses message.thinking (Qwen3 via Ollama-shim
 // inside LocalAI). The driver must surface it on ReasonContent too.
 func TestLocalAIChatExtractsThinking(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"choices":[{"message":{
 			"role":"assistant",
@@ -398,7 +412,7 @@ func TestLocalAIChatExtractsThinking(t *testing.T) {
 	defer srv.Close()
 
 	l := newLocalAIForTest(srv.URL)
-	resp, err := l.ChatWithMessages("qwen3-32b",
+	resp, err := l.ChatWithMessages(ctx, "qwen3-32b",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -414,6 +428,7 @@ func TestLocalAIChatExtractsThinking(t *testing.T) {
 // non-reasoning model) must produce empty ReasonContent without
 // crashing or erroring.
 func TestLocalAIChatHandlesAbsentReasoning(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"choices":[{"message":{
 			"role":"assistant","content":"hello"
@@ -422,7 +437,7 @@ func TestLocalAIChatHandlesAbsentReasoning(t *testing.T) {
 	defer srv.Close()
 
 	l := newLocalAIForTest(srv.URL)
-	resp, err := l.ChatWithMessages("llama-3-8b-instruct",
+	resp, err := l.ChatWithMessages(ctx, "llama-3-8b-instruct",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -441,6 +456,7 @@ func TestLocalAIChatHandlesAbsentReasoning(t *testing.T) {
 // chunks and delta.content chunks (kimi-k2.6, o-series shape).
 // Reasoning must reach the sender's 2nd arg, content the 1st.
 func TestLocalAIStreamExtractsReasoningContentDelta(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		w.WriteHeader(http.StatusOK)
@@ -456,7 +472,7 @@ func TestLocalAIStreamExtractsReasoningContentDelta(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	var content, reasoning []string
-	err := l.ChatStreamlyWithSender("kimi-k2.6",
+	err := l.ChatStreamlyWithSender(ctx, "kimi-k2.6",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 		func(c *string, r *string) error {
@@ -486,6 +502,7 @@ func TestLocalAIStreamExtractsReasoningContentDelta(t *testing.T) {
 // Streaming chat where the upstream uses delta.thinking (Qwen3 shape).
 // The same handler must work.
 func TestLocalAIStreamExtractsThinkingDelta(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		w.WriteHeader(http.StatusOK)
@@ -499,7 +516,7 @@ func TestLocalAIStreamExtractsThinkingDelta(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	var got []string
-	err := l.ChatStreamlyWithSender("qwen3-32b",
+	err := l.ChatStreamlyWithSender(ctx, "qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 		func(c *string, r *string) error {
@@ -524,6 +541,7 @@ func TestLocalAIStreamExtractsThinkingDelta(t *testing.T) {
 // Request-side: ChatConfig.Effort must flow into request body as
 // reasoning_effort.
 func TestLocalAIChatPropagatesReasoningEffort(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, err := io.ReadAll(r.Body)
@@ -541,7 +559,7 @@ func TestLocalAIChatPropagatesReasoningEffort(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	effort := "high"
-	_, err := l.ChatWithMessages("kimi-k2.6",
+	_, err := l.ChatWithMessages(ctx, "kimi-k2.6",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, &ChatConfig{Effort: &effort}, nil)
 	if err != nil {
@@ -558,6 +576,7 @@ func TestLocalAIChatPropagatesReasoningEffort(t *testing.T) {
 // Request-side: ChatConfig.Thinking must flow into request body as
 // enable_thinking (Qwen3-style).
 func TestLocalAIChatPropagatesEnableThinking(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, err := io.ReadAll(r.Body)
@@ -575,7 +594,7 @@ func TestLocalAIChatPropagatesEnableThinking(t *testing.T) {
 
 	l := newLocalAIForTest(srv.URL)
 	think := true
-	_, err := l.ChatWithMessages("qwen3-32b",
+	_, err := l.ChatWithMessages(ctx, "qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, &ChatConfig{Thinking: &think}, nil)
 	if err != nil {
@@ -588,6 +607,7 @@ func TestLocalAIChatPropagatesEnableThinking(t *testing.T) {
 
 // Stream request also propagates the reasoning params.
 func TestLocalAIStreamPropagatesReasoningParams(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, err := io.ReadAll(r.Body)
@@ -610,7 +630,7 @@ func TestLocalAIStreamPropagatesReasoningParams(t *testing.T) {
 	l := newLocalAIForTest(srv.URL)
 	effort := "medium"
 	think := true
-	err := l.ChatStreamlyWithSender("kimi-k2.6",
+	err := l.ChatStreamlyWithSender(ctx, "kimi-k2.6",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, &ChatConfig{Effort: &effort, Thinking: &think}, nil,
 		func(*string, *string) error { return nil },
diff --git a/internal/entity/models/longcat.go b/internal/entity/models/longcat.go
index 856b6c81ca..5aac51be6c 100644
--- a/internal/entity/models/longcat.go
+++ b/internal/entity/models/longcat.go
@@ -52,7 +52,7 @@ func (l *LongCatModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (l *LongCatModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (l *LongCatModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -99,7 +99,7 @@ func (l *LongCatModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -167,7 +167,7 @@ func (l *LongCatModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender sends messages and streams the response via the
-func (l *LongCatModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LongCatModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -224,7 +224,7 @@ func (l *LongCatModel) ChatStreamlyWithSender(modelName string, messages []Messa
 	}
 
 	// SSE streams are long-lived.
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -309,7 +309,7 @@ type longCatListModelsResponse struct {
 
 const longCatMaxListModelsResponseBytes = 1 << 20
 
-func (l *LongCatModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (l *LongCatModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := l.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -321,7 +321,7 @@ func (l *LongCatModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, l.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -362,56 +362,56 @@ func (l *LongCatModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (l *LongCatModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := l.ListModels(apiConfig)
+func (l *LongCatModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := l.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (l *LongCatModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (l *LongCatModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // Rerank is not exposed by the LongCat API.
-func (l *LongCatModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (l *LongCatModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // Balance is not exposed by the LongCat API.
-func (l *LongCatModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (l *LongCatModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // TranscribeAudio (ASR) is not exposed by the LongCat API.
-func (l *LongCatModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (l *LongCatModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LongCatModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LongCatModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
 // AudioSpeech (TTS) is not exposed by the LongCat API.
-func (l *LongCatModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (l *LongCatModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LongCatModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (l *LongCatModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LongCatModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (l *LongCatModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
 // ParseFile parse file
-func (l *LongCatModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (l *LongCatModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LongCatModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (l *LongCatModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
 
-func (l *LongCatModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (l *LongCatModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", l.Name())
 }
diff --git a/internal/entity/models/longcat_test.go b/internal/entity/models/longcat_test.go
index c397f4a7ac..3d2ffafbf4 100644
--- a/internal/entity/models/longcat_test.go
+++ b/internal/entity/models/longcat_test.go
@@ -100,6 +100,7 @@ func TestLongCatNewModelWithCustomDefaultTransport(t *testing.T) {
 }
 
 func TestLongCatChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatServer(t, "/openai/v1/chat/completions", func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "LongCat-Flash-Chat" {
 			t.Errorf("model=%v", body["model"])
@@ -117,7 +118,7 @@ func TestLongCatChatHappyPath(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("LongCat-Flash-Chat",
+	resp, err := m.ChatWithMessages(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -135,6 +136,7 @@ func TestLongCatChatHappyPath(t *testing.T) {
 }
 
 func TestLongCatChatExtractsReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	// LongCat-Flash-Thinking returns the chain-of-thought in
 	// message.reasoning_content (OpenAI o-series shape). Live-probed
 	// against api.longcat.chat; the fixture mimics the actual response
@@ -157,7 +159,7 @@ func TestLongCatChatExtractsReasoningContent(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("LongCat-Flash-Thinking",
+	resp, err := m.ChatWithMessages(ctx, "LongCat-Flash-Thinking",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -180,6 +182,7 @@ func TestLongCatChatExtractsReasoningContent(t *testing.T) {
 // top_p — anything else is undocumented and must not be sent, since
 // the maintainer specifically flagged this on PR #14809.
 func TestLongCatChatDropsUndocumentedFields(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatServer(t, "/openai/v1/chat/completions", func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		for _, k := range []string{"stop", "reasoning_effort", "response_format", "tools", "tool_choice", "presence_penalty", "frequency_penalty", "n", "logprobs"} {
 			if _, present := body[k]; present {
@@ -207,7 +210,7 @@ func TestLongCatChatDropsUndocumentedFields(t *testing.T) {
 	topP := 0.9
 	stop := []string{"END"}
 	effort := "high"
-	_, err := m.ChatWithMessages("LongCat-Flash-Chat",
+	_, err := m.ChatWithMessages(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		// Deliberately pass Stop/Effort to prove they are filtered out.
@@ -220,8 +223,9 @@ func TestLongCatChatDropsUndocumentedFields(t *testing.T) {
 }
 
 func TestLongCatChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
-	_, err := m.ChatWithMessages("LongCat-Flash-Chat",
+	_, err := m.ChatWithMessages(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 	)
@@ -231,15 +235,17 @@ func TestLongCatChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestLongCatChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("LongCat-Flash-Chat", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "LongCat-Flash-Chat", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
 }
 
 func TestLongCatChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatServer(t, "/openai/v1/chat/completions", func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -248,7 +254,7 @@ func TestLongCatChatRejectsHTTPError(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("LongCat-Flash-Chat",
+	_, err := m.ChatWithMessages(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
@@ -257,6 +263,7 @@ func TestLongCatChatRejectsHTTPError(t *testing.T) {
 }
 
 func TestLongCatStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"content":"Hello"}}]}`+"\n"+
@@ -269,7 +276,7 @@ func TestLongCatStreamHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone bool
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(c *string, _ *string) error {
@@ -295,6 +302,7 @@ func TestLongCatStreamHappyPath(t *testing.T) {
 }
 
 func TestLongCatStreamExtractsReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	// Fixture matches the shape captured live from
 	// LongCat-Flash-Thinking against api.longcat.chat: deltas
 	// interleave reasoning_content and content within the stream.
@@ -310,7 +318,7 @@ func TestLongCatStreamExtractsReasoningContent(t *testing.T) {
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
 	var content, reasoning []string
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Thinking",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Thinking",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(c *string, r *string) error {
@@ -337,10 +345,11 @@ func TestLongCatStreamExtractsReasoningContent(t *testing.T) {
 }
 
 func TestLongCatStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	apiKey := "test-key"
 	stream := false
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Stream: &stream},
@@ -352,9 +361,10 @@ func TestLongCatStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestLongCatStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -363,6 +373,7 @@ func TestLongCatStreamRequiresSender(t *testing.T) {
 }
 
 func TestLongCatStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"delta":{"content":"half"}}]}`+"\n",
 	)
@@ -370,7 +381,7 @@ func TestLongCatStreamFailsWithoutTerminal(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -383,6 +394,7 @@ func TestLongCatStreamFailsWithoutTerminal(t *testing.T) {
 // which masked truncated or corrupted streams. The driver must now
 // fail hard with a "longcat: invalid SSE event" wrapper.
 func TestLongCatStreamRejectsMalformedFrame(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"delta":{"content":"ok"}}]}`+"\n"+
 			`data: {this is not valid json}`+"\n",
@@ -391,7 +403,7 @@ func TestLongCatStreamRejectsMalformedFrame(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -404,6 +416,7 @@ func TestLongCatStreamRejectsMalformedFrame(t *testing.T) {
 // the "no choices" continue and leave the caller with a generic
 // truncation error. The driver must surface the upstream error verbatim.
 func TestLongCatStreamSurfacesUpstreamError(t *testing.T) {
+	ctx := t.Context()
 	srv := newLongCatSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"delta":{"content":"partial "}}]}`+"\n"+
 			`data: {"error":{"message":"rate limit exceeded","type":"rate_limit_error"}}`+"\n",
@@ -412,7 +425,7 @@ func TestLongCatStreamSurfacesUpstreamError(t *testing.T) {
 
 	m := newLongCatForTest(srv.URL)
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("LongCat-Flash-Chat",
+	err := m.ChatStreamlyWithSender(ctx, "LongCat-Flash-Chat",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -425,6 +438,7 @@ func TestLongCatStreamSurfacesUpstreamError(t *testing.T) {
 }
 
 func TestLongCatListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	var requests int
 	srv := newLongCatServer(t, "/openai/v1/models", func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		requests++
@@ -463,19 +477,20 @@ func TestLongCatListModelsAndCheckConnection(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newLongCatForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newLongCatForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if got := joinModelNames(models, ","); got != "LongCat-Flash-Chat,LongCat-Flash-Thinking-2601" {
 		t.Errorf("models=%q", got)
 	}
-	if err := newLongCatForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newLongCatForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestLongCatListModelsRejectsInvalidResponses(t *testing.T) {
+	ctx := t.Context()
 	for name, response := range map[string]string{
 		"missing data": `{}`,
 		"null data":    `{"data":null}`,
@@ -488,7 +503,7 @@ func TestLongCatListModelsRejectsInvalidResponses(t *testing.T) {
 			defer srv.Close()
 
 			apiKey := "test-key"
-			_, err := newLongCatForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+			_, err := newLongCatForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 			if err == nil {
 				t.Fatalf("expected error")
 			}
@@ -497,17 +512,18 @@ func TestLongCatListModelsRejectsInvalidResponses(t *testing.T) {
 }
 
 func TestLongCatListModelsRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	for name, cfg := range map[string]*APIConfig{
 		"nil config": nil,
 		"nil key":    {},
 		"empty key":  {ApiKey: new(string)},
 	} {
 		t.Run(name, func(t *testing.T) {
-			_, err := newLongCatForTest("http://unused").ListModels(cfg)
+			_, err := newLongCatForTest("http://unused").ListModels(ctx, cfg)
 			if err == nil || !strings.Contains(err.Error(), "api key is required") {
 				t.Errorf("expected api-key error, got %v", err)
 			}
-			err = newLongCatForTest("http://unused").CheckConnection(cfg)
+			err = newLongCatForTest("http://unused").CheckConnection(ctx, cfg)
 			if err == nil || !strings.Contains(err.Error(), "api key is required") {
 				t.Errorf("CheckConnection expected api-key error, got %v", err)
 			}
@@ -516,41 +532,45 @@ func TestLongCatListModelsRequiresAPIKey(t *testing.T) {
 }
 
 func TestLongCatEmbedReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	model := "x"
-	_, err := m.Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: want 'no such method', got %v", err)
 	}
 }
 
 func TestLongCatRerankReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	model := "x"
-	_, err := m.Rerank(&model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil)
+	_, err := m.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, &RerankConfig{TopN: 1}, nil)
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: want 'no such method', got %v", err)
 	}
 }
 
 func TestLongCatBalanceReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
-	_, err := m.Balance(&APIConfig{})
+	_, err := m.Balance(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: want 'no such method', got %v", err)
 	}
 }
 
 func TestLongCatAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newLongCatForTest("http://unused")
 	model := "x"
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: want 'no such method', got %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: want 'no such method', got %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: want 'no such method', got %v", err)
 	}
 }
diff --git a/internal/entity/models/mineru.go b/internal/entity/models/mineru.go
index bafe664e77..d4c39a2583 100644
--- a/internal/entity/models/mineru.go
+++ b/internal/entity/models/mineru.go
@@ -48,51 +48,51 @@ func (m *MinerUModel) Name() string {
 	return "mineru.net"
 }
 
-func (m *MinerUModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *MinerUModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerUModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *MinerUModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *MinerUModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *MinerUModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerUModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *MinerUModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerUModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *MinerUModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *MinerUModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *MinerUModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerUModel) CheckConnection(apiConfig *APIConfig) error {
+func (m *MinerUModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
@@ -105,7 +105,7 @@ type mineruTaskSubmitResponse struct {
 	TraceID string `json:"trace_id"`
 }
 
-func (m *MinerUModel) ParseFile(modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *MinerUModel) ParseFile(ctx context.Context, modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -137,7 +137,7 @@ func (m *MinerUModel) ParseFile(modelName *string, content []byte, documentURL *
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", apiURL, bytes.NewBuffer(jsonData))
@@ -177,7 +177,7 @@ func (m *MinerUModel) ParseFile(modelName *string, content []byte, documentURL *
 	}, nil
 }
 
-func (m *MinerUModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *MinerUModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
@@ -196,7 +196,7 @@ type mineruTaskQueryResponse struct {
 	Msg string `json:"msg"`
 }
 
-func (m *MinerUModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *MinerUModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -208,7 +208,7 @@ func (m *MinerUModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskRespon
 	}
 	apiURL := fmt.Sprintf("%s/api/%s/%s", resolvedBaseURL, m.baseModel.URLSuffix.DocumentParse, taskID)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", apiURL, nil)
diff --git a/internal/entity/models/mineru_local.go b/internal/entity/models/mineru_local.go
index 125fce4ebe..f478c7eeb8 100644
--- a/internal/entity/models/mineru_local.go
+++ b/internal/entity/models/mineru_local.go
@@ -50,55 +50,55 @@ func (m *MinerULocalModel) Name() string {
 	return "mineru"
 }
 
-func (m *MinerULocalModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *MinerULocalModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerULocalModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *MinerULocalModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *MinerULocalModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *MinerULocalModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerULocalModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *MinerULocalModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinerULocalModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *MinerULocalModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *MinerULocalModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *MinerULocalModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) CheckConnection(apiConfig *APIConfig) error {
+func (m *MinerULocalModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) ParseFile(modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *MinerULocalModel) ParseFile(ctx context.Context, modelName *string, content []byte, documentURL *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -135,7 +135,7 @@ func (m *MinerULocalModel) ParseFile(modelName *string, content []byte, document
 		return nil, fmt.Errorf("failed to close multipart writer: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", apiURL, &body)
@@ -187,11 +187,11 @@ func (m *MinerULocalModel) ParseFile(modelName *string, content []byte, document
 	}, nil
 }
 
-func (m *MinerULocalModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *MinerULocalModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", m.Name())
 }
 
-func (m *MinerULocalModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *MinerULocalModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -205,7 +205,7 @@ func (m *MinerULocalModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskR
 		return nil, err
 	}
 	url := fmt.Sprintf("%s/%s/%s/result", resolvedBaseURL, m.baseModel.URLSuffix.Task, taskID)
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
diff --git a/internal/entity/models/minimax.go b/internal/entity/models/minimax.go
index e2f8db1e80..052b0812e5 100644
--- a/internal/entity/models/minimax.go
+++ b/internal/entity/models/minimax.go
@@ -60,7 +60,7 @@ func validateMinimaxModelName(modelName string) (string, error) {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (m *MinimaxModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *MinimaxModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -141,7 +141,7 @@ func (m *MinimaxModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -224,7 +224,7 @@ func (m *MinimaxModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (m *MinimaxModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinimaxModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -312,7 +312,7 @@ func (m *MinimaxModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -392,11 +392,11 @@ func (m *MinimaxModel) ChatStreamlyWithSender(modelName string, messages []Messa
 }
 
 // Embed embeds a list of texts into embeddings
-func (m *MinimaxModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *MinimaxModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
-func (m *MinimaxModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *MinimaxModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -408,7 +408,7 @@ func (m *MinimaxModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, m.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -449,31 +449,31 @@ func (m *MinimaxModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	return models, nil
 }
 
-func (m *MinimaxModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *MinimaxModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MinimaxModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := m.ListModels(apiConfig)
+func (m *MinimaxModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := m.ListModels(ctx, apiConfig)
 	return err
 }
 
 // Rerank calculates similarity scores between query and documents
-func (m *MinimaxModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *MinimaxModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", m.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (m *MinimaxModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *MinimaxModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MinimaxModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinimaxModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
 // AudioSpeech convert text to audio
-func (m *MinimaxModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *MinimaxModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -508,7 +508,7 @@ func (m *MinimaxModel) AudioSpeech(modelName *string, audioContent *string, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -563,7 +563,7 @@ func (m *MinimaxModel) AudioSpeech(modelName *string, audioContent *string, apiC
 	}, nil
 }
 
-func (m *MinimaxModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MinimaxModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -609,7 +609,7 @@ func (m *MinimaxModel) AudioSpeechWithSender(modelName *string, audioContent *st
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -658,19 +658,19 @@ func (m *MinimaxModel) AudioSpeechWithSender(modelName *string, audioContent *st
 }
 
 // OCRFile OCR file
-func (m *MinimaxModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *MinimaxModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
 // ParseFile parse file
-func (m *MinimaxModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *MinimaxModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MinimaxModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *MinimaxModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MinimaxModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *MinimaxModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
diff --git a/internal/entity/models/minimax_test.go b/internal/entity/models/minimax_test.go
index df24d993f0..fc922d2d39 100644
--- a/internal/entity/models/minimax_test.go
+++ b/internal/entity/models/minimax_test.go
@@ -85,6 +85,7 @@ func TestMinimaxNewInstancePreservesConfig(t *testing.T) {
 }
 
 func TestMinimaxChatForcesNonStreaming(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -113,6 +114,7 @@ func TestMinimaxChatForcesNonStreaming(t *testing.T) {
 	stream := true
 	thinking := true
 	resp, err := newMinimaxForTest(srv.URL).ChatWithMessages(
+		ctx,
 		" MiniMax-M3 ",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -131,6 +133,7 @@ func TestMinimaxChatForcesNonStreaming(t *testing.T) {
 }
 
 func TestMinimaxChatRejectsEmptyChoices(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"choices": []map[string]interface{}{}})
 	})
@@ -138,6 +141,7 @@ func TestMinimaxChatRejectsEmptyChoices(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newMinimaxForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"MiniMax-M3",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -150,6 +154,7 @@ func TestMinimaxChatRejectsEmptyChoices(t *testing.T) {
 }
 
 func TestMinimaxStreamForcesStreaming(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -177,6 +182,7 @@ func TestMinimaxStreamForcesStreaming(t *testing.T) {
 	stream := false
 	var content, reasoning []string
 	err := newMinimaxForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"MiniMax-M3",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -204,6 +210,7 @@ func TestMinimaxStreamForcesStreaming(t *testing.T) {
 }
 
 func TestMinimaxStreamAcceptsNilConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["stream"] != true {
 			t.Errorf("stream=%v, want true", body["stream"])
@@ -215,6 +222,7 @@ func TestMinimaxStreamAcceptsNilConfig(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newMinimaxForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"MiniMax-M3",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -228,6 +236,7 @@ func TestMinimaxStreamAcceptsNilConfig(t *testing.T) {
 }
 
 func TestMinimaxListModelsUsesBodylessGet(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -245,7 +254,7 @@ func TestMinimaxListModelsUsesBodylessGet(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newMinimaxForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newMinimaxForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -255,6 +264,7 @@ func TestMinimaxListModelsUsesBodylessGet(t *testing.T) {
 }
 
 func TestMinimaxListModelsRejectsMalformedResponse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	for name, response := range map[string]interface{}{
 		"missing data": map[string]interface{}{"object": "list"},
@@ -266,7 +276,7 @@ func TestMinimaxListModelsRejectsMalformedResponse(t *testing.T) {
 			})
 			defer srv.Close()
 
-			if _, err := newMinimaxForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey}); err == nil {
+			if _, err := newMinimaxForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 				t.Fatal("expected malformed response error")
 			}
 		})
@@ -274,6 +284,7 @@ func TestMinimaxListModelsRejectsMalformedResponse(t *testing.T) {
 }
 
 func TestMinimaxCheckConnectionUsesListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := newMinimaxServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -288,12 +299,13 @@ func TestMinimaxCheckConnectionUsesListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newMinimaxForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newMinimaxForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestMinimaxValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	emptyKey := " "
 	send := func(*string, *string) error { return nil }
@@ -306,7 +318,7 @@ func TestMinimaxValidatesInputs(t *testing.T) {
 		{
 			name: "chat api key",
 			run: func() error {
-				_, err := newMinimaxForTest("http://unused").ChatWithMessages("MiniMax-M3", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+				_, err := newMinimaxForTest("http://unused").ChatWithMessages(ctx, "MiniMax-M3", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -314,7 +326,7 @@ func TestMinimaxValidatesInputs(t *testing.T) {
 		{
 			name: "chat model",
 			run: func() error {
-				_, err := newMinimaxForTest("http://unused").ChatWithMessages(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newMinimaxForTest("http://unused").ChatWithMessages(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -322,28 +334,28 @@ func TestMinimaxValidatesInputs(t *testing.T) {
 		{
 			name: "stream api key",
 			run: func() error {
-				return newMinimaxForTest("http://unused").ChatStreamlyWithSender("MiniMax-M3", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
+				return newMinimaxForTest("http://unused").ChatStreamlyWithSender(ctx, "MiniMax-M3", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
 			},
 			want: "api key is required",
 		},
 		{
 			name: "stream model",
 			run: func() error {
-				return newMinimaxForTest("http://unused").ChatStreamlyWithSender(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, send)
+				return newMinimaxForTest("http://unused").ChatStreamlyWithSender(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, send)
 			},
 			want: "model name is required",
 		},
 		{
 			name: "stream sender",
 			run: func() error {
-				return newMinimaxForTest("http://unused").ChatStreamlyWithSender("MiniMax-M3", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
+				return newMinimaxForTest("http://unused").ChatStreamlyWithSender(ctx, "MiniMax-M3", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 			},
 			want: "sender is required",
 		},
 		{
 			name: "models api key",
 			run: func() error {
-				_, err := newMinimaxForTest("http://unused").ListModels(&APIConfig{})
+				_, err := newMinimaxForTest("http://unused").ListModels(ctx, &APIConfig{})
 				return err
 			},
 			want: "api key is required",
diff --git a/internal/entity/models/mistral.go b/internal/entity/models/mistral.go
index d9197a3723..88eadbabf6 100644
--- a/internal/entity/models/mistral.go
+++ b/internal/entity/models/mistral.go
@@ -55,7 +55,7 @@ func (m *MistralModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (m *MistralModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *MistralModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -108,7 +108,7 @@ func (m *MistralModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -208,7 +208,7 @@ func extractMistralContent(raw interface{}) (string, string, error) {
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (m *MistralModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MistralModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -266,7 +266,7 @@ func (m *MistralModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -343,7 +343,7 @@ type mistralEmbeddingResponse struct {
 }
 
 // Embed turns a list of texts into embedding vectors
-func (m *MistralModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *MistralModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -376,7 +376,7 @@ func (m *MistralModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -432,7 +432,7 @@ func (m *MistralModel) Embed(modelName *string, texts []string, apiConfig *APICo
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (m *MistralModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *MistralModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -444,7 +444,7 @@ func (m *MistralModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, m.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -482,13 +482,13 @@ func (m *MistralModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 }
 
 // Balance is not exposed by the Mistral API, so this returns "no such method".
-func (m *MistralModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *MistralModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (m *MistralModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := m.ListModels(apiConfig)
+func (m *MistralModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := m.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -496,30 +496,30 @@ func (m *MistralModel) CheckConnection(apiConfig *APIConfig) error {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (m *MistralModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *MistralModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (m *MistralModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *MistralModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MistralModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MistralModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
 // AudioSpeech convert text to audio
-func (m *MistralModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *MistralModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MistralModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MistralModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
 // OCRFile OCR file
-func (m *MistralModel) OCRFile(modelName *string, content []byte, urls *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *MistralModel) OCRFile(ctx context.Context, modelName *string, content []byte, urls *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -556,7 +556,7 @@ func (m *MistralModel) OCRFile(modelName *string, content []byte, urls *string,
 		return nil, fmt.Errorf("failed to marshal json payload: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), 120*time.Second)
+	ctx, cancel := context.WithTimeout(ctx, 120*time.Second)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -606,14 +606,14 @@ func (m *MistralModel) OCRFile(modelName *string, content []byte, urls *string,
 	}, nil
 }
 
-func (m *MistralModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *MistralModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MistralModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *MistralModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MistralModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *MistralModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
diff --git a/internal/entity/models/mistral_test.go b/internal/entity/models/mistral_test.go
index 9d15590b4c..38759ec581 100644
--- a/internal/entity/models/mistral_test.go
+++ b/internal/entity/models/mistral_test.go
@@ -65,6 +65,7 @@ func TestMistralName(t *testing.T) {
 }
 
 func TestMistralChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "mistral-large-latest" {
 			t.Errorf("expected model=mistral-large-latest, got %v", body["model"])
@@ -87,7 +88,7 @@ func TestMistralChatHappyPath(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("mistral-large-latest", []Message{
+	resp, err := m.ChatWithMessages(ctx, "mistral-large-latest", []Message{
 		{Role: "user", Content: "ping"},
 	}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -102,6 +103,7 @@ func TestMistralChatHappyPath(t *testing.T) {
 }
 
 func TestMistralChatPropagatesConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["max_tokens"] != float64(64) {
 			t.Errorf("max_tokens=%v want 64", body["max_tokens"])
@@ -128,7 +130,7 @@ func TestMistralChatPropagatesConfig(t *testing.T) {
 	temp := 0.3
 	topP := 0.9
 	stop := []string{"END"}
-	_, err := m.ChatWithMessages("mistral-large-latest", []Message{{Role: "user", Content: "ping"}},
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest", []Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{MaxTokens: &mt, Temperature: &temp, TopP: &topP, Stop: &stop},
 		nil,
@@ -139,28 +141,31 @@ func TestMistralChatPropagatesConfig(t *testing.T) {
 }
 
 func TestMistralChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
-	_, err := m.ChatWithMessages("mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 	emptyKey := ""
-	_, err = m.ChatWithMessages("mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+	_, err = m.ChatWithMessages(ctx, "mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("empty key: expected api-key error, got %v", err)
 	}
 }
 
 func TestMistralChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("mistral-large-latest", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
 }
 
 func TestMistralChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -169,13 +174,14 @@ func TestMistralChatRejectsHTTPError(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := m.ChatWithMessages("mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestMistralChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
+	ctx := t.Context()
 	// Empty *Region pointer must fall back to the "default" entry, not
 	// be treated as an explicit "" region (which would miss the lookup).
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -188,7 +194,7 @@ func TestMistralChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	emptyRegion := ""
-	_, err := m.ChatWithMessages("mistral-large-latest",
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &emptyRegion}, nil, nil,
 	)
@@ -198,6 +204,7 @@ func TestMistralChatFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 }
 
 func TestMistralListModelsFallsBackToDefaultOnEmptyRegion(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"data": []map[string]interface{}{{"id": "x"}}})
 	})
@@ -206,15 +213,16 @@ func TestMistralListModelsFallsBackToDefaultOnEmptyRegion(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	emptyRegion := ""
-	if _, err := m.ListModels(&APIConfig{ApiKey: &apiKey, Region: &emptyRegion}); err != nil {
+	if _, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: &emptyRegion}); err != nil {
 		t.Errorf("empty Region: expected fallback to default, got %v", err)
 	}
 }
 
 func TestMistralStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("mistral-large-latest",
+	err := m.ChatStreamlyWithSender(ctx, "mistral-large-latest",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -223,10 +231,11 @@ func TestMistralStreamRequiresSender(t *testing.T) {
 }
 
 func TestMistralChatRejectsUnknownRegion(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	apiKey := "test-key"
 	region := "eu"
-	_, err := m.ChatWithMessages("mistral-large-latest", []Message{{Role: "user", Content: "x"}},
+	_, err := m.ChatWithMessages(ctx, "mistral-large-latest", []Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey, Region: &region}, nil, nil,
 	)
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured for region") {
@@ -235,6 +244,7 @@ func TestMistralChatRejectsUnknownRegion(t *testing.T) {
 }
 
 func TestMistralStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -262,7 +272,7 @@ func TestMistralStreamHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone int32
-	err := m.ChatStreamlyWithSender("mistral-large-latest",
+	err := m.ChatStreamlyWithSender(ctx, "mistral-large-latest",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(content *string, _ *string) error {
@@ -289,10 +299,11 @@ func TestMistralStreamHappyPath(t *testing.T) {
 }
 
 func TestMistralStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	apiKey := "test-key"
 	stream := false
-	err := m.ChatStreamlyWithSender("mistral-large-latest",
+	err := m.ChatStreamlyWithSender(ctx, "mistral-large-latest",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Stream: &stream},
@@ -305,6 +316,7 @@ func TestMistralStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestMistralStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	// Body closes before [DONE] or a finish_reason -> driver must complain
 	// instead of pretending the stream finished cleanly.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -315,7 +327,7 @@ func TestMistralStreamFailsWithoutTerminal(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	err := m.ChatStreamlyWithSender("mistral-large-latest",
+	err := m.ChatStreamlyWithSender(ctx, "mistral-large-latest",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(*string, *string) error { return nil },
@@ -326,6 +338,7 @@ func TestMistralStreamFailsWithoutTerminal(t *testing.T) {
 }
 
 func TestMistralListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -339,7 +352,7 @@ func TestMistralListModelsHappyPath(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	ids, err := m.ListModels(&APIConfig{ApiKey: &apiKey})
+	ids, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -349,13 +362,15 @@ func TestMistralListModelsHappyPath(t *testing.T) {
 }
 
 func TestMistralListModelsRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
-	if _, err := m.ListModels(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := m.ListModels(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestMistralCheckConnectionDelegatesToListModels(t *testing.T) {
+	ctx := t.Context()
 	// 200 -> CheckConnection succeeds; 401 -> CheckConnection propagates.
 	okSrv := newMistralServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"data": []map[string]interface{}{{"id": "x"}}})
@@ -368,33 +383,36 @@ func TestMistralCheckConnectionDelegatesToListModels(t *testing.T) {
 
 	apiKey := "test-key"
 	mOK := newMistralForTest(okSrv.URL)
-	if err := mOK.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := mOK.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection(ok): %v", err)
 	}
 	mFail := newMistralForTest(failSrv.URL)
-	if err := mFail.CheckConnection(&APIConfig{ApiKey: &apiKey}); err == nil {
+	if err := mFail.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err == nil {
 		t.Error("CheckConnection(fail): expected error, got nil")
 	}
 }
 
 func TestMistralBalanceReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
-	_, err := m.Balance(&APIConfig{})
+	_, err := m.Balance(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected 'no such method', got %v", err)
 	}
 }
 
 func TestMistralRerankReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	q := "mistral-large-latest"
-	_, err := m.Rerank(&q, "what is rag?", []string{"a", "b"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
+	_, err := m.Rerank(ctx, &q, "what is rag?", []string{"a", "b"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: expected 'no such method', got %v", err)
 	}
 }
 
 func TestMistralUnsupportedDefaultsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	modelName := "mistral-large-latest"
 
@@ -403,29 +421,29 @@ func TestMistralUnsupportedDefaultsReturnNoSuchMethod(t *testing.T) {
 		call func() error
 	}{
 		{"TranscribeAudio", func() error {
-			_, err := m.TranscribeAudio(&modelName, nil, &APIConfig{}, nil, nil)
+			_, err := m.TranscribeAudio(ctx, &modelName, nil, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"TranscribeAudioWithSender", func() error {
-			return m.TranscribeAudioWithSender(&modelName, nil, &APIConfig{}, nil, nil, nil)
+			return m.TranscribeAudioWithSender(ctx, &modelName, nil, &APIConfig{}, nil, nil, nil)
 		}},
 		{"AudioSpeech", func() error {
-			_, err := m.AudioSpeech(&modelName, nil, &APIConfig{}, nil, nil)
+			_, err := m.AudioSpeech(ctx, &modelName, nil, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"AudioSpeechWithSender", func() error {
-			return m.AudioSpeechWithSender(&modelName, nil, &APIConfig{}, nil, nil, nil)
+			return m.AudioSpeechWithSender(ctx, &modelName, nil, &APIConfig{}, nil, nil, nil)
 		}},
 		{"ParseFile", func() error {
-			_, err := m.ParseFile(&modelName, nil, nil, &APIConfig{}, nil, nil)
+			_, err := m.ParseFile(ctx, &modelName, nil, nil, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"ListTasks", func() error {
-			_, err := m.ListTasks(&APIConfig{})
+			_, err := m.ListTasks(ctx, &APIConfig{})
 			return err
 		}},
 		{"ShowTask", func() error {
-			_, err := m.ShowTask("task-id", &APIConfig{})
+			_, err := m.ShowTask(ctx, "task-id", &APIConfig{})
 			return err
 		}},
 	}
@@ -438,6 +456,7 @@ func TestMistralUnsupportedDefaultsReturnNoSuchMethod(t *testing.T) {
 }
 
 func TestMistralEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "mistral-embed" {
 			t.Errorf("model=%v want mistral-embed", body["model"])
@@ -459,7 +478,7 @@ func TestMistralEmbedHappyPath(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	vecs, err := m.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := m.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -472,6 +491,7 @@ func TestMistralEmbedHappyPath(t *testing.T) {
 }
 
 func TestMistralEmbedReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	// Upstream returns the three vectors in shuffled order. The driver
 	// must reorder them so the slot at position i corresponds to input i.
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -488,7 +508,7 @@ func TestMistralEmbedReordersByIndex(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	vecs, err := m.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := m.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -500,6 +520,7 @@ func TestMistralEmbedReordersByIndex(t *testing.T) {
 }
 
 func TestMistralEmbedEmptyInputShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	// Empty input must NOT make an HTTP call; the test fails the request
 	// rather than the assertion if it does.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
@@ -511,7 +532,7 @@ func TestMistralEmbedEmptyInputShortCircuits(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	vecs, err := m.Embed(&model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := m.Embed(ctx, &model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed([]): %v", err)
 	}
@@ -521,29 +542,32 @@ func TestMistralEmbedEmptyInputShortCircuits(t *testing.T) {
 }
 
 func TestMistralEmbedRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	model := "mistral-embed"
-	_, err := m.Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestMistralEmbedRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	m := newMistralForTest("http://unused")
 	apiKey := "test-key"
-	_, err := m.Embed(nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 	empty := ""
-	_, err = m.Embed(&empty, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err = m.Embed(ctx, &empty, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("empty model: expected model-name error, got %v", err)
 	}
 }
 
 func TestMistralEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	// A malformed upstream that repeats data[*].index would silently
 	// overwrite the earlier vector; the driver must fail loudly instead.
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
@@ -559,13 +583,14 @@ func TestMistralEmbedRejectsDuplicateIndex(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index 0") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
 }
 
 func TestMistralEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -578,13 +603,14 @@ func TestMistralEmbedRejectsOutOfRangeIndex(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestMistralEmbedRejectsMissingSlot(t *testing.T) {
+	ctx := t.Context()
 	// Upstream returns only one of the two requested embeddings.
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
@@ -598,13 +624,14 @@ func TestMistralEmbedRejectsMissingSlot(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	_, err := m.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding for input index 1") {
 		t.Errorf("expected missing-embedding error for slot 1, got %v", err)
 	}
 }
 
 func TestMistralEmbedRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -614,7 +641,7 @@ func TestMistralEmbedRejectsHTTPError(t *testing.T) {
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
 	model := "mistral-embed"
-	_, err := m.Embed(&model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := m.Embed(ctx, &model, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "Mistral embeddings API error") {
 		t.Errorf("expected Mistral embeddings API error, got %v", err)
 	}
@@ -625,6 +652,7 @@ func TestMistralEmbedRejectsHTTPError(t *testing.T) {
 // Regression net: the existing string-content path stays green for every
 // non-reasoning Mistral model.
 func TestMistralChatHandlesStringContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{"message": map[string]interface{}{
@@ -637,7 +665,7 @@ func TestMistralChatHandlesStringContent(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("ministral-3b-latest",
+	resp, err := m.ChatWithMessages(ctx, "ministral-3b-latest",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -659,6 +687,7 @@ func TestMistralChatHandlesStringContent(t *testing.T) {
 // from api.mistral.ai/v1/chat/completions against magistral-medium-latest
 // with the prompt "When do two trains meet?".
 func TestMistralChatExtractsReasoningFromStructuredContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{"message": map[string]interface{}{
@@ -681,7 +710,7 @@ func TestMistralChatExtractsReasoningFromStructuredContent(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("magistral-medium-latest",
+	resp, err := m.ChatWithMessages(ctx, "magistral-medium-latest",
 		[]Message{{Role: "user", Content: "When do they meet?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -700,6 +729,7 @@ func TestMistralChatExtractsReasoningFromStructuredContent(t *testing.T) {
 // magistral with a trivial answer that needed no reasoning returns the
 // structured shape with only a `text` part. ReasonContent must be empty.
 func TestMistralChatHandlesStructuredContentWithoutThinking(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{"message": map[string]interface{}{
@@ -714,7 +744,7 @@ func TestMistralChatHandlesStructuredContentWithoutThinking(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("magistral-small-latest",
+	resp, err := m.ChatWithMessages(ctx, "magistral-small-latest",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -732,6 +762,7 @@ func TestMistralChatHandlesStructuredContentWithoutThinking(t *testing.T) {
 // the parser forward-compatible with new Mistral content variants
 // (audio chunks, citations, etc.) that ragflow doesn't surface yet.
 func TestMistralChatIgnoresUnknownContentPartTypes(t *testing.T) {
+	ctx := t.Context()
 	srv := newMistralServer(t, "/chat/completions", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{"message": map[string]interface{}{
@@ -748,7 +779,7 @@ func TestMistralChatIgnoresUnknownContentPartTypes(t *testing.T) {
 
 	m := newMistralForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := m.ChatWithMessages("magistral-small-latest",
+	resp, err := m.ChatWithMessages(ctx, "magistral-small-latest",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
diff --git a/internal/entity/models/modelscope.go b/internal/entity/models/modelscope.go
index fc9bd6c99c..3a3c533ad4 100644
--- a/internal/entity/models/modelscope.go
+++ b/internal/entity/models/modelscope.go
@@ -141,7 +141,7 @@ func buildModelScopeChatBody(modelName string, messages []Message, stream bool,
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (m *ModelScopeModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *ModelScopeModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -163,7 +163,7 @@ func (m *ModelScopeModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -211,7 +211,7 @@ func (m *ModelScopeModel) ChatWithMessages(modelName string, messages []Message,
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender.
-func (m *ModelScopeModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *ModelScopeModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -331,41 +331,41 @@ func (m *ModelScopeModel) ChatStreamlyWithSender(modelName string, messages []Me
 	return sender(&endOfStream, nil)
 }
 
-func (m *ModelScopeModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *ModelScopeModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *ModelScopeModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *ModelScopeModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *ModelScopeModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, asrConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *ModelScopeModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, asrConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *ModelScopeModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *ModelScopeModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *ModelScopeModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
 // ListModels returns the model IDs exposed by ModelScope's OpenAI-compatible
 // /v1/models endpoint.
-func (m *ModelScopeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *ModelScopeModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -377,7 +377,7 @@ func (m *ModelScopeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = normalizeModelScopeBaseURL(baseURL)
 	url := fmt.Sprintf("%s/%s", baseURL, m.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -412,19 +412,19 @@ func (m *ModelScopeModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (m *ModelScopeModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *ModelScopeModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := m.ListModels(apiConfig)
+func (m *ModelScopeModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := m.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (m *ModelScopeModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *ModelScopeModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *ModelScopeModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *ModelScopeModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
diff --git a/internal/entity/models/modelscope_test.go b/internal/entity/models/modelscope_test.go
index 1be310552d..704909b6f1 100644
--- a/internal/entity/models/modelscope_test.go
+++ b/internal/entity/models/modelscope_test.go
@@ -94,6 +94,7 @@ func TestModelScopeNewModelWithCustomDefaultTransport(t *testing.T) {
 }
 
 func TestModelScopeChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/chat/completions" {
@@ -120,7 +121,7 @@ func TestModelScopeChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T)
 	m := newModelScopeForTest(srv.URL)
 	maxTokens := 32
 	temp := 0.2
-	resp, err := m.ChatWithMessages("Qwen/Qwen2.5-7B-Instruct",
+	resp, err := m.ChatWithMessages(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{},
 		&ChatConfig{MaxTokens: &maxTokens, Temperature: &temp}, nil)
@@ -145,6 +146,7 @@ func TestModelScopeChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T)
 }
 
 func TestModelScopeChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if got := r.Header.Get("Authorization"); got != "Bearer ms-test" {
 			t.Errorf("Authorization=%q, want Bearer ms-test", got)
@@ -155,7 +157,7 @@ func TestModelScopeChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 
 	m := newModelScopeForTest(srv.URL + "/v1")
 	key := "ms-test"
-	_, err := m.ChatWithMessages("Qwen/Qwen2.5-7B-Instruct",
+	_, err := m.ChatWithMessages(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil)
 	if err != nil {
@@ -164,6 +166,7 @@ func TestModelScopeChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 }
 
 func TestModelScopeChatExtractsReasoningFields(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"choices":[{"message":{
 			"content":"12",
@@ -173,7 +176,7 @@ func TestModelScopeChatExtractsReasoningFields(t *testing.T) {
 	defer srv.Close()
 
 	m := newModelScopeForTest(srv.URL)
-	resp, err := m.ChatWithMessages("Qwen/Qwen3-8B",
+	resp, err := m.ChatWithMessages(ctx, "Qwen/Qwen3-8B",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{}, nil, nil)
 	if err != nil {
@@ -185,6 +188,7 @@ func TestModelScopeChatExtractsReasoningFields(t *testing.T) {
 }
 
 func TestModelScopeStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -218,7 +222,7 @@ func TestModelScopeStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	var sawDone bool
-	err := m.ChatStreamlyWithSender("Qwen/Qwen2.5-7B-Instruct",
+	err := m.ChatStreamlyWithSender(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{}, nil, nil,
 		func(c *string, r *string) error {
@@ -248,9 +252,10 @@ func TestModelScopeStreamHappyPath(t *testing.T) {
 }
 
 func TestModelScopeStreamRejectsFalseStreamConfig(t *testing.T) {
+	ctx := t.Context()
 	m := newModelScopeForTest("http://unused")
 	stream := false
-	err := m.ChatStreamlyWithSender("Qwen/Qwen2.5-7B-Instruct",
+	err := m.ChatStreamlyWithSender(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{},
 		&ChatConfig{Stream: &stream},
@@ -262,6 +267,7 @@ func TestModelScopeStreamRejectsFalseStreamConfig(t *testing.T) {
 }
 
 func TestModelScopeStreamCancelsOnIdle(t *testing.T) {
+	ctx := t.Context()
 	withModelScopeIdleTimeout(t, 200*time.Millisecond)
 
 	hold := make(chan struct{})
@@ -281,7 +287,7 @@ func TestModelScopeStreamCancelsOnIdle(t *testing.T) {
 	t.Cleanup(func() { close(hold) })
 
 	m := newModelScopeForTest(srv.URL)
-	err := m.ChatStreamlyWithSender("Qwen/Qwen2.5-7B-Instruct",
+	err := m.ChatStreamlyWithSender(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil,
 		func(*string, *string) error { return nil })
@@ -291,6 +297,7 @@ func TestModelScopeStreamCancelsOnIdle(t *testing.T) {
 }
 
 func TestModelScopeListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/models" {
 			t.Errorf("path=%s, want /v1/models", r.URL.Path)
@@ -305,21 +312,22 @@ func TestModelScopeListModelsAndCheckConnection(t *testing.T) {
 	m := newModelScopeForTest(srv.URL)
 	key := "ms-test"
 	apiConfig := &APIConfig{ApiKey: &key}
-	models, err := m.ListModels(apiConfig)
+	models, err := m.ListModels(ctx, apiConfig)
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "Qwen/Qwen2.5-7B-Instruct,Qwen/Qwen3-8B" {
 		t.Errorf("models=%v", models)
 	}
-	if err := m.CheckConnection(apiConfig); err != nil {
+	if err := m.CheckConnection(ctx, apiConfig); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestModelScopeMissingBaseURLFailsClearly(t *testing.T) {
+	ctx := t.Context()
 	m := NewModelScopeModel(map[string]string{}, URLSuffix{Chat: "v1/chat/completions"})
-	_, err := m.ChatWithMessages("Qwen/Qwen2.5-7B-Instruct",
+	_, err := m.ChatWithMessages(ctx, "Qwen/Qwen2.5-7B-Instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "base URL") {
@@ -328,31 +336,32 @@ func TestModelScopeMissingBaseURLFailsClearly(t *testing.T) {
 }
 
 func TestModelScopeUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newModelScopeForTest("http://unused")
 	model := "Qwen/Qwen2.5-7B-Instruct"
 
-	if _, err := m.Embed(&model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: expected no such method, got %v", err)
 	}
-	if _, err := m.Rerank(&model, "q", []string{"d"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, &model, "q", []string{"d"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: expected no such method, got %v", err)
 	}
-	if _, err := m.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected no such method, got %v", err)
 	}
-	if _, err := m.TranscribeAudio(&model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: expected no such method, got %v", err)
 	}
-	if err := m.TranscribeAudioWithSender(&model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.TranscribeAudioWithSender(ctx, &model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudioWithSender: expected no such method, got %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: expected no such method, got %v", err)
 	}
-	if err := m.AudioSpeechWithSender(&model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.AudioSpeechWithSender(ctx, &model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeechWithSender: expected no such method, got %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: expected no such method, got %v", err)
 	}
 }
diff --git a/internal/entity/models/moonshot.go b/internal/entity/models/moonshot.go
index 3e789d1a0c..c6b7a1f22f 100644
--- a/internal/entity/models/moonshot.go
+++ b/internal/entity/models/moonshot.go
@@ -58,7 +58,7 @@ func validateMoonshotModelName(modelName string) (string, error) {
 	return strings.TrimSpace(modelName), nil
 }
 
-func (m *MoonshotModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (m *MoonshotModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -143,7 +143,7 @@ func (m *MoonshotModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -228,7 +228,7 @@ func (m *MoonshotModel) ChatWithMessages(modelName string, messages []Message, a
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (m *MoonshotModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MoonshotModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -320,7 +320,7 @@ func (m *MoonshotModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -407,11 +407,11 @@ func (m *MoonshotModel) ChatStreamlyWithSender(modelName string, messages []Mess
 }
 
 // Embed embeds a list of texts into embeddings
-func (m *MoonshotModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (m *MoonshotModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
-func (m *MoonshotModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (m *MoonshotModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -423,7 +423,7 @@ func (m *MoonshotModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, m.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -464,7 +464,7 @@ func (m *MoonshotModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	return models, nil
 }
 
-func (m *MoonshotModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (m *MoonshotModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := m.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -476,7 +476,7 @@ func (m *MoonshotModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 	}
 	url := fmt.Sprintf("%s/%s", baseURL, m.baseModel.URLSuffix.Balance)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -522,8 +522,8 @@ func (m *MoonshotModel) Balance(apiConfig *APIConfig) (map[string]interface{}, e
 	return response, nil
 }
 
-func (m *MoonshotModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := m.ListModels(apiConfig)
+func (m *MoonshotModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := m.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -531,42 +531,42 @@ func (m *MoonshotModel) CheckConnection(apiConfig *APIConfig) error {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (m *MoonshotModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (m *MoonshotModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", m.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (m *MoonshotModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (m *MoonshotModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MoonshotModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MoonshotModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
 // AudioSpeech convert text to audio
-func (m *MoonshotModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (m *MoonshotModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MoonshotModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (m *MoonshotModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", m.Name())
 }
 
 // OCRFile OCR file
-func (m *MoonshotModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (m *MoonshotModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
 // ParseFile parse file
-func (m *MoonshotModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (m *MoonshotModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MoonshotModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (m *MoonshotModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
 
-func (m *MoonshotModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (m *MoonshotModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", m.Name())
 }
diff --git a/internal/entity/models/moonshot_test.go b/internal/entity/models/moonshot_test.go
index cf48d520ca..bb6b563bc7 100644
--- a/internal/entity/models/moonshot_test.go
+++ b/internal/entity/models/moonshot_test.go
@@ -85,6 +85,7 @@ func TestMoonshotNewInstancePreservesConfig(t *testing.T) {
 }
 
 func TestMoonshotChatForcesNonStreaming(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -113,6 +114,7 @@ func TestMoonshotChatForcesNonStreaming(t *testing.T) {
 	stream := true
 	thinking := true
 	resp, err := newMoonshotForTest(srv.URL).ChatWithMessages(
+		ctx,
 		" kimi-k2.6 ",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -130,6 +132,7 @@ func TestMoonshotChatForcesNonStreaming(t *testing.T) {
 }
 
 func TestMoonshotChatSupportsTools(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		tools, ok := body["tools"].([]interface{})
 		if !ok || len(tools) != 1 {
@@ -164,6 +167,7 @@ func TestMoonshotChatSupportsTools(t *testing.T) {
 		},
 	}}
 	resp, err := newMoonshotForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"kimi-k2.6",
 		[]Message{{Role: "user", Content: "北京今天天气怎么样？"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -181,6 +185,7 @@ func TestMoonshotChatSupportsTools(t *testing.T) {
 }
 
 func TestMoonshotStreamForcesStreaming(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -209,6 +214,7 @@ func TestMoonshotStreamForcesStreaming(t *testing.T) {
 	var content, reasoning []string
 	var sawDone bool
 	err := newMoonshotForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"kimi-k2.6",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -243,6 +249,7 @@ func TestMoonshotStreamForcesStreaming(t *testing.T) {
 }
 
 func TestMoonshotStreamDoesNotSendDoneAfterScannerError(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["stream"] != true {
 			t.Errorf("stream=%v, want true", body["stream"])
@@ -255,6 +262,7 @@ func TestMoonshotStreamDoesNotSendDoneAfterScannerError(t *testing.T) {
 	apiKey := "test-key"
 	var sawDone bool
 	err := newMoonshotForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"kimi-k2.6",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -276,6 +284,7 @@ func TestMoonshotStreamDoesNotSendDoneAfterScannerError(t *testing.T) {
 }
 
 func TestMoonshotListModelsUsesBodylessGet(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -293,7 +302,7 @@ func TestMoonshotListModelsUsesBodylessGet(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newMoonshotForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newMoonshotForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -303,6 +312,7 @@ func TestMoonshotListModelsUsesBodylessGet(t *testing.T) {
 }
 
 func TestMoonshotBalanceUsesBodylessGet(t *testing.T) {
+	ctx := t.Context()
 	srv := newMoonshotServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -323,7 +333,7 @@ func TestMoonshotBalanceUsesBodylessGet(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	balance, err := newMoonshotForTest(srv.URL).Balance(&APIConfig{ApiKey: &apiKey})
+	balance, err := newMoonshotForTest(srv.URL).Balance(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("Balance: %v", err)
 	}
@@ -336,6 +346,7 @@ func TestMoonshotBalanceUsesBodylessGet(t *testing.T) {
 }
 
 func TestMoonshotRejectsMalformedResponses(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	tests := []struct {
 		name     string
@@ -346,7 +357,7 @@ func TestMoonshotRejectsMalformedResponses(t *testing.T) {
 			name:     "models missing data",
 			response: map[string]interface{}{"object": "list"},
 			run: func(m *MoonshotModel) error {
-				_, err := m.ListModels(&APIConfig{ApiKey: &apiKey})
+				_, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 				return err
 			},
 		},
@@ -356,7 +367,7 @@ func TestMoonshotRejectsMalformedResponses(t *testing.T) {
 				"data": []map[string]string{{"id": ""}},
 			},
 			run: func(m *MoonshotModel) error {
-				_, err := m.ListModels(&APIConfig{ApiKey: &apiKey})
+				_, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 				return err
 			},
 		},
@@ -366,7 +377,7 @@ func TestMoonshotRejectsMalformedResponses(t *testing.T) {
 				"data": map[string]float64{"cash_balance": 3},
 			},
 			run: func(m *MoonshotModel) error {
-				_, err := m.Balance(&APIConfig{ApiKey: &apiKey})
+				_, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey})
 				return err
 			},
 		},
@@ -387,6 +398,7 @@ func TestMoonshotRejectsMalformedResponses(t *testing.T) {
 }
 
 func TestMoonshotValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	emptyKey := " "
 	send := func(*string, *string) error { return nil }
@@ -399,7 +411,7 @@ func TestMoonshotValidatesInputs(t *testing.T) {
 		{
 			name: "chat api key",
 			run: func() error {
-				_, err := newMoonshotForTest("http://unused").ChatWithMessages("kimi-k2.6", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
+				_, err := newMoonshotForTest("http://unused").ChatWithMessages(ctx, "kimi-k2.6", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &emptyKey}, nil, nil)
 				return err
 			},
 			want: "api key is required",
@@ -407,7 +419,7 @@ func TestMoonshotValidatesInputs(t *testing.T) {
 		{
 			name: "chat model",
 			run: func() error {
-				_, err := newMoonshotForTest("http://unused").ChatWithMessages(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+				_, err := newMoonshotForTest("http://unused").ChatWithMessages(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
 			},
 			want: "model name is required",
@@ -415,28 +427,28 @@ func TestMoonshotValidatesInputs(t *testing.T) {
 		{
 			name: "stream api key",
 			run: func() error {
-				return newMoonshotForTest("http://unused").ChatStreamlyWithSender("kimi-k2.6", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
+				return newMoonshotForTest("http://unused").ChatStreamlyWithSender(ctx, "kimi-k2.6", []Message{{Role: "user", Content: "x"}}, nil, nil, nil, send)
 			},
 			want: "api key is required",
 		},
 		{
 			name: "stream model",
 			run: func() error {
-				return newMoonshotForTest("http://unused").ChatStreamlyWithSender(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, send)
+				return newMoonshotForTest("http://unused").ChatStreamlyWithSender(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, send)
 			},
 			want: "model name is required",
 		},
 		{
 			name: "stream sender",
 			run: func() error {
-				return newMoonshotForTest("http://unused").ChatStreamlyWithSender("kimi-k2.6", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
+				return newMoonshotForTest("http://unused").ChatStreamlyWithSender(ctx, "kimi-k2.6", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 			},
 			want: "sender is required",
 		},
 		{
 			name: "models api key",
 			run: func() error {
-				_, err := newMoonshotForTest("http://unused").ListModels(&APIConfig{})
+				_, err := newMoonshotForTest("http://unused").ListModels(ctx, &APIConfig{})
 				return err
 			},
 			want: "api key is required",
@@ -444,7 +456,7 @@ func TestMoonshotValidatesInputs(t *testing.T) {
 		{
 			name: "balance api key",
 			run: func() error {
-				_, err := newMoonshotForTest("http://unused").Balance(nil)
+				_, err := newMoonshotForTest("http://unused").Balance(ctx, &APIConfig{})
 				return err
 			},
 			want: "api key is required",
diff --git a/internal/entity/models/n1n.go b/internal/entity/models/n1n.go
index 7cc0f7e0a8..42c43dce05 100644
--- a/internal/entity/models/n1n.go
+++ b/internal/entity/models/n1n.go
@@ -159,7 +159,7 @@ type n1nChatResponse struct {
 
 // ChatWithMessages sends a single, non-streaming chat completion
 // against n1n.ai's /v1/chat/completions endpoint.
-func (n *N1NModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (n *N1NModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -178,7 +178,7 @@ func (n *N1NModel) ChatWithMessages(modelName string, messages []Message, apiCon
 
 	reqBody := buildN1NChatRequest(modelName, messages, false, chatModelConfig)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newN1NJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
@@ -224,7 +224,7 @@ func (n *N1NModel) ChatWithMessages(modelName string, messages []Message, apiCon
 }
 
 // ChatStreamlyWithSender sends a streaming chat completion.
-func (n *N1NModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *N1NModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -251,7 +251,7 @@ func (n *N1NModel) ChatStreamlyWithSender(modelName string, messages []Message,
 
 	reqBody := buildN1NChatRequest(modelName, messages, true, chatModelConfig)
 
-	req, err := newN1NJSONRequest(context.Background(), "POST", endpoint, reqBody, apiKey)
+	req, err := newN1NJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
 	if err != nil {
 		return err
 	}
@@ -322,7 +322,7 @@ type n1nEmbeddingRequest struct {
 }
 
 // Embed turns a list of texts into embedding vectors
-func (n *N1NModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (n *N1NModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -348,7 +348,7 @@ func (n *N1NModel) Embed(modelName *string, texts []string, apiConfig *APIConfig
 		reqBody.Dimensions = embeddingConfig.Dimension
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newN1NJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
@@ -416,7 +416,7 @@ type n1nRerankRequest struct {
 
 // Rerank scores a query against a list of documents using
 // n1n.ai's /v1/rerank endpoint (Cohere-shaped response).
-func (n *N1NModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (n *N1NModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -443,7 +443,7 @@ func (n *N1NModel) Rerank(modelName *string, query string, documents []string, a
 		reqBody.TopN = rerankConfig.TopN
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newN1NJSONRequest(ctx, "POST", endpoint, reqBody, apiKey)
@@ -497,7 +497,7 @@ type n1nModelCatalogResponse struct {
 }
 
 // ListModels returns the live n1n.ai model catalog
-func (n *N1NModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (n *N1NModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -508,7 +508,7 @@ func (n *N1NModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error)
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := newN1NJSONRequest(ctx, "GET", endpoint, nil, apiKey)
@@ -539,48 +539,48 @@ func (n *N1NModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error)
 }
 
 // CheckConnection verifies the API key
-func (n *N1NModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := n.ListModels(apiConfig)
+func (n *N1NModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := n.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (n *N1NModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (n *N1NModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
 // TranscribeAudio: n1n.ai exposes /v1/audio/transcriptions
-func (n *N1NModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (n *N1NModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *N1NModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *N1NModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
 // AudioSpeech: n1n.ai exposes /v1/audio/speech
-func (n *N1NModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (n *N1NModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *N1NModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *N1NModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
 // OCRFile is not exposed by the n1n.ai API.
-func (n *N1NModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (n *N1NModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
 // ParseFile is not exposed by the n1n.ai API.
-func (n *N1NModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (n *N1NModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
 // ListTasks: n1n.ai has /v1/contents/generations/tasks
-func (n *N1NModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (n *N1NModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *N1NModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (n *N1NModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
diff --git a/internal/entity/models/novita.go b/internal/entity/models/novita.go
index 3617924121..547a43fd08 100644
--- a/internal/entity/models/novita.go
+++ b/internal/entity/models/novita.go
@@ -188,7 +188,7 @@ func (e *novitaThinkExtractor) Flush() *novitaThinkSegment {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (n *NovitaModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (n *NovitaModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -249,7 +249,7 @@ func (n *NovitaModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -332,7 +332,7 @@ func (n *NovitaModel) ChatWithMessages(modelName string, messages []Message, api
 //     inline-style models): a stateful extractor splits tag bytes
 //     across SSE chunk boundaries, then routes content/reasoning to
 //     the first/second sender arg respectively.
-func (n *NovitaModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NovitaModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -392,7 +392,7 @@ func (n *NovitaModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -493,7 +493,7 @@ func (n *NovitaModel) ChatStreamlyWithSender(modelName string, messages []Messag
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (n *NovitaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (n *NovitaModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -505,7 +505,7 @@ func (n *NovitaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, n.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -542,8 +542,8 @@ func (n *NovitaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (n *NovitaModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := n.ListModels(apiConfig)
+func (n *NovitaModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := n.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -562,7 +562,7 @@ type novitaEmbeddingResponse struct {
 // Embed turns a list of texts into embedding vectors using the Novita
 // /v3/embeddings endpoint. The output has one vector per input, in the
 // same order the inputs were given.
-func (n *NovitaModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (n *NovitaModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -595,7 +595,7 @@ func (n *NovitaModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -666,7 +666,7 @@ type novitaRerankResponse struct {
 // /openai/v1/rerank endpoint and returns one RerankResult per scored
 // document in the API's ranking order. Caller may sort by Index to
 // recover original input order.
-func (n *NovitaModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (n *NovitaModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -705,7 +705,7 @@ func (n *NovitaModel) Rerank(modelName *string, query string, documents []string
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -756,7 +756,7 @@ func (n *NovitaModel) Rerank(modelName *string, query string, documents []string
 }
 
 // Balance Get remaining credit
-func (n *NovitaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (n *NovitaModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -826,36 +826,36 @@ func (n *NovitaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, err
 	return response, nil
 }
 
-func (n *NovitaModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (n *NovitaModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NovitaModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NovitaModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NovitaModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (n *NovitaModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NovitaModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NovitaModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
 // OCRFile OCR file
-func (n *NovitaModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (n *NovitaModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
 // ParseFile parse file
-func (n *NovitaModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (n *NovitaModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NovitaModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (n *NovitaModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NovitaModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (n *NovitaModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
diff --git a/internal/entity/models/novita_test.go b/internal/entity/models/novita_test.go
index e0ad544d86..b1a2eda284 100644
--- a/internal/entity/models/novita_test.go
+++ b/internal/entity/models/novita_test.go
@@ -255,6 +255,7 @@ func TestNovitaName(t *testing.T) {
 }
 
 func TestNovitaChatPureText(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -266,6 +267,7 @@ func TestNovitaChatPureText(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newNovitaForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"meta-llama/llama-3.3-70b-instruct",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -278,6 +280,7 @@ func TestNovitaChatPureText(t *testing.T) {
 }
 
 func TestNovitaChatExtractsThinkTags(t *testing.T) {
+	ctx := t.Context()
 	// qwen3-style response: <think>...</think> embedded in content.
 	// Driver must split it into Answer + ReasonContent.
 	srv := newNovitaServer(t, "/openai/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
@@ -294,6 +297,7 @@ func TestNovitaChatExtractsThinkTags(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newNovitaForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen/qwen3-30b-a3b-fp8",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -314,6 +318,7 @@ func TestNovitaChatExtractsThinkTags(t *testing.T) {
 // to ChatResponse.ReasonContent. Live-confirmed against
 // api.novita.ai/openai/v1/chat/completions with deepseek/deepseek-v3.1.
 func TestNovitaChatExtractsReasoningContentField(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -329,6 +334,7 @@ func TestNovitaChatExtractsReasoningContentField(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newNovitaForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-v3.1",
 		[]Message{{Role: "user", Content: "2+2?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -350,6 +356,7 @@ func TestNovitaChatExtractsReasoningContentField(t *testing.T) {
 // (not delta.content with <think> tags). The driver must forward
 // those chunks via the sender's second arg.
 func TestNovitaStreamExtractsDeltaReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"step 1. "}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"reasoning_content":"step 2."}}]}`+"\n"+
@@ -361,6 +368,7 @@ func TestNovitaStreamExtractsDeltaReasoningContent(t *testing.T) {
 	apiKey := "test-key"
 	var content, reasoning []string
 	err := newNovitaForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-v3.1",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -393,6 +401,7 @@ func TestNovitaStreamExtractsDeltaReasoningContent(t *testing.T) {
 // so a tenant can switch a deepseek-v3.1 / glm-4.5 / qwen3 deployment
 // out of its default thinking mode without prompt-level hacks.
 func TestNovitaChatPropagatesEnableThinking(t *testing.T) {
+	ctx := t.Context()
 	cases := []struct {
 		name  string
 		value bool
@@ -421,6 +430,7 @@ func TestNovitaChatPropagatesEnableThinking(t *testing.T) {
 			apiKey := "test-key"
 			thinking := tc.value
 			_, err := newNovitaForTest(srv.URL).ChatWithMessages(
+				ctx,
 				"qwen/qwen3-30b-a3b-fp8",
 				[]Message{{Role: "user", Content: "x"}},
 				&APIConfig{ApiKey: &apiKey},
@@ -436,6 +446,7 @@ func TestNovitaChatPropagatesEnableThinking(t *testing.T) {
 // silently flip behavior for downstream proxies that distinguish
 // "field absent" from "field present with default". Leave it out.
 func TestNovitaChatOmitsEnableThinkingWhenUnset(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if _, present := body["enable_thinking"]; present {
 			t.Errorf("enable_thinking must be absent when Thinking unset, got %v", body["enable_thinking"])
@@ -449,7 +460,7 @@ func TestNovitaChatOmitsEnableThinkingWhenUnset(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newNovitaForTest(srv.URL).ChatWithMessages("m",
+	_, err := newNovitaForTest(srv.URL).ChatWithMessages(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{}, nil) // no Thinking
@@ -462,6 +473,7 @@ func TestNovitaChatOmitsEnableThinkingWhenUnset(t *testing.T) {
 // case for ChatStreamlyWithSender so callers get the same toggle
 // regardless of streaming mode.
 func TestNovitaStreamPropagatesEnableThinking(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, _ := io.ReadAll(r.Body)
@@ -477,6 +489,7 @@ func TestNovitaStreamPropagatesEnableThinking(t *testing.T) {
 	apiKey := "test-key"
 	thinking := false
 	err := newNovitaForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-v3.1",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -492,7 +505,8 @@ func TestNovitaStreamPropagatesEnableThinking(t *testing.T) {
 }
 
 func TestNovitaChatRequiresAPIKey(t *testing.T) {
-	_, err := newNovitaForTest("http://unused").ChatWithMessages("m",
+	ctx := t.Context()
+	_, err := newNovitaForTest("http://unused").ChatWithMessages(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("got %v", err)
@@ -500,8 +514,9 @@ func TestNovitaChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestNovitaChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newNovitaForTest("http://unused").ChatWithMessages("m", nil,
+	_, err := newNovitaForTest("http://unused").ChatWithMessages(ctx, "m", nil,
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("got %v", err)
@@ -509,6 +524,7 @@ func TestNovitaChatRequiresMessages(t *testing.T) {
 }
 
 func TestNovitaChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"detail":"unauthorized"}`))
@@ -516,7 +532,7 @@ func TestNovitaChatRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newNovitaForTest(srv.URL).ChatWithMessages("m",
+	_, err := newNovitaForTest(srv.URL).ChatWithMessages(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
@@ -528,6 +544,7 @@ func TestNovitaChatRejectsHTTPError(t *testing.T) {
 // delta.content must surface reasoning chunks through the sender's
 // second arg, and visible content through the first.
 func TestNovitaStreamSplitsThinkTags(t *testing.T) {
+	ctx := t.Context()
 	// Simulate the realistic case where tags span deltas — split
 	// "<think>" across two chunks, and split "</think>" too.
 	srv := newNovitaSSEServer(t, "/openai/v1/chat/completions",
@@ -546,6 +563,7 @@ func TestNovitaStreamSplitsThinkTags(t *testing.T) {
 	apiKey := "test-key"
 	var content, reasoning []string
 	err := newNovitaForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen/qwen3-30b-a3b-fp8",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -577,6 +595,7 @@ func TestNovitaStreamSplitsThinkTags(t *testing.T) {
 // Streaming for a non-reasoning model that emits only content chunks
 // must continue to work unchanged.
 func TestNovitaStreamPureContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaSSEServer(t, "/openai/v1/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"content":"Hello "}}]}`+"\n"+
@@ -588,7 +607,7 @@ func TestNovitaStreamPureContent(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	var sawDone bool
-	err := newNovitaForTest(srv.URL).ChatStreamlyWithSender("meta-llama/llama-3.3-70b-instruct",
+	err := newNovitaForTest(srv.URL).ChatStreamlyWithSender(ctx, "meta-llama/llama-3.3-70b-instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(c *string, _ *string) error {
@@ -614,8 +633,9 @@ func TestNovitaStreamPureContent(t *testing.T) {
 }
 
 func TestNovitaStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	err := newNovitaForTest("http://unused").ChatStreamlyWithSender("m",
+	err := newNovitaForTest("http://unused").ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "sender is required") {
@@ -624,9 +644,10 @@ func TestNovitaStreamRequiresSender(t *testing.T) {
 }
 
 func TestNovitaStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
-	err := newNovitaForTest("http://unused").ChatStreamlyWithSender("m",
+	err := newNovitaForTest("http://unused").ChatStreamlyWithSender(ctx, "m",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Stream: &stream},
@@ -638,6 +659,7 @@ func TestNovitaStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestNovitaListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -650,7 +672,7 @@ func TestNovitaListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	ids, err := newNovitaForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	ids, err := newNovitaForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -660,18 +682,20 @@ func TestNovitaListModelsHappyPath(t *testing.T) {
 }
 
 func TestNovitaCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"data": []map[string]interface{}{{"id": "x"}}})
 	})
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newNovitaForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newNovitaForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 }
 
 func TestNovitaRerankHappyPathReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "baai/bge-reranker-v2-m3" {
 			t.Errorf("model=%v", body["model"])
@@ -692,7 +716,7 @@ func TestNovitaRerankHappyPathReordersByIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "baai/bge-reranker-v2-m3"
-	resp, err := newNovitaForTest(srv.URL).Rerank(&model, "what is rag", []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newNovitaForTest(srv.URL).Rerank(ctx, &model, "what is rag", []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
@@ -708,6 +732,7 @@ func TestNovitaRerankHappyPathReordersByIndex(t *testing.T) {
 }
 
 func TestNovitaRerankRespectsTopNConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["top_n"] != float64(2) {
 			t.Errorf("top_n=%v, want 2", body["top_n"])
@@ -718,15 +743,16 @@ func TestNovitaRerankRespectsTopNConfig(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "baai/bge-reranker-v2-m3"
-	if _, err := newNovitaForTest(srv.URL).Rerank(&model, "q", []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil); err != nil {
+	if _, err := newNovitaForTest(srv.URL).Rerank(ctx, &model, "q", []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil); err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
 }
 
 func TestNovitaRerankEmptyDocumentsShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := "x"
-	resp, err := newNovitaForTest("http://unused").Rerank(&model, "q", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newNovitaForTest("http://unused").Rerank(ctx, &model, "q", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("expected nil error for empty docs, got %v", err)
 	}
@@ -736,22 +762,25 @@ func TestNovitaRerankEmptyDocumentsShortCircuits(t *testing.T) {
 }
 
 func TestNovitaRerankRequiresApiKey(t *testing.T) {
+	ctx := t.Context()
 	model := "x"
-	_, err := newNovitaForTest("http://unused").Rerank(&model, "q", []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := newNovitaForTest("http://unused").Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaRerankRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newNovitaForTest("http://unused").Rerank(nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newNovitaForTest("http://unused").Rerank(ctx, nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaRerankRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_, _ = io.WriteString(w, `{"results":[{"index":5,"relevance_score":0.5}]}`)
 	})
@@ -759,13 +788,14 @@ func TestNovitaRerankRejectsOutOfRangeIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "x"
-	_, err := newNovitaForTest(srv.URL).Rerank(&model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newNovitaForTest(srv.URL).Rerank(ctx, &model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaRerankRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newNovitaServer(t, "/openai/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_, _ = io.WriteString(w, `{"results":[{"index":0,"relevance_score":0.9},{"index":0,"relevance_score":0.5}]}`)
 	})
@@ -773,13 +803,14 @@ func TestNovitaRerankRejectsDuplicateIndex(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "x"
-	_, err := newNovitaForTest(srv.URL).Rerank(&model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newNovitaForTest(srv.URL).Rerank(ctx, &model, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaRerankSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = io.WriteString(w, `{"error":"bad key"}`)
@@ -788,46 +819,49 @@ func TestNovitaRerankSurfacesHTTPError(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "x"
-	_, err := newNovitaForTest(srv.URL).Rerank(&model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newNovitaForTest(srv.URL).Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "Novita rerank API error") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaRerankRejectsMissingRerankSuffix(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := "x"
 	driver := NewNovitaModel(
 		map[string]string{"default": "http://unused"},
 		URLSuffix{Chat: "openai/v1/chat/completions"},
 	)
-	_, err := driver.Rerank(&model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := driver.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no rerank URL suffix configured") {
 		t.Errorf("got %v", err)
 	}
 }
 
 func TestNovitaBalanceReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	// Balance IS implemented (makes HTTP call), not a "no such method" stub.
 	// With dummy key it should reach the HTTP call stage, not fail at APIConfigCheck.
 	apiKey := "test-key"
-	_, err := newNovitaForTest("http://unused").Balance(&APIConfig{ApiKey: &apiKey})
+	_, err := newNovitaForTest("http://unused").Balance(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected non-api-key error (e.g. connection refused), got %v", err)
 	}
 }
 
 func TestNovitaAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := "x"
 	apiKey := "test-key"
 	v := newNovitaForTest("http://unused")
-	if _, err := v.TranscribeAudio(&m, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := v.TranscribeAudio(ctx, &m, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
-	if _, err := v.AudioSpeech(&m, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := v.AudioSpeech(ctx, &m, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: %v", err)
 	}
-	if _, err := v.OCRFile(&m, nil, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := v.OCRFile(ctx, &m, nil, &m, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
@@ -839,6 +873,7 @@ func TestNovitaAudioOCRReturnNoSuchMethod(t *testing.T) {
 // slash. baseURLForRegion now trims the trailing "/" so all three
 // endpoint builders (Chat, Stream, ListModels) emit clean paths.
 func TestNovitaBaseURLTrimsTrailingSlash(t *testing.T) {
+	ctx := t.Context()
 	cases := []struct {
 		name        string
 		path        string
@@ -854,7 +889,7 @@ func TestNovitaBaseURLTrimsTrailingSlash(t *testing.T) {
 			method:    http.MethodPost,
 			urlSuffix: URLSuffix{Chat: "openai/v1/chat/completions"},
 			invoke: func(n *NovitaModel, apiKey string) error {
-				_, err := n.ChatWithMessages("m",
+				_, err := n.ChatWithMessages(ctx, "m",
 					[]Message{{Role: "user", Content: "x"}},
 					&APIConfig{ApiKey: &apiKey}, nil, nil)
 				return err
@@ -867,7 +902,7 @@ func TestNovitaBaseURLTrimsTrailingSlash(t *testing.T) {
 			method:    http.MethodGet,
 			urlSuffix: URLSuffix{Models: "openai/v1/models"},
 			invoke: func(n *NovitaModel, apiKey string) error {
-				_, err := n.ListModels(&APIConfig{ApiKey: &apiKey})
+				_, err := n.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 				return err
 			},
 			respBody: `{"data":[]}`,
@@ -878,7 +913,7 @@ func TestNovitaBaseURLTrimsTrailingSlash(t *testing.T) {
 			method:    http.MethodPost,
 			urlSuffix: URLSuffix{Chat: "openai/v1/chat/completions"},
 			invoke: func(n *NovitaModel, apiKey string) error {
-				return n.ChatStreamlyWithSender("m",
+				return n.ChatStreamlyWithSender(ctx, "m",
 					[]Message{{Role: "user", Content: "x"}},
 					&APIConfig{ApiKey: &apiKey}, nil, nil,
 					func(*string, *string) error { return nil })
diff --git a/internal/entity/models/nvidia.go b/internal/entity/models/nvidia.go
index f699924cc6..d40923a5dd 100644
--- a/internal/entity/models/nvidia.go
+++ b/internal/entity/models/nvidia.go
@@ -51,7 +51,7 @@ func (n NvidiaModel) Name() string {
 	return "nvidia"
 }
 
-func (n *NvidiaModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (n *NvidiaModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -114,7 +114,7 @@ func (n *NvidiaModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -184,7 +184,7 @@ func (n *NvidiaModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (n *NvidiaModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NvidiaModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -252,7 +252,7 @@ func (n *NvidiaModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -324,7 +324,7 @@ type nvidiaEmbeddingResponse struct {
 	} `json:"data"`
 }
 
-func (n NvidiaModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (n NvidiaModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -364,7 +364,7 @@ func (n NvidiaModel) Embed(modelName *string, texts []string, apiConfig *APIConf
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -441,7 +441,7 @@ type nvidiaRerankResponse struct {
 // RerankResult entries are in the API's ranking order; callers that
 // need original-input order should sort by Index. Same return-shape
 // contract as the Aliyun and ZhipuAI Rerank drivers.
-func (n NvidiaModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (n NvidiaModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -487,7 +487,7 @@ func (n NvidiaModel) Rerank(modelName *string, query string, documents []string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -533,30 +533,30 @@ func (n NvidiaModel) Rerank(modelName *string, query string, documents []string,
 }
 
 // TranscribeAudio transcribe audio
-func (n *NvidiaModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (n *NvidiaModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NvidiaModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NvidiaModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
 // AudioSpeech convert text to audio
-func (n *NvidiaModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (n *NvidiaModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NvidiaModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (n *NvidiaModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", n.Name())
 }
 
 // OCRFile OCR file
-func (n *NvidiaModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (n *NvidiaModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
 // ParseFile parse file
-func (n *NvidiaModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (n *NvidiaModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
@@ -564,7 +564,7 @@ func (n *NvidiaModel) ParseFile(modelName *string, content []byte, url *string,
 // and returns the list of available model ids. The endpoint is
 // OpenAI-compatible, so the parsing follows the same shape used by
 // the moonshot, xai, and openai drivers.
-func (n NvidiaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (n NvidiaModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := n.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -580,7 +580,7 @@ func (n NvidiaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 
 	url := fmt.Sprintf("%s/%s", baseURL, n.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -617,7 +617,7 @@ func (n NvidiaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return ParseListModel(modelList), nil
 }
 
-func (n NvidiaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (n NvidiaModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
@@ -625,15 +625,15 @@ func (n NvidiaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, erro
 // is reachable and that the API key is accepted, by issuing a
 // lightweight ListModels call. Mirrors the pattern used by the xai,
 // moonshot, deepseek, aliyun, and gitee drivers.
-func (n NvidiaModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := n.ListModels(apiConfig)
+func (n NvidiaModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := n.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (n *NvidiaModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (n *NvidiaModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
 
-func (n *NvidiaModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (n *NvidiaModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", n.Name())
 }
diff --git a/internal/entity/models/nvidia_rerank_test.go b/internal/entity/models/nvidia_rerank_test.go
index 10edee00dc..ad3f95ddfb 100644
--- a/internal/entity/models/nvidia_rerank_test.go
+++ b/internal/entity/models/nvidia_rerank_test.go
@@ -52,6 +52,7 @@ func newNvidiaModelForTest(baseURL string) *NvidiaModel {
 }
 
 func TestNvidiaRerankHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newNvidiaRerankServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "nvidia/nv-rerankqa-mistral-4b-v3" {
 			t.Errorf("expected model=nvidia/nv-rerankqa-mistral-4b-v3, got %v", body["model"])
@@ -86,6 +87,7 @@ func TestNvidiaRerankHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
 	resp, err := model.Rerank(
+		ctx,
 		&modelName,
 		"What is RAPTOR?",
 		[]string{"doc-zero", "doc-one", "doc-two"},
@@ -108,6 +110,7 @@ func TestNvidiaRerankHappyPath(t *testing.T) {
 }
 
 func TestNvidiaRerankTopNClamp(t *testing.T) {
+	ctx := t.Context()
 	srv := newNvidiaRerankServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["top_n"] != float64(2) {
 			t.Errorf("expected top_n clamp to RerankConfig.TopN=2, got %v", body["top_n"])
@@ -120,6 +123,7 @@ func TestNvidiaRerankTopNClamp(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
 	if _, err := model.Rerank(
+		ctx,
 		&modelName, "q",
 		[]string{"a", "b", "c", "d"},
 		&APIConfig{ApiKey: &apiKey},
@@ -131,10 +135,11 @@ func TestNvidiaRerankTopNClamp(t *testing.T) {
 }
 
 func TestNvidiaRerankEmptyDocuments(t *testing.T) {
+	ctx := t.Context()
 	model := newNvidiaModelForTest("http://unused")
 	apiKey := "test-key"
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
-	resp, err := model.Rerank(&modelName, "q", nil, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	resp, err := model.Rerank(ctx, &modelName, "q", nil, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err != nil {
 		t.Fatalf("expected nil error for empty documents, got %v", err)
 	}
@@ -144,24 +149,27 @@ func TestNvidiaRerankEmptyDocuments(t *testing.T) {
 }
 
 func TestNvidiaRerankRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	model := newNvidiaModelForTest("http://unused")
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
-	_, err := model.Rerank(&modelName, "q", []string{"a"}, &APIConfig{}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, &modelName, "q", []string{"a"}, &APIConfig{}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestNvidiaRerankRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	model := newNvidiaModelForTest("http://unused")
 	apiKey := "test-key"
-	_, err := model.Rerank(nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestNvidiaRerankRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newNvidiaRerankServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"unauthorized"}`))
@@ -171,13 +179,14 @@ func TestNvidiaRerankRejectsHTTPError(t *testing.T) {
 	model := newNvidiaModelForTest(srv.URL)
 	apiKey := "test-key"
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
-	_, err := model.Rerank(&modelName, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, &modelName, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "Nvidia rerank API error") {
 		t.Errorf("expected API error, got %v", err)
 	}
 }
 
 func TestNvidiaRerankRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newNvidiaRerankServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"rankings": []map[string]interface{}{
@@ -190,7 +199,7 @@ func TestNvidiaRerankRejectsOutOfRangeIndex(t *testing.T) {
 	model := newNvidiaModelForTest(srv.URL)
 	apiKey := "test-key"
 	modelName := "nvidia/nv-rerankqa-mistral-4b-v3"
-	_, err := model.Rerank(&modelName, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, &modelName, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "unexpected rerank index") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
diff --git a/internal/entity/models/ollama.go b/internal/entity/models/ollama.go
index e04e02f89c..b917e8fafd 100644
--- a/internal/entity/models/ollama.go
+++ b/internal/entity/models/ollama.go
@@ -73,7 +73,7 @@ func (o *OllamaModel) Name() string {
 	return "Ollama"
 }
 
-func (o *OllamaModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (o *OllamaModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if len(messages) == 0 {
 		return nil, fmt.Errorf("message is nil")
 	}
@@ -144,7 +144,7 @@ func (o *OllamaModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -195,7 +195,7 @@ func (o *OllamaModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (o *OllamaModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OllamaModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if len(messages) == 0 {
 		return fmt.Errorf("messages is empty")
 	}
@@ -265,7 +265,7 @@ func (o *OllamaModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -330,7 +330,7 @@ func (o *OllamaModel) ChatStreamlyWithSender(modelName string, messages []Messag
 	return scanner.Err()
 }
 
-func (o *OllamaModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (o *OllamaModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -370,7 +370,7 @@ func (o *OllamaModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -427,39 +427,39 @@ func (o *OllamaModel) Embed(modelName *string, texts []string, apiConfig *APICon
 	return embeddings, nil
 }
 
-func (o *OllamaModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (o *OllamaModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (o *OllamaModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (o *OllamaModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OllamaModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OllamaModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", o.Name())
 }
 
 // AudioSpeech convert text to audio
-func (o *OllamaModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (o *OllamaModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OllamaModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OllamaModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", o.Name())
 }
 
 // OCRFile OCR file
-func (o *OllamaModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (o *OllamaModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
 // ParseFile parse file
-func (o *OllamaModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (o *OllamaModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OllamaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (o *OllamaModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 
 	baseURL, err := o.baseModel.GetBaseURL(apiConfig)
 	if err != nil {
@@ -471,7 +471,7 @@ func (o *OllamaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), o.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -526,20 +526,20 @@ func (o *OllamaModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	return ParseListModel(modelList), nil
 }
 
-func (o *OllamaModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (o *OllamaModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection verifies that the configured Ollama base URL is reachable
-func (o *OllamaModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := o.ListModels(apiConfig)
+func (o *OllamaModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := o.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (o *OllamaModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (o *OllamaModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OllamaModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (o *OllamaModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
diff --git a/internal/entity/models/ollama_test.go b/internal/entity/models/ollama_test.go
index cdfc2066de..daf65d9e69 100644
--- a/internal/entity/models/ollama_test.go
+++ b/internal/entity/models/ollama_test.go
@@ -28,6 +28,7 @@ func newOllamaForListModelsTest(baseURL string) *OllamaModel {
 }
 
 func TestOllamaListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -40,7 +41,7 @@ func TestOllamaListModels(t *testing.T) {
 	}))
 	defer srv.Close()
 
-	models, err := newOllamaForListModelsTest(srv.URL).ListModels(&APIConfig{})
+	models, err := newOllamaForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -53,13 +54,14 @@ func TestOllamaListModels(t *testing.T) {
 }
 
 func TestOllamaListModelsFallsBackToModelField(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		// Some entries may carry only the "model" field; it should be used as the name.
 		_, _ = io.WriteString(w, `{"models":[{"model":"phi3:mini"},{"name":""},{"name":"  "}]}`)
 	}))
 	defer srv.Close()
 
-	models, err := newOllamaForListModelsTest(srv.URL).ListModels(&APIConfig{})
+	models, err := newOllamaForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -72,20 +74,22 @@ func TestOllamaListModelsFallsBackToModelField(t *testing.T) {
 }
 
 func TestOllamaListModelsRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusInternalServerError)
 		_, _ = io.WriteString(w, `boom`)
 	}))
 	defer srv.Close()
 
-	if _, err := newOllamaForListModelsTest(srv.URL).ListModels(&APIConfig{}); err == nil {
+	if _, err := newOllamaForListModelsTest(srv.URL).ListModels(ctx, &APIConfig{}); err == nil {
 		t.Fatal("ListModels: expected error for HTTP 500, got nil")
 	}
 }
 
 func TestOllamaListModelsRequiresBaseURL(t *testing.T) {
+	ctx := t.Context()
 	m := NewOllamaModel(map[string]string{}, URLSuffix{Models: "api/tags"})
-	if _, err := m.ListModels(&APIConfig{}); err == nil {
+	if _, err := m.ListModels(ctx, &APIConfig{}); err == nil {
 		t.Fatal("ListModels: expected error for missing base URL, got nil")
 	}
 }
diff --git a/internal/entity/models/openai.go b/internal/entity/models/openai.go
index 22f47f8d91..b3769c0d2d 100644
--- a/internal/entity/models/openai.go
+++ b/internal/entity/models/openai.go
@@ -58,7 +58,7 @@ func (o *OpenAIModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response
-func (o *OpenAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (o *OpenAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -139,7 +139,7 @@ func (o *OpenAIModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -230,7 +230,7 @@ func (o *OpenAIModel) ChatWithMessages(modelName string, messages []Message, api
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (o *OpenAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -320,7 +320,7 @@ func (o *OpenAIModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -455,7 +455,7 @@ type openaiUsage struct {
 	TotalTokens  int `json:"total_tokens"`
 }
 
-func (o *OpenAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (o *OpenAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -488,7 +488,7 @@ func (o *OpenAIModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -531,7 +531,7 @@ func (o *OpenAIModel) Embed(modelName *string, texts []string, apiConfig *APICon
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (o *OpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (o *OpenAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -543,7 +543,7 @@ func (o *OpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, o.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -582,13 +582,13 @@ func (o *OpenAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 }
 
 // Balance is not exposed by the OpenAI API, so this returns "no such method".
-func (o *OpenAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (o *OpenAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (o *OpenAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := o.ListModels(apiConfig)
+func (o *OpenAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := o.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -597,13 +597,13 @@ func (o *OpenAIModel) CheckConnection(apiConfig *APIConfig) error {
 
 // Rerank calculates similarity scores between query and documents. OpenAI does
 // not expose a rerank API, so this is left unimplemented.
-func (o *OpenAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (o *OpenAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", o.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (o *OpenAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+func (o *OpenAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, responseFormat, err := o.newOpenAIASRRequest(ctx, modelName, file, apiConfig, asrConfig, false)
@@ -631,12 +631,12 @@ func (o *OpenAIModel) TranscribeAudio(modelName *string, file *string, apiConfig
 	return decodeOpenAIASRResponse(respBody, responseFormat)
 }
 
-func (o *OpenAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	req, responseFormat, err := o.newOpenAIASRRequest(context.Background(), modelName, file, apiConfig, asrConfig, true)
+	req, responseFormat, err := o.newOpenAIASRRequest(ctx, modelName, file, apiConfig, asrConfig, true)
 	if err != nil {
 		return err
 	}
@@ -723,8 +723,8 @@ func decodeOpenAIASRResponse(respBody []byte, responseFormat string) (*ASRRespon
 }
 
 // AudioSpeech convert text to audio
-func (o *OpenAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+func (o *OpenAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, _, err := o.newOpenAITTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, false)
@@ -750,12 +750,12 @@ func (o *OpenAIModel) AudioSpeech(modelName *string, audioContent *string, apiCo
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (o *OpenAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	req, streamFormat, err := o.newOpenAITTSRequest(context.Background(), modelName, audioContent, apiConfig, ttsConfig, true)
+	req, streamFormat, err := o.newOpenAITTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, true)
 	if err != nil {
 		return err
 	}
@@ -1014,20 +1014,20 @@ func writeOpenAIMultipartField(writer *multipart.Writer, key string, value inter
 }
 
 // OCRFile OCR file
-func (o *OpenAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (o *OpenAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
 // ParseFile parse file
-func (o *OpenAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (o *OpenAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OpenAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (o *OpenAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OpenAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (o *OpenAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
diff --git a/internal/entity/models/openai_test.go b/internal/entity/models/openai_test.go
index b7c0ceb4a3..a717c29810 100644
--- a/internal/entity/models/openai_test.go
+++ b/internal/entity/models/openai_test.go
@@ -68,6 +68,7 @@ func TestOpenAIConfigAdvertisedAudioModelsHaveSuffixes(t *testing.T) {
 }
 
 func TestOpenAITranscribeAudioPostsMultipartToAudioEndpoint(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -126,6 +127,7 @@ func TestOpenAITranscribeAudioPostsMultipartToAudioEndpoint(t *testing.T) {
 	apiKey := "test-key"
 	model := "whisper-1"
 	resp, err := newOpenAIForTest(srv.URL).TranscribeAudio(
+		ctx,
 		&model,
 		&audioPath,
 		&APIConfig{ApiKey: &apiKey},
@@ -144,6 +146,7 @@ func TestOpenAITranscribeAudioPostsMultipartToAudioEndpoint(t *testing.T) {
 }
 
 func TestOpenAITranscribeAudioWithSenderStreamsDeltas(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -189,6 +192,7 @@ func TestOpenAITranscribeAudioWithSenderStreamsDeltas(t *testing.T) {
 	model := "gpt-4o-mini-transcribe"
 	var chunks []string
 	err := newOpenAIForTest(srv.URL).TranscribeAudioWithSender(
+		ctx,
 		&model,
 		&audioPath,
 		&APIConfig{ApiKey: &apiKey},
@@ -210,6 +214,7 @@ func TestOpenAITranscribeAudioWithSenderStreamsDeltas(t *testing.T) {
 }
 
 func TestOpenAIAudioSpeechPostsJSONToAudioEndpoint(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s, want POST", r.Method)
@@ -254,6 +259,7 @@ func TestOpenAIAudioSpeechPostsJSONToAudioEndpoint(t *testing.T) {
 	model := "tts-1"
 	input := "hello"
 	resp, err := newOpenAIForTest(srv.URL).AudioSpeech(
+		ctx,
 		&model,
 		&input,
 		&APIConfig{ApiKey: &apiKey},
@@ -278,8 +284,10 @@ func TestOpenAIAudioSpeechRequiresVoice(t *testing.T) {
 	apiKey := "test-key"
 	model := "tts-1"
 	input := "hello"
+	ctx := t.Context()
 
 	_, err := newOpenAIForTest("http://unused").AudioSpeech(
+		ctx,
 		&model,
 		&input,
 		&APIConfig{ApiKey: &apiKey},
@@ -292,11 +300,13 @@ func TestOpenAIAudioSpeechRequiresVoice(t *testing.T) {
 }
 
 func TestOpenAIAudioSpeechRejectsNonStringVoice(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := "tts-1"
 	input := "hello"
 
 	_, err := newOpenAIForTest("http://unused").AudioSpeech(
+		ctx,
 		&model,
 		&input,
 		&APIConfig{ApiKey: &apiKey},
@@ -347,9 +357,11 @@ func TestOpenAIAudioSpeechWithSenderStreamsRawAudio(t *testing.T) {
 	apiKey := "test-key"
 	model := "tts-1"
 	input := "hello"
+	ctx := t.Context()
 
 	var chunks []string
 	err := newOpenAIForTest(srv.URL).AudioSpeechWithSender(
+		ctx,
 		&model,
 		&input,
 		&APIConfig{ApiKey: &apiKey},
@@ -371,6 +383,7 @@ func TestOpenAIAudioSpeechWithSenderStreamsRawAudio(t *testing.T) {
 }
 
 func TestOpenAIAudioSpeechWithSenderStreamsSSEAudioDeltas(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if got := r.Header.Get("Accept"); got != "text/event-stream" {
 			t.Errorf("Accept=%q, want text/event-stream", got)
@@ -397,6 +410,7 @@ func TestOpenAIAudioSpeechWithSenderStreamsSSEAudioDeltas(t *testing.T) {
 	input := "hello"
 	var chunks []string
 	err := newOpenAIForTest(srv.URL).AudioSpeechWithSender(
+		ctx,
 		&model,
 		&input,
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/openrouter.go b/internal/entity/models/openrouter.go
index 638aa6cd26..965fce3819 100644
--- a/internal/entity/models/openrouter.go
+++ b/internal/entity/models/openrouter.go
@@ -54,7 +54,7 @@ func (o *OpenRouterModel) Name() string {
 	return "openrouter"
 }
 
-func (o *OpenRouterModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (o *OpenRouterModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -118,7 +118,7 @@ func (o *OpenRouterModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -189,7 +189,7 @@ func (o *OpenRouterModel) ChatWithMessages(modelName string, messages []Message,
 	return chatResponse, nil
 }
 
-func (o *OpenRouterModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenRouterModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -267,7 +267,7 @@ func (o *OpenRouterModel) ChatStreamlyWithSender(modelName string, messages []Me
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -353,7 +353,7 @@ type openrouterUsage struct {
 	TotalTokens  int `json:"total_tokens"`
 }
 
-func (o *OpenRouterModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (o *OpenRouterModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -385,7 +385,7 @@ func (o *OpenRouterModel) Embed(modelName *string, texts []string, apiConfig *AP
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -448,7 +448,7 @@ type OpenRouterRerankResponse struct {
 	} `json:"results"`
 }
 
-func (o *OpenRouterModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (o *OpenRouterModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -480,7 +480,7 @@ func (o *OpenRouterModel) Rerank(modelName *string, query string, documents []st
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(resolvedBaseURL, "/"), o.baseModel.URLSuffix.Rerank)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -543,7 +543,7 @@ func openRouterAudioFormat(file string, asrConfig *ASRConfig) string {
 }
 
 // TranscribeAudio transcribe audio
-func (o *OpenRouterModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (o *OpenRouterModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -591,7 +591,7 @@ func (o *OpenRouterModel) TranscribeAudio(modelName *string, file *string, apiCo
 		return nil, err
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(resolvedBaseURL, "/"), o.baseModel.URLSuffix.ASR)
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -625,12 +625,12 @@ func (o *OpenRouterModel) TranscribeAudio(modelName *string, file *string, apiCo
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (o *OpenRouterModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenRouterModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", o.Name())
 }
 
 // AudioSpeech convert text to audio
-func (o *OpenRouterModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (o *OpenRouterModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -664,7 +664,7 @@ func (o *OpenRouterModel) AudioSpeech(modelName *string, audioContent *string, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -693,21 +693,21 @@ func (o *OpenRouterModel) AudioSpeech(modelName *string, audioContent *string, a
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (o *OpenRouterModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OpenRouterModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", o.Name())
 }
 
 // OCRFile OCR file
-func (o *OpenRouterModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (o *OpenRouterModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
 // ParseFile parse file
-func (o *OpenRouterModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (o *OpenRouterModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OpenRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (o *OpenRouterModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -726,7 +726,7 @@ func (o *OpenRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -765,7 +765,7 @@ func (o *OpenRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(modelList), nil
 }
 
-func (o *OpenRouterModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (o *OpenRouterModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -776,7 +776,7 @@ func (o *OpenRouterModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 	}
 	url := fmt.Sprintf("%s/%s", baseURL, o.baseModel.URLSuffix.Balance)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -823,15 +823,15 @@ func (o *OpenRouterModel) Balance(apiConfig *APIConfig) (map[string]interface{},
 	}, nil
 }
 
-func (o *OpenRouterModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := o.Balance(apiConfig)
+func (o *OpenRouterModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := o.Balance(ctx, apiConfig)
 	return err
 }
 
-func (o *OpenRouterModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (o *OpenRouterModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
 
-func (o *OpenRouterModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (o *OpenRouterModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", o.Name())
 }
diff --git a/internal/entity/models/openrouter_test.go b/internal/entity/models/openrouter_test.go
index 3aef619eb7..e3c51f695a 100644
--- a/internal/entity/models/openrouter_test.go
+++ b/internal/entity/models/openrouter_test.go
@@ -77,6 +77,7 @@ func TestOpenRouterAudioFormatUsesConfiguredValue(t *testing.T) {
 }
 
 func TestOpenRouterTranscribeAudioHappyPath(t *testing.T) {
+	ctx := t.Context()
 	audio := []byte("RIFF test audio")
 	srv := newOpenRouterServer(t, "/audio/transcriptions", func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodPost {
@@ -120,6 +121,7 @@ func TestOpenRouterTranscribeAudioHappyPath(t *testing.T) {
 	modelName := "openai/whisper-large-v3"
 	file := writeOpenRouterAudioFile(t, "sample.wav", audio)
 	resp, err := newOpenRouterForTest(srv.URL).TranscribeAudio(
+		ctx,
 		&modelName,
 		&file,
 		&APIConfig{ApiKey: &apiKey},
@@ -141,6 +143,7 @@ func TestOpenRouterTranscribeAudioHappyPath(t *testing.T) {
 }
 
 func TestOpenRouterTranscribeAudioInfersFormat(t *testing.T) {
+	ctx := t.Context()
 	srv := newOpenRouterServer(t, "/audio/transcriptions", func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		inputAudio, ok := body["input_audio"].(map[string]interface{})
 		if !ok {
@@ -157,13 +160,14 @@ func TestOpenRouterTranscribeAudioInfersFormat(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "openai/whisper-large-v3"
 	file := writeOpenRouterAudioFile(t, "sample.mp3", []byte("audio"))
-	_, err := newOpenRouterForTest(srv.URL).TranscribeAudio(&modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newOpenRouterForTest(srv.URL).TranscribeAudio(ctx, &modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("TranscribeAudio: %v", err)
 	}
 }
 
 func TestOpenRouterTranscribeAudioValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	modelName := "openai/whisper-large-v3"
 	apiKey := "test-key"
 	file := "sample.wav"
@@ -182,7 +186,7 @@ func TestOpenRouterTranscribeAudioValidatesInputs(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			_, err := newOpenRouterForTest("http://unused").TranscribeAudio(tt.modelName, tt.file, tt.apiConfig, nil, nil)
+			_, err := newOpenRouterForTest("http://unused").TranscribeAudio(ctx, tt.modelName, tt.file, tt.apiConfig, nil, nil)
 			if err == nil || !strings.Contains(err.Error(), tt.want) {
 				t.Fatalf("err=%v, want %q", err, tt.want)
 			}
@@ -191,18 +195,20 @@ func TestOpenRouterTranscribeAudioValidatesInputs(t *testing.T) {
 }
 
 func TestOpenRouterTranscribeAudioValidatesASRSuffix(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "openai/whisper-large-v3"
 	file := "sample.wav"
 	model := NewOpenRouterModel(map[string]string{"default": "http://unused"}, URLSuffix{})
 
-	_, err := model.TranscribeAudio(&modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := model.TranscribeAudio(ctx, &modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "OpenRouter ASR url suffix is missing") {
 		t.Fatalf("err=%v", err)
 	}
 }
 
 func TestOpenRouterTranscribeAudioHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newOpenRouterServer(t, "/audio/transcriptions", func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		http.Error(w, `{"error":"unauthorized"}`, http.StatusUnauthorized)
 	})
@@ -211,7 +217,7 @@ func TestOpenRouterTranscribeAudioHTTPError(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "openai/whisper-large-v3"
 	file := writeOpenRouterAudioFile(t, "sample.wav", []byte("audio"))
-	_, err := newOpenRouterForTest(srv.URL).TranscribeAudio(&modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newOpenRouterForTest(srv.URL).TranscribeAudio(ctx, &modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil ||
 		!strings.Contains(err.Error(), "OpenRouter ASR API error") ||
 		!strings.Contains(err.Error(), "401 Unauthorized") ||
@@ -225,6 +231,7 @@ func TestOpenRouterTranscribeAudioHTTPError(t *testing.T) {
 // and reasoning is requested via OpenRouter's standard `reasoning` object rather
 // than the non-standard `thinking` key that the API silently ignores.
 func TestOpenRouterChatStreamlyRequest(t *testing.T) {
+	ctx := t.Context()
 	var gotPath string
 	var gotBody map[string]interface{}
 	srv := newOpenRouterServer(t, "/chat/completions", func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
@@ -242,6 +249,7 @@ func TestOpenRouterChatStreamlyRequest(t *testing.T) {
 	var reasoning strings.Builder
 	// "qwen_max" would have triggered the removed async-routing branch (empty suffix).
 	err := newOpenRouterForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen_max",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/orcarouter.go b/internal/entity/models/orcarouter.go
index 27d2d185f1..67e022e4ea 100644
--- a/internal/entity/models/orcarouter.go
+++ b/internal/entity/models/orcarouter.go
@@ -48,7 +48,7 @@ func (o *OrcaRouterModel) Name() string {
 	return "orcarouter"
 }
 
-func (o *OrcaRouterModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (o *OrcaRouterModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -106,7 +106,7 @@ func (o *OrcaRouterModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -168,7 +168,7 @@ func (o *OrcaRouterModel) ChatWithMessages(modelName string, messages []Message,
 	return chatResponse, nil
 }
 
-func (o *OrcaRouterModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OrcaRouterModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -226,7 +226,7 @@ func (o *OrcaRouterModel) ChatStreamlyWithSender(modelName string, messages []Me
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -296,23 +296,23 @@ func (o *OrcaRouterModel) ChatStreamlyWithSender(modelName string, messages []Me
 	return nil
 }
 
-func (o *OrcaRouterModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (o *OrcaRouterModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (o *OrcaRouterModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (o *OrcaRouterModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OrcaRouterModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (o *OrcaRouterModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -347,7 +347,7 @@ func (o *OrcaRouterModel) AudioSpeech(modelName *string, audioContent *string, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -376,19 +376,19 @@ func (o *OrcaRouterModel) AudioSpeech(modelName *string, audioContent *string, a
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (o *OrcaRouterModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (o *OrcaRouterModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (o *OrcaRouterModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (o *OrcaRouterModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (o *OrcaRouterModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := o.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -406,7 +406,7 @@ func (o *OrcaRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -445,19 +445,19 @@ func (o *OrcaRouterModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(modelList), nil
 }
 
-func (o *OrcaRouterModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (o *OrcaRouterModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := o.ListModels(apiConfig)
+func (o *OrcaRouterModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := o.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (o *OrcaRouterModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (o *OrcaRouterModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
 
-func (o *OrcaRouterModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (o *OrcaRouterModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", o.Name())
 }
diff --git a/internal/entity/models/paddleocr.go b/internal/entity/models/paddleocr.go
index 51a11cf39e..77e6c664a7 100644
--- a/internal/entity/models/paddleocr.go
+++ b/internal/entity/models/paddleocr.go
@@ -53,35 +53,35 @@ func (p *PaddleOCRModel) Name() string {
 	return "paddle_ocr.net"
 }
 
-func (p *PaddleOCRModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (p *PaddleOCRModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (p *PaddleOCRModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (p *PaddleOCRModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (p *PaddleOCRModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (p *PaddleOCRModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
@@ -116,7 +116,7 @@ type paddleJsonlLine struct {
 	} `json:"result"`
 }
 
-func (p *PaddleOCRModel) OCRFile(modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (p *PaddleOCRModel) OCRFile(ctx context.Context, modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -140,7 +140,7 @@ func (p *PaddleOCRModel) OCRFile(modelName *string, content []byte, fileURL *str
 
 	// One generous deadline bounds the whole OCR operation (submit + poll +
 	// result download), so the poll loop below can no longer spin forever.
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	var req *http.Request
@@ -319,26 +319,26 @@ func (p *PaddleOCRModel) OCRFile(modelName *string, content []byte, fileURL *str
 	return &OCRFileResponse{Text: &extractedText}, nil
 }
 
-func (p *PaddleOCRModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (p *PaddleOCRModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (p *PaddleOCRModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (p *PaddleOCRModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) CheckConnection(apiConfig *APIConfig) error {
+func (p *PaddleOCRModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (p *PaddleOCRModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PaddleOCRModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (p *PaddleOCRModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
diff --git a/internal/entity/models/paddleocr_local.go b/internal/entity/models/paddleocr_local.go
index 8e78554677..d14cf9bf46 100644
--- a/internal/entity/models/paddleocr_local.go
+++ b/internal/entity/models/paddleocr_local.go
@@ -50,35 +50,35 @@ func (p *PaddleOCRLocalModel) Name() string {
 	return "paddleocr"
 }
 
-func (p *PaddleOCRLocalModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (p *PaddleOCRLocalModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRLocalModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (p *PaddleOCRLocalModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (p *PaddleOCRLocalModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (p *PaddleOCRLocalModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRLocalModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (p *PaddleOCRLocalModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PaddleOCRLocalModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", p.Name())
 }
 
@@ -98,7 +98,7 @@ type paddleLocalOCRResponse struct {
 	} `json:"result"`
 }
 
-func (p *PaddleOCRLocalModel) OCRFile(modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (p *PaddleOCRLocalModel) OCRFile(ctx context.Context, modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if len(content) == 0 {
 		return nil, fmt.Errorf("local PaddleOCR requires file content, but content is empty")
 	}
@@ -133,7 +133,7 @@ func (p *PaddleOCRLocalModel) OCRFile(modelName *string, content []byte, fileURL
 		return nil, fmt.Errorf("failed to marshal local PaddleOCR request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -185,26 +185,26 @@ func (p *PaddleOCRLocalModel) OCRFile(modelName *string, content []byte, fileURL
 	}, nil
 }
 
-func (p *PaddleOCRLocalModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (p *PaddleOCRLocalModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (p *PaddleOCRLocalModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (p *PaddleOCRLocalModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) CheckConnection(apiConfig *APIConfig) error {
+func (p *PaddleOCRLocalModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (p *PaddleOCRLocalModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
 
-func (p *PaddleOCRLocalModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (p *PaddleOCRLocalModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", p.Name())
 }
diff --git a/internal/entity/models/perplexity.go b/internal/entity/models/perplexity.go
index 6f10d9c074..601f72fb39 100644
--- a/internal/entity/models/perplexity.go
+++ b/internal/entity/models/perplexity.go
@@ -114,7 +114,7 @@ type perplexityChatResponse struct {
 	FinishReason string                 `json:"finish_reason"`
 }
 
-func (p *PerplexityModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (p *PerplexityModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -135,7 +135,7 @@ func (p *PerplexityModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -181,7 +181,7 @@ func (p *PerplexityModel) ChatWithMessages(modelName string, messages []Message,
 	}, nil
 }
 
-func (p *PerplexityModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PerplexityModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -212,7 +212,7 @@ func (p *PerplexityModel) ChatStreamlyWithSender(modelName string, messages []Me
 	// ResponseHeaderTimeout caps the initial header wait. This context
 	// also caps the body-read phase so a stalled SSE stream cannot hold
 	// the caller's goroutine and connection indefinitely.
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -284,7 +284,7 @@ type perplexityModelListResponse struct {
 	Data []ModelListItem `json:"data"`
 }
 
-func (p *PerplexityModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (p *PerplexityModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -296,7 +296,7 @@ func (p *PerplexityModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, p.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -333,8 +333,8 @@ func (p *PerplexityModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(ModelList{Models: bare}), nil
 }
 
-func (p *PerplexityModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := p.ListModels(apiConfig)
+func (p *PerplexityModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := p.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -350,7 +350,7 @@ type perplexityEmbeddingResponse struct {
 	Error interface{}                   `json:"error"`
 }
 
-func (p *PerplexityModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (p *PerplexityModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -382,7 +382,7 @@ func (p *PerplexityModel) Embed(modelName *string, texts []string, apiConfig *AP
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -424,42 +424,42 @@ func (p *PerplexityModel) Embed(modelName *string, texts []string, apiConfig *AP
 	return embeddings, nil
 }
 
-func (p *PerplexityModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (p *PerplexityModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (p *PerplexityModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (p *PerplexityModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PerplexityModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (p *PerplexityModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PerplexityModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (p *PerplexityModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (p *PerplexityModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (p *PerplexityModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PerplexityModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (p *PerplexityModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
diff --git a/internal/entity/models/perplexity_test.go b/internal/entity/models/perplexity_test.go
index 81030a4624..77cfca0804 100644
--- a/internal/entity/models/perplexity_test.go
+++ b/internal/entity/models/perplexity_test.go
@@ -60,6 +60,7 @@ func TestPerplexityFactory(t *testing.T) {
 }
 
 func TestPerplexityChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/sonar" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -91,6 +92,7 @@ func TestPerplexityChatHappyPath(t *testing.T) {
 	stop := []string{"END"}
 	effort := "high"
 	resp, err := newPerplexityForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"sonar-reasoning-pro",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -109,6 +111,7 @@ func TestPerplexityChatHappyPath(t *testing.T) {
 }
 
 func TestPerplexityChatSkipsReasoningEffortForNonReasoningModel(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "sonar" {
 			t.Errorf("model=%v", body["model"])
@@ -129,6 +132,7 @@ func TestPerplexityChatSkipsReasoningEffortForNonReasoningModel(t *testing.T) {
 	apiKey := "test-key"
 	effort := "high"
 	resp, err := newPerplexityForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"sonar",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -144,21 +148,24 @@ func TestPerplexityChatSkipsReasoningEffortForNonReasoningModel(t *testing.T) {
 }
 
 func TestPerplexityChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newPerplexityForTest("http://unused").ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPerplexityForTest("http://unused").ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestPerplexityChatRequiresApiKey(t *testing.T) {
-	_, err := newPerplexityForTest("http://unused").ChatWithMessages("sonar", []Message{{Role: "user", Content: "x"}}, nil, nil, nil)
+	ctx := t.Context()
+	_, err := newPerplexityForTest("http://unused").ChatWithMessages(ctx, "sonar", []Message{{Role: "user", Content: "x"}}, nil, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestPerplexityStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/sonar" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -182,6 +189,7 @@ func TestPerplexityStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	err := newPerplexityForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"sonar",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -207,6 +215,7 @@ func TestPerplexityStreamHappyPath(t *testing.T) {
 }
 
 func TestPerplexityStreamStopsOnDoneMarker(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w,
@@ -219,6 +228,7 @@ func TestPerplexityStreamStopsOnDoneMarker(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	err := newPerplexityForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"sonar",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -238,6 +248,7 @@ func TestPerplexityStreamStopsOnDoneMarker(t *testing.T) {
 }
 
 func TestPerplexityListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s", r.Method)
@@ -257,19 +268,20 @@ func TestPerplexityListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newPerplexityForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "sonar,sonar-pro,pplx-embed-v1-0.6b" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestPerplexityListModelsAcceptsBareArray(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode([]map[string]interface{}{
 			{"id": "sonar"},
@@ -279,7 +291,7 @@ func TestPerplexityListModelsAcceptsBareArray(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newPerplexityForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newPerplexityForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -289,6 +301,7 @@ func TestPerplexityListModelsAcceptsBareArray(t *testing.T) {
 }
 
 func TestPerplexityEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newPerplexityServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/v1/embeddings" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -315,6 +328,7 @@ func TestPerplexityEmbedHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "pplx-embed-v1-0.6b"
 	out, err := newPerplexityForTest(srv.URL).Embed(
+		ctx,
 		&modelName,
 		[]string{"hello", "world"},
 		&APIConfig{ApiKey: &apiKey},
@@ -336,9 +350,10 @@ func TestPerplexityEmbedHappyPath(t *testing.T) {
 }
 
 func TestPerplexityEmbedEmptyTextsReturnsEmpty(t *testing.T) {
+	ctx := t.Context()
 	modelName := "pplx-embed-v1-0.6b"
 	apiKey := "test-key"
-	out, err := newPerplexityForTest("http://unused").Embed(&modelName, nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	out, err := newPerplexityForTest("http://unused").Embed(ctx, &modelName, nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -348,19 +363,21 @@ func TestPerplexityEmbedEmptyTextsReturnsEmpty(t *testing.T) {
 }
 
 func TestPerplexityEmbedRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newPerplexityForTest("http://unused").Embed(nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPerplexityForTest("http://unused").Embed(ctx, nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestPerplexityUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newPerplexityForTest("http://unused")
-	if _, err := m.Rerank(nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank error=%v", err)
 	}
-	if _, err := m.Balance(nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance error=%v", err)
 	}
 }
diff --git a/internal/entity/models/ppio.go b/internal/entity/models/ppio.go
index aa3a6d590b..62dc1528db 100644
--- a/internal/entity/models/ppio.go
+++ b/internal/entity/models/ppio.go
@@ -112,7 +112,7 @@ type ppioChatResponse struct {
 	FinishReason string           `json:"finish_reason"`
 }
 
-func (p *PPIOModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (p *PPIOModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -133,7 +133,7 @@ func (p *PPIOModel) ChatWithMessages(modelName string, messages []Message, apiCo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -180,7 +180,7 @@ func (p *PPIOModel) ChatWithMessages(modelName string, messages []Message, apiCo
 	}, nil
 }
 
-func (p *PPIOModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PPIOModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -208,7 +208,7 @@ func (p *PPIOModel) ChatStreamlyWithSender(modelName string, messages []Message,
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -279,7 +279,7 @@ type ppioListModelsResponse struct {
 	Error interface{}     `json:"error"`
 }
 
-func (p *PPIOModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (p *PPIOModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := p.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -289,7 +289,7 @@ func (p *PPIOModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -324,51 +324,51 @@ func (p *PPIOModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (p *PPIOModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := p.ListModels(apiConfig)
+func (p *PPIOModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := p.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (p *PPIOModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (p *PPIOModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (p *PPIOModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (p *PPIOModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (p *PPIOModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PPIOModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (p *PPIOModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (p *PPIOModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (p *PPIOModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (p *PPIOModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (p *PPIOModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
 
-func (p *PPIOModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (p *PPIOModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", p.Name())
 }
diff --git a/internal/entity/models/ppio_test.go b/internal/entity/models/ppio_test.go
index 9c8f6921bf..3dede3b64b 100644
--- a/internal/entity/models/ppio_test.go
+++ b/internal/entity/models/ppio_test.go
@@ -75,6 +75,7 @@ func TestPPIONewModelWithCustomDefaultTransport(t *testing.T) {
 }
 
 func TestPPIOChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -130,6 +131,7 @@ func TestPPIOChatHappyPath(t *testing.T) {
 	stop := []string{"END"}
 	effort := "high"
 	resp, err := newPPIOForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -148,6 +150,7 @@ func TestPPIOChatHappyPath(t *testing.T) {
 }
 
 func TestPPIOChatUsesReasoningFallback(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -162,6 +165,7 @@ func TestPPIOChatUsesReasoningFallback(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newPPIOForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -177,35 +181,39 @@ func TestPPIOChatUsesReasoningFallback(t *testing.T) {
 }
 
 func TestPPIOChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newPPIOForTest("http://unused").ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPPIOForTest("http://unused").ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestPPIOChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newPPIOForTest("http://unused").ChatWithMessages("deepseek/deepseek-r1", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPPIOForTest("http://unused").ChatWithMessages(ctx, "deepseek/deepseek-r1", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages error, got %v", err)
 	}
 }
 
 func TestPPIOChatSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		http.Error(w, "bad key", http.StatusUnauthorized)
 	})
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newPPIOForTest(srv.URL).ChatWithMessages("deepseek/deepseek-r1", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPPIOForTest(srv.URL).ChatWithMessages(ctx, "deepseek/deepseek-r1", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "status 401") {
 		t.Errorf("expected HTTP status error, got %v", err)
 	}
 }
 
 func TestPPIOChatRejectsProviderError(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"error": map[string]interface{}{"message": "invalid model"},
@@ -214,13 +222,14 @@ func TestPPIOChatRejectsProviderError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newPPIOForTest(srv.URL).ChatWithMessages("deepseek/deepseek-r1", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newPPIOForTest(srv.URL).ChatWithMessages(ctx, "deepseek/deepseek-r1", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "upstream error") {
 		t.Errorf("expected upstream error, got %v", err)
 	}
 }
 
 func TestPPIOStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -245,6 +254,7 @@ func TestPPIOStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	err := newPPIOForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -273,6 +283,7 @@ func TestPPIOStreamHappyPath(t *testing.T) {
 }
 
 func TestPPIOStreamSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		http.Error(w, "bad key", http.StatusUnauthorized)
 	})
@@ -280,6 +291,7 @@ func TestPPIOStreamSurfacesHTTPError(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newPPIOForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -291,6 +303,7 @@ func TestPPIOStreamSurfacesHTTPError(t *testing.T) {
 }
 
 func TestPPIOStreamStopsOnSenderError(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w, `data: {"choices":[{"delta":{"content":"partial"}}]}`+"\n")
@@ -299,6 +312,7 @@ func TestPPIOStreamStopsOnSenderError(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newPPIOForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -310,9 +324,11 @@ func TestPPIOStreamStopsOnSenderError(t *testing.T) {
 }
 
 func TestPPIOStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
 	err := newPPIOForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -326,8 +342,10 @@ func TestPPIOStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestPPIOStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newPPIOForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil,
@@ -338,6 +356,7 @@ func TestPPIOStreamRequiresSender(t *testing.T) {
 }
 
 func TestPPIOStreamRequiresTerminalEvent(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w, `data: {"choices":[{"delta":{"content":"partial"}}]}`+"\n")
@@ -346,6 +365,7 @@ func TestPPIOStreamRequiresTerminalEvent(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newPPIOForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek/deepseek-r1",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -357,6 +377,7 @@ func TestPPIOStreamRequiresTerminalEvent(t *testing.T) {
 }
 
 func TestPPIOListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s", r.Method)
@@ -375,26 +396,28 @@ func TestPPIOListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newPPIOForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "deepseek/deepseek-r1,qwen/qwen-2.5-72b-instruct" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestPPIOListModelsRequiresAPIKey(t *testing.T) {
-	_, err := newPPIOForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newPPIOForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestPPIOListModelsRejectsProviderError(t *testing.T) {
+	ctx := t.Context()
 	srv := newPPIOServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"error": map[string]interface{}{"message": "unauthorized"},
@@ -403,7 +426,7 @@ func TestPPIOListModelsRejectsProviderError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newPPIOForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := newPPIOForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "upstream error") {
 		t.Errorf("expected upstream error, got %v", err)
 	}
@@ -473,38 +496,39 @@ func TestPPIOMissingRegionBaseURL(t *testing.T) {
 }
 
 func TestPPIOUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newPPIOForTest("http://unused")
-	if _, err := m.Embed(nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Embed(ctx, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed error=%v", err)
 	}
-	if _, err := m.Rerank(nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, nil, "", nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank error=%v", err)
 	}
-	if _, err := m.Balance(nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance error=%v", err)
 	}
-	if _, err := m.TranscribeAudio(nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio error=%v", err)
 	}
-	if err := m.TranscribeAudioWithSender(nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.TranscribeAudioWithSender(ctx, nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudioWithSender error=%v", err)
 	}
-	if _, err := m.AudioSpeech(nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech error=%v", err)
 	}
-	if err := m.AudioSpeechWithSender(nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := m.AudioSpeechWithSender(ctx, nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeechWithSender error=%v", err)
 	}
-	if _, err := m.OCRFile(nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile error=%v", err)
 	}
-	if _, err := m.ParseFile(nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ParseFile(ctx, nil, nil, nil, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ParseFile error=%v", err)
 	}
-	if _, err := m.ListTasks(nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ListTasks(ctx, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ListTasks error=%v", err)
 	}
-	if _, err := m.ShowTask("", nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.ShowTask(ctx, "", nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ShowTask error=%v", err)
 	}
 }
diff --git a/internal/entity/models/qiniu.go b/internal/entity/models/qiniu.go
index 6f6e16cfe4..bb0f648d7a 100644
--- a/internal/entity/models/qiniu.go
+++ b/internal/entity/models/qiniu.go
@@ -140,7 +140,7 @@ func applyQiniuThinkingConfig(reqBody map[string]interface{}, modelName string,
 	}
 }
 
-func (q *QiniuModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (q *QiniuModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := q.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -187,7 +187,7 @@ func (q *QiniuModel) ChatWithMessages(modelName string, messages []Message, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -260,7 +260,7 @@ func (q *QiniuModel) ChatWithMessages(modelName string, messages []Message, apiC
 	return chatResponse, nil
 }
 
-func (q *QiniuModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (q *QiniuModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := q.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -308,7 +308,7 @@ func (q *QiniuModel) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -385,39 +385,39 @@ func (q *QiniuModel) ChatStreamlyWithSender(modelName string, messages []Message
 	return nil
 }
 
-func (q *QiniuModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (q *QiniuModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (q *QiniuModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (q *QiniuModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (q *QiniuModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (q *QiniuModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (q *QiniuModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (q *QiniuModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (q *QiniuModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (q *QiniuModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := q.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -429,7 +429,7 @@ func (q *QiniuModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	baseURL := strings.TrimSuffix(resolvedBaseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, q.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -464,22 +464,22 @@ func (q *QiniuModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return ParseListModel(modelList), nil
 }
 
-func (q *QiniuModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (q *QiniuModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := q.ListModels(apiConfig)
+func (q *QiniuModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := q.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
 	return nil
 }
 
-func (q *QiniuModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (q *QiniuModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
 
-func (q *QiniuModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (q *QiniuModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", q.Name())
 }
diff --git a/internal/entity/models/qiniu_test.go b/internal/entity/models/qiniu_test.go
index a0436a927f..0181d913ab 100644
--- a/internal/entity/models/qiniu_test.go
+++ b/internal/entity/models/qiniu_test.go
@@ -36,6 +36,7 @@ func TestQiniuToolCalls(t *testing.T) {
 }
 
 func TestQiniuChatStreamRejectsTruncatedResponse(t *testing.T) {
+	ctx := t.Context()
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = w.Write([]byte(`data: {"choices":[{"delta":{"content":"partial"}}]}` + "\n\n"))
@@ -44,13 +45,14 @@ func TestQiniuChatStreamRejectsTruncatedResponse(t *testing.T) {
 
 	apiKey := "test-key"
 	model := NewQiniuModel(map[string]string{"default": server.URL}, URLSuffix{Chat: "chat/completions"})
-	err := model.ChatStreamlyWithSender("model", []Message{{Role: "user", Content: "hi"}}, &APIConfig{ApiKey: &apiKey}, &ChatConfig{}, nil, func(_, _ *string) error { return nil })
+	err := model.ChatStreamlyWithSender(ctx, "model", []Message{{Role: "user", Content: "hi"}}, &APIConfig{ApiKey: &apiKey}, &ChatConfig{}, nil, func(_, _ *string) error { return nil })
 	if err == nil || !strings.Contains(err.Error(), "stream ended before [DONE] or finish_reason") {
 		t.Fatalf("error = %v", err)
 	}
 }
 
 func TestQiniuCheckConnectionUsesListModels(t *testing.T) {
+	ctx := t.Context()
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method = %s, want GET", r.Method)
@@ -67,12 +69,13 @@ func TestQiniuCheckConnectionUsesListModels(t *testing.T) {
 
 	apiKey := "test-key"
 	model := NewQiniuModel(map[string]string{"default": server.URL}, URLSuffix{Models: "models"})
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection() error = %v", err)
 	}
 }
 
 func TestQiniuCheckConnectionPropagatesListModelsError(t *testing.T) {
+	ctx := t.Context()
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		http.Error(w, `{"message":"invalid api key"}`, http.StatusUnauthorized)
 	}))
@@ -80,7 +83,7 @@ func TestQiniuCheckConnectionPropagatesListModelsError(t *testing.T) {
 
 	apiKey := "invalid-key"
 	model := NewQiniuModel(map[string]string{"default": server.URL}, URLSuffix{Models: "models"})
-	err := model.CheckConnection(&APIConfig{ApiKey: &apiKey})
+	err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "status 401") {
 		t.Fatalf("CheckConnection() error = %v, want status 401", err)
 	}
diff --git a/internal/entity/models/ragcon.go b/internal/entity/models/ragcon.go
index be1e6cdfeb..ddf666f4cc 100644
--- a/internal/entity/models/ragcon.go
+++ b/internal/entity/models/ragcon.go
@@ -127,7 +127,7 @@ func ragconChatPayload(modelName string, messages []Message, stream bool, chatMo
 	return reqBody
 }
 
-func (r *RAGconModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (r *RAGconModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -148,7 +148,7 @@ func (r *RAGconModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -223,7 +223,7 @@ func (r *RAGconModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (r *RAGconModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *RAGconModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -250,7 +250,7 @@ func (r *RAGconModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), http.MethodPost, url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -349,7 +349,7 @@ type ragconEmbeddingResponse struct {
 	} `json:"data"`
 }
 
-func (r *RAGconModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (r *RAGconModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -380,7 +380,7 @@ func (r *RAGconModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -417,7 +417,7 @@ func (r *RAGconModel) Embed(modelName *string, texts []string, apiConfig *APICon
 }
 
 // Rerank POSTs to RAGcon's /rerank endpoint (LiteLLM proxy passthrough).
-func (r *RAGconModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (r *RAGconModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -452,7 +452,7 @@ func (r *RAGconModel) Rerank(modelName *string, query string, documents []string
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -497,7 +497,7 @@ func (r *RAGconModel) Rerank(modelName *string, query string, documents []string
 }
 
 // ListModels returns the model ids visible to the configured API key.
-func (r *RAGconModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (r *RAGconModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -509,7 +509,7 @@ func (r *RAGconModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, r.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -541,13 +541,13 @@ func (r *RAGconModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	return ParseListModel(modelList), nil
 }
 
-func (r *RAGconModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := r.ListModels(apiConfig)
+func (r *RAGconModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := r.ListModels(ctx, apiConfig)
 	return err
 }
 
 // Balance is not exposed by RAGcon's LiteLLM proxy.
-func (r *RAGconModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (r *RAGconModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
@@ -623,8 +623,8 @@ func (r *RAGconModel) newASRRequest(ctx context.Context, modelName *string, file
 	return req, responseFormat, nil
 }
 
-func (r *RAGconModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+func (r *RAGconModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, responseFormat, err := r.newASRRequest(ctx, modelName, file, apiConfig, asrConfig, false)
@@ -650,12 +650,12 @@ func (r *RAGconModel) TranscribeAudio(modelName *string, file *string, apiConfig
 	return decodeOpenAIASRResponse(respBody, responseFormat)
 }
 
-func (r *RAGconModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *RAGconModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	req, responseFormat, err := r.newASRRequest(context.Background(), modelName, file, apiConfig, asrConfig, true)
+	req, responseFormat, err := r.newASRRequest(ctx, modelName, file, apiConfig, asrConfig, true)
 	if err != nil {
 		return err
 	}
@@ -786,8 +786,8 @@ func (r *RAGconModel) newTTSRequest(ctx context.Context, modelName *string, audi
 	return req, streamFormat, nil
 }
 
-func (r *RAGconModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+func (r *RAGconModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, _, err := r.newTTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, false)
@@ -812,12 +812,12 @@ func (r *RAGconModel) AudioSpeech(modelName *string, audioContent *string, apiCo
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (r *RAGconModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *RAGconModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	req, streamFormat, err := r.newTTSRequest(context.Background(), modelName, audioContent, apiConfig, ttsConfig, true)
+	req, streamFormat, err := r.newTTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, true)
 	if err != nil {
 		return err
 	}
@@ -843,19 +843,19 @@ func (r *RAGconModel) AudioSpeechWithSender(modelName *string, audioContent *str
 }
 
 // OCRFile is not offered by RAGcon.
-func (r *RAGconModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (r *RAGconModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
 // ParseFile is not offered by RAGcon.
-func (r *RAGconModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (r *RAGconModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *RAGconModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (r *RAGconModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *RAGconModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (r *RAGconModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
diff --git a/internal/entity/models/ragcon_test.go b/internal/entity/models/ragcon_test.go
index e30a23465c..04b9414008 100644
--- a/internal/entity/models/ragcon_test.go
+++ b/internal/entity/models/ragcon_test.go
@@ -82,6 +82,8 @@ func TestRAGconFactory(t *testing.T) {
 }
 
 func TestRAGconChatHappyPath(t *testing.T) {
+	ctx := t.Context()
+
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s, want /chat/completions", r.URL.Path)
@@ -110,6 +112,7 @@ func TestRAGconChatHappyPath(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newRAGconForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"llama-4-maverick",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -130,7 +133,9 @@ func TestRAGconChatHappyPath(t *testing.T) {
 }
 
 func TestRAGconChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newRAGconForTest("http://unused").ChatWithMessages(
+		ctx,
 		"llama-4-maverick",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{},
@@ -142,8 +147,10 @@ func TestRAGconChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestRAGconChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	_, err := newRAGconForTest("http://unused").ChatWithMessages(
+		ctx,
 		"llama-4-maverick",
 		[]Message{},
 		&APIConfig{ApiKey: &apiKey},
@@ -155,6 +162,7 @@ func TestRAGconChatRequiresMessages(t *testing.T) {
 }
 
 func TestRAGconChatSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = io.WriteString(w, `{"error":"invalid key"}`)
@@ -163,6 +171,7 @@ func TestRAGconChatSurfacesHTTPError(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newRAGconForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"llama-4-maverick",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -174,6 +183,7 @@ func TestRAGconChatSurfacesHTTPError(t *testing.T) {
 }
 
 func TestRAGconStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -195,6 +205,7 @@ func TestRAGconStreamHappyPath(t *testing.T) {
 	var content, reasoning []string
 	var sawDone bool
 	err := newRAGconForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"llama-4-maverick",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -228,8 +239,10 @@ func TestRAGconStreamHappyPath(t *testing.T) {
 }
 
 func TestRAGconStreamRejectsNilSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newRAGconForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"llama-4-maverick",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -243,6 +256,7 @@ func TestRAGconStreamRejectsNilSender(t *testing.T) {
 }
 
 func TestRAGconEmbed(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/embeddings" {
 			t.Errorf("path=%s, want /embeddings", r.URL.Path)
@@ -258,7 +272,7 @@ func TestRAGconEmbed(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	embeddings, err := newRAGconForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	embeddings, err := newRAGconForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -268,6 +282,7 @@ func TestRAGconEmbed(t *testing.T) {
 }
 
 func TestRAGconRerank(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/rerank" {
 			t.Errorf("path=%s, want /rerank", r.URL.Path)
@@ -286,7 +301,7 @@ func TestRAGconRerank(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "rerank-v1"
-	resp, err := newRAGconForTest(srv.URL).Rerank(&model, "q", []string{"doc0", "doc1"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil)
+	resp, err := newRAGconForTest(srv.URL).Rerank(ctx, &model, "q", []string{"doc0", "doc1"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
@@ -296,6 +311,7 @@ func TestRAGconRerank(t *testing.T) {
 }
 
 func TestRAGconListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/models" {
 			t.Errorf("path=%s, want /models", r.URL.Path)
@@ -306,19 +322,20 @@ func TestRAGconListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	cfg := &APIConfig{ApiKey: &apiKey}
-	models, err := newRAGconForTest(srv.URL).ListModels(cfg)
+	models, err := newRAGconForTest(srv.URL).ListModels(ctx, cfg)
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "llama-4-maverick,gpt-oss-120b" {
 		t.Errorf("models=%v", models)
 	}
-	if err := newRAGconForTest(srv.URL).CheckConnection(cfg); err != nil {
+	if err := newRAGconForTest(srv.URL).CheckConnection(ctx, cfg); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestRAGconTranscribeAudioPostsMultipart(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/audio/transcriptions" {
 			t.Errorf("path=%s, want /audio/transcriptions", r.URL.Path)
@@ -352,7 +369,7 @@ func TestRAGconTranscribeAudioPostsMultipart(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "whisper-1"
-	resp, err := newRAGconForTest(srv.URL).TranscribeAudio(&model, &audioPath, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newRAGconForTest(srv.URL).TranscribeAudio(ctx, &model, &audioPath, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("TranscribeAudio: %v", err)
 	}
@@ -362,6 +379,7 @@ func TestRAGconTranscribeAudioPostsMultipart(t *testing.T) {
 }
 
 func TestRAGconAudioSpeechPostsJSON(t *testing.T) {
+	ctx := t.Context()
 	srv := newRAGconServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/audio/speech" {
 			t.Errorf("path=%s, want /audio/speech", r.URL.Path)
@@ -376,7 +394,7 @@ func TestRAGconAudioSpeechPostsJSON(t *testing.T) {
 	apiKey := "test-key"
 	model := "tts-1"
 	text := "hello"
-	resp, err := newRAGconForTest(srv.URL).AudioSpeech(&model, &text, &APIConfig{ApiKey: &apiKey}, &TTSConfig{Params: map[string]interface{}{"voice": "alloy"}}, nil)
+	resp, err := newRAGconForTest(srv.URL).AudioSpeech(ctx, &model, &text, &APIConfig{ApiKey: &apiKey}, &TTSConfig{Params: map[string]interface{}{"voice": "alloy"}}, nil)
 	if err != nil {
 		t.Fatalf("AudioSpeech: %v", err)
 	}
@@ -386,22 +404,23 @@ func TestRAGconAudioSpeechPostsJSON(t *testing.T) {
 }
 
 func TestRAGconUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	r := newRAGconForTest("http://unused")
 	model := "llama-4-maverick"
 
-	if _, err := r.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := r.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected no such method, got %v", err)
 	}
-	if _, err := r.OCRFile(&model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := r.OCRFile(ctx, &model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: expected no such method, got %v", err)
 	}
-	if _, err := r.ParseFile(&model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := r.ParseFile(ctx, &model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ParseFile: expected no such method, got %v", err)
 	}
-	if _, err := r.ListTasks(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := r.ListTasks(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ListTasks: expected no such method, got %v", err)
 	}
-	if _, err := r.ShowTask("t1", &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := r.ShowTask(ctx, "t1", &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("ShowTask: expected no such method, got %v", err)
 	}
 }
diff --git a/internal/entity/models/reasoning_family_provider_test.go b/internal/entity/models/reasoning_family_provider_test.go
index 2687429925..37a9b21aaf 100644
--- a/internal/entity/models/reasoning_family_provider_test.go
+++ b/internal/entity/models/reasoning_family_provider_test.go
@@ -47,6 +47,7 @@ func newReasoningFamilyChatServer(t *testing.T, handler func(t *testing.T, body
 }
 
 func TestGiteeChatExtractsQwenThinkingFromInlineContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newReasoningFamilyChatServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "qwen3-8b" {
 			t.Errorf("model=%v, want qwen3-8b", body["model"])
@@ -72,6 +73,7 @@ func TestGiteeChatExtractsQwenThinkingFromInlineContent(t *testing.T) {
 		map[string]string{"default": srv.URL},
 		URLSuffix{Chat: "chat/completions"},
 	).ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -85,6 +87,7 @@ func TestGiteeChatExtractsQwenThinkingFromInlineContent(t *testing.T) {
 }
 
 func TestSiliconflowChatExtractsProviderPrefixedQwenThinkingFromInlineContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newReasoningFamilyChatServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "qwen/qwen3-8b" {
 			t.Errorf("model=%v, want qwen/qwen3-8b", body["model"])
@@ -110,6 +113,7 @@ func TestSiliconflowChatExtractsProviderPrefixedQwenThinkingFromInlineContent(t
 		map[string]string{"default": srv.URL},
 		URLSuffix{Chat: "chat/completions"},
 	).ChatWithMessages(
+		ctx,
 		"qwen/qwen3-8b",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/replicate.go b/internal/entity/models/replicate.go
index b208bd4c8c..22f7be9327 100644
--- a/internal/entity/models/replicate.go
+++ b/internal/entity/models/replicate.go
@@ -325,7 +325,7 @@ func (r *ReplicateModel) waitForPrediction(ctx context.Context, prediction *repl
 	}
 }
 
-func (r *ReplicateModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (r *ReplicateModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -341,7 +341,7 @@ func (r *ReplicateModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	prediction, err := r.createPrediction(ctx, url, version, replicateInputFromMessages(messages, chatModelConfig), false, *apiConfig.ApiKey, true)
@@ -364,7 +364,7 @@ func (r *ReplicateModel) ChatWithMessages(modelName string, messages []Message,
 	return &ChatResponse{Answer: &answer, ReasonContent: &reasonContent}, nil
 }
 
-func (r *ReplicateModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *ReplicateModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -387,12 +387,12 @@ func (r *ReplicateModel) ChatStreamlyWithSender(modelName string, messages []Mes
 		return err
 	}
 
-	prediction, err := r.createPrediction(context.Background(), url, version, replicateInputFromMessages(messages, chatModelConfig), true, *apiConfig.ApiKey, false)
+	prediction, err := r.createPrediction(ctx, url, version, replicateInputFromMessages(messages, chatModelConfig), true, *apiConfig.ApiKey, false)
 	if err != nil {
 		return err
 	}
 	if prediction.URLs.Stream == "" {
-		ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+		ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 		defer cancel()
 		prediction, err = r.waitForPrediction(ctx, prediction, *apiConfig.ApiKey)
 		if err != nil {
@@ -411,11 +411,11 @@ func (r *ReplicateModel) ChatStreamlyWithSender(modelName string, messages []Mes
 		return sender(&endOfStream, nil)
 	}
 
-	return r.readPredictionStream(prediction.URLs.Stream, *apiConfig.ApiKey, sender)
+	return r.readPredictionStream(ctx, prediction.URLs.Stream, *apiConfig.ApiKey, sender)
 }
 
-func (r *ReplicateModel) readPredictionStream(url string, apiKey string, sender func(*string, *string) error) error {
-	req, err := http.NewRequestWithContext(context.Background(), http.MethodGet, url, nil)
+func (r *ReplicateModel) readPredictionStream(ctx context.Context, url string, apiKey string, sender func(*string, *string) error) error {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -499,7 +499,7 @@ func dispatchReplicateSSEEvent(event replicateSSEEvent, sender func(*string, *st
 	}
 }
 
-func (r *ReplicateModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (r *ReplicateModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -509,7 +509,7 @@ func (r *ReplicateModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -567,8 +567,8 @@ func (r *ReplicateModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(modelList), nil
 }
 
-func (r *ReplicateModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := r.ListModels(apiConfig)
+func (r *ReplicateModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := r.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -674,7 +674,7 @@ func replicateKeys(m map[string]interface{}) []string {
 }
 
 // Embed turns a list of texts into embedding vectors via Replicate's
-func (r *ReplicateModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (r *ReplicateModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -696,7 +696,7 @@ func (r *ReplicateModel) Embed(modelName *string, texts []string, apiConfig *API
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	prediction, err := r.createPrediction(ctx, url, version, input, false, *apiConfig.ApiKey, true)
@@ -765,7 +765,7 @@ func replicateScoresFromInterface(arr []interface{}, n int) ([]float64, error) {
 }
 
 // Rerank scores a query against a list of documents
-func (r *ReplicateModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (r *ReplicateModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := r.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -787,7 +787,7 @@ func (r *ReplicateModel) Rerank(modelName *string, query string, documents []str
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	prediction, err := r.createPrediction(ctx, url, version, input, false, *apiConfig.ApiKey, true)
@@ -829,38 +829,38 @@ func (r *ReplicateModel) Rerank(modelName *string, query string, documents []str
 	return &RerankResponse{Data: results}, nil
 }
 
-func (r *ReplicateModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (r *ReplicateModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (r *ReplicateModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *ReplicateModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (r *ReplicateModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (r *ReplicateModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (r *ReplicateModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (r *ReplicateModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (r *ReplicateModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
 
-func (r *ReplicateModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (r *ReplicateModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", r.Name())
 }
diff --git a/internal/entity/models/replicate_test.go b/internal/entity/models/replicate_test.go
index 020a4fb91a..489b86b33d 100644
--- a/internal/entity/models/replicate_test.go
+++ b/internal/entity/models/replicate_test.go
@@ -122,7 +122,9 @@ func TestReplicateOfficialChatHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	maxTokens := 128
 	stop := []string{"END"}
+	ctx := t.Context()
 	resp, err := newReplicateForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"meta/meta-llama-3-70b-instruct",
 		[]Message{{Role: "system", Content: "be helpful"}, {Role: "user", Content: "hello"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -161,9 +163,10 @@ func TestReplicateCommunityChatUsesVersionEndpoint(t *testing.T) {
 		})
 	}))
 	defer srv.Close()
-
+	ctx := t.Context()
 	apiKey := "test-key"
 	resp, err := newReplicateForTest(srv.URL).ChatWithMessages(
+		ctx,
 		version,
 		[]Message{{Role: "user", Content: "hello"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -201,9 +204,10 @@ func TestReplicateChatPollsUntilSucceeded(t *testing.T) {
 		}
 	}))
 	defer srv.Close()
-
+	ctx := t.Context()
 	apiKey := "test-key"
 	resp, err := newReplicateForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"meta/meta-llama-3-70b-instruct",
 		[]Message{{Role: "user", Content: "hello"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -256,10 +260,11 @@ func TestReplicateStreamHappyPath(t *testing.T) {
 		})
 	}))
 	defer apiSrv.Close()
-
+	ctx := t.Context()
 	apiKey := "test-key"
 	var chunks []string
 	err := newReplicateForTest(apiSrv.URL).ChatStreamlyWithSender(
+		ctx,
 		"meta/meta-llama-3-70b-instruct",
 		[]Message{{Role: "user", Content: "hello"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -278,6 +283,7 @@ func TestReplicateStreamHappyPath(t *testing.T) {
 }
 
 func TestReplicateListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/models" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -296,28 +302,29 @@ func TestReplicateListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newReplicateForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "meta/meta-llama-3-70b-instruct,replicate/hello-world" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestReplicateUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newReplicateForTest("http://unused")
 	apiKey := "test-key"
 	// Rerank IS implemented; with empty documents it short-circuits (no error).
 	// Pass non-empty docs + nil modelName to trigger model-name validation.
-	if _, err := m.Rerank(nil, "", []string{"d"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := m.Rerank(ctx, nil, "", []string{"d"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("Rerank error=%v", err)
 	}
 	// Balance IS a stub → "no such method"
-	if _, err := m.Balance(&APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance error=%v", err)
 	}
 }
diff --git a/internal/entity/models/siliconflow.go b/internal/entity/models/siliconflow.go
index cb6196743f..424a027982 100644
--- a/internal/entity/models/siliconflow.go
+++ b/internal/entity/models/siliconflow.go
@@ -67,7 +67,7 @@ type SiliconflowRerankRequest struct {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (s *SiliconflowModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (s *SiliconflowModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -132,7 +132,7 @@ func (s *SiliconflowModel) ChatWithMessages(modelName string, messages []Message
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -213,7 +213,7 @@ func (s *SiliconflowModel) ChatWithMessages(modelName string, messages []Message
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (s *SiliconflowModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *SiliconflowModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -275,7 +275,7 @@ func (s *SiliconflowModel) ChatStreamlyWithSender(modelName string, messages []M
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -380,7 +380,7 @@ type siliconflowUsage struct {
 const siliconflowMaxBatchSize = 32
 
 // Embed embeds a list of texts into embeddings
-func (s *SiliconflowModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (s *SiliconflowModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -417,7 +417,7 @@ func (s *SiliconflowModel) Embed(modelName *string, texts []string, apiConfig *A
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -462,7 +462,7 @@ func (s *SiliconflowModel) Embed(modelName *string, texts []string, apiConfig *A
 	return embeddings, nil
 }
 
-func (s *SiliconflowModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (s *SiliconflowModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -481,7 +481,7 @@ func (s *SiliconflowModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -526,7 +526,7 @@ type siliconflowBalanceResponse struct {
 	} `json:"data"`
 }
 
-func (s *SiliconflowModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (s *SiliconflowModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -538,7 +538,7 @@ func (s *SiliconflowModel) Balance(apiConfig *APIConfig) (map[string]interface{}
 
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), s.baseModel.URLSuffix.Balance)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -595,8 +595,8 @@ func (s *SiliconflowModel) Balance(apiConfig *APIConfig) (map[string]interface{}
 	}, nil
 }
 
-func (s *SiliconflowModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := s.ListModels(apiConfig)
+func (s *SiliconflowModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := s.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -630,7 +630,7 @@ type SiliconflowRerankResponse struct {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (s *SiliconflowModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (s *SiliconflowModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -667,7 +667,7 @@ func (s *SiliconflowModel) Rerank(modelName *string, query string, documents []s
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(resolvedBaseURL, "/"), s.baseModel.URLSuffix.Rerank)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, strings.NewReader(string(jsonData)))
@@ -709,7 +709,7 @@ func (s *SiliconflowModel) Rerank(modelName *string, query string, documents []s
 }
 
 // TranscribeAudio transcribe audio
-func (s *SiliconflowModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (s *SiliconflowModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -790,7 +790,7 @@ func (s *SiliconflowModel) TranscribeAudio(modelName *string, file *string, apiC
 	}
 
 	// build request
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, &body)
@@ -830,12 +830,12 @@ func (s *SiliconflowModel) TranscribeAudio(modelName *string, file *string, apiC
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (s *SiliconflowModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *SiliconflowModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", s.Name())
 }
 
 // AudioSpeech convert text to audio
-func (s *SiliconflowModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (s *SiliconflowModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -870,7 +870,7 @@ func (s *SiliconflowModel) AudioSpeech(modelName *string, audioContent *string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -899,7 +899,7 @@ func (s *SiliconflowModel) AudioSpeech(modelName *string, audioContent *string,
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (s *SiliconflowModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *SiliconflowModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -934,7 +934,7 @@ func (s *SiliconflowModel) AudioSpeechWithSender(modelName *string, audioContent
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -979,19 +979,19 @@ func (s *SiliconflowModel) AudioSpeechWithSender(modelName *string, audioContent
 }
 
 // OCRFile OCR file
-func (s *SiliconflowModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (s *SiliconflowModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
 // ParseFile parse file
-func (s *SiliconflowModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (s *SiliconflowModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
-func (s *SiliconflowModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (s *SiliconflowModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
-func (s *SiliconflowModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (s *SiliconflowModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
diff --git a/internal/entity/models/siliconflow_test.go b/internal/entity/models/siliconflow_test.go
index 28bb872478..d40863f1b6 100644
--- a/internal/entity/models/siliconflow_test.go
+++ b/internal/entity/models/siliconflow_test.go
@@ -42,7 +42,8 @@ func TestSiliconflowChatRejectsMissingContent(t *testing.T) {
 
 	apiKey := "test-key"
 	model := NewSiliconflowModel(map[string]string{"default": server.URL}, URLSuffix{Chat: "chat/completions"})
-	if _, err := model.ChatWithMessages("model", []Message{{Role: "user", Content: "hi"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil {
+	ctx := t.Context()
+	if _, err := model.ChatWithMessages(ctx, "model", []Message{{Role: "user", Content: "hi"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil {
 		t.Fatal("expected missing content error")
 	}
 }
diff --git a/internal/entity/models/sse_scanner_buffer_test.go b/internal/entity/models/sse_scanner_buffer_test.go
index e2aee7de24..89a2b1410b 100644
--- a/internal/entity/models/sse_scanner_buffer_test.go
+++ b/internal/entity/models/sse_scanner_buffer_test.go
@@ -1,6 +1,7 @@
 package models
 
 import (
+	"context"
 	"io"
 	"net/http"
 	"net/http/httptest"
@@ -13,7 +14,7 @@ import (
 // provider. The buffer regression below exercises it through a table so a new
 // provider only needs one row.
 type chatStreamer interface {
-	ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
+	ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
 }
 
 // largeSSEStreamServer streams a single SSE "data:" line whose content delta is
@@ -63,6 +64,7 @@ func TestChatStreamLargeChunkNotTruncated(t *testing.T) {
 		{"Jiekou.AI", build(func(b map[string]string, s URLSuffix) chatStreamer { return NewJieKouAIModel(b, s) })},
 	}
 
+	ctx := t.Context()
 	for _, tc := range cases {
 		tc := tc
 		t.Run(tc.name, func(t *testing.T) {
@@ -72,6 +74,7 @@ func TestChatStreamLargeChunkNotTruncated(t *testing.T) {
 			apiKey := "test-key"
 			var got strings.Builder
 			err := tc.build(srv.URL).ChatStreamlyWithSender(
+				ctx,
 				"test-model",
 				[]Message{{Role: "user", Content: "hi"}},
 				&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/stepfun.go b/internal/entity/models/stepfun.go
index 7b8f1004a8..512e435aae 100644
--- a/internal/entity/models/stepfun.go
+++ b/internal/entity/models/stepfun.go
@@ -53,7 +53,7 @@ func (s *StepFunModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (s *StepFunModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (s *StepFunModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -103,7 +103,7 @@ func (s *StepFunModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -162,7 +162,7 @@ func (s *StepFunModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (s *StepFunModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *StepFunModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -220,7 +220,7 @@ func (s *StepFunModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -285,12 +285,12 @@ func (s *StepFunModel) ChatStreamlyWithSender(modelName string, messages []Messa
 }
 
 // Embed is left as a stub.
-func (s *StepFunModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (s *StepFunModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (s *StepFunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (s *StepFunModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -302,7 +302,7 @@ func (s *StepFunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, s.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -340,13 +340,13 @@ func (s *StepFunModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 }
 
 // Balance is not exposed by the StepFun API, so this returns "no such method".
-func (s *StepFunModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (s *StepFunModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (s *StepFunModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := s.ListModels(apiConfig)
+func (s *StepFunModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := s.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -355,21 +355,21 @@ func (s *StepFunModel) CheckConnection(apiConfig *APIConfig) error {
 
 // Rerank calculates similarity scores between query and documents. StepFun
 // does not expose a public rerank API, so this returns "no such method".
-func (s *StepFunModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (s *StepFunModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (s *StepFunModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (s *StepFunModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
-func (s *StepFunModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *StepFunModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", s.Name())
 }
 
 // AudioSpeech convert text to audio
-func (s *StepFunModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (s *StepFunModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -431,7 +431,7 @@ func (s *StepFunModel) AudioSpeech(modelName *string, audioContent *string, apiC
 }
 
 // AudioSpeechWithSender for Streaming TTS
-func (s *StepFunModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (s *StepFunModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := s.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -513,19 +513,19 @@ func (s *StepFunModel) AudioSpeechWithSender(modelName *string, audioContent *st
 }
 
 // OCRFile OCR file
-func (s *StepFunModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (s *StepFunModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
 // ParseFile parse file
-func (s *StepFunModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (s *StepFunModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
-func (s *StepFunModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (s *StepFunModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
 
-func (s *StepFunModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (s *StepFunModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", s.Name())
 }
diff --git a/internal/entity/models/timeout_test.go b/internal/entity/models/timeout_test.go
index 2b2097970d..d41e3bbcfc 100644
--- a/internal/entity/models/timeout_test.go
+++ b/internal/entity/models/timeout_test.go
@@ -59,6 +59,7 @@ func newTimeoutTestGroq(baseURL string) *GroqModel {
 // nonStreamCallTimeout (the old 120s-wall footgun, just relocated), the
 // context would cancel mid-stream and this test would fail.
 func TestStreamNotTruncatedByNonStreamTimeout(t *testing.T) {
+	ctx := t.Context()
 	// Stream emits for ~240ms, far past the 60ms non-stream deadline but well
 	// inside the 10s stream deadline.
 	withTestTimeouts(t, 60*time.Millisecond, 10*time.Second)
@@ -86,6 +87,7 @@ func TestStreamNotTruncatedByNonStreamTimeout(t *testing.T) {
 	apiKey := "test-key"
 	var got strings.Builder
 	err := newTimeoutTestGroq(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"llama-3.3-70b-versatile",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -116,6 +118,7 @@ func TestStreamNotTruncatedByNonStreamTimeout(t *testing.T) {
 // watchdog so a broken timeout surfaces as a direct test failure
 // rather than relying on the package's global test timeout.
 func TestNonStreamHonorsShortDeadline(t *testing.T) {
+	ctx := t.Context()
 	withTestTimeouts(t, 100*time.Millisecond, 10*time.Second)
 
 	var hits atomic.Int32
@@ -130,6 +133,7 @@ func TestNonStreamHonorsShortDeadline(t *testing.T) {
 	done := make(chan error, 1)
 	go func() {
 		_, err := newTimeoutTestGroq(srv.URL).ChatWithMessages(
+			ctx,
 			"llama-3.3-70b-versatile",
 			[]Message{{Role: "user", Content: "hi"}},
 			&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/togetherai.go b/internal/entity/models/togetherai.go
index e137f543d4..cf880f8412 100644
--- a/internal/entity/models/togetherai.go
+++ b/internal/entity/models/togetherai.go
@@ -127,7 +127,7 @@ type togetherAIChatResponse struct {
 	FinishReason string                 `json:"finish_reason"`
 }
 
-func (t *TogetherAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (t *TogetherAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -148,7 +148,7 @@ func (t *TogetherAIModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -194,7 +194,7 @@ func (t *TogetherAIModel) ChatWithMessages(modelName string, messages []Message,
 	}, nil
 }
 
-func (t *TogetherAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TogetherAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -222,7 +222,7 @@ func (t *TogetherAIModel) ChatStreamlyWithSender(modelName string, messages []Me
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -289,7 +289,7 @@ type togetherAIModelInfo struct {
 	ID string `json:"id"`
 }
 
-func (t *TogetherAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (t *TogetherAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -301,7 +301,7 @@ func (t *TogetherAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, t.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
@@ -333,8 +333,8 @@ func (t *TogetherAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(ModelList{Models: result}), nil
 }
 
-func (t *TogetherAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := t.ListModels(apiConfig)
+func (t *TogetherAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := t.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -350,7 +350,7 @@ type togetherAIEmbeddingResponse struct {
 	Object string                    `json:"object"`
 }
 
-func (t *TogetherAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (t *TogetherAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -383,7 +383,7 @@ func (t *TogetherAIModel) Embed(modelName *string, texts []string, apiConfig *AP
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -438,7 +438,7 @@ func (t *TogetherAIModel) Embed(modelName *string, texts []string, apiConfig *AP
 	return embeddings, nil
 }
 
-func (t *TogetherAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (t *TogetherAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -516,11 +516,11 @@ func (t *TogetherAIModel) Rerank(modelName *string, query string, documents []st
 	return &rerankResponse, nil
 }
 
-func (t *TogetherAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (t *TogetherAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TogetherAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (t *TogetherAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -623,11 +623,11 @@ func (t *TogetherAIModel) TranscribeAudio(modelName *string, file *string, apiCo
 	}, nil
 }
 
-func (t *TogetherAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TogetherAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TogetherAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (t *TogetherAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -687,7 +687,7 @@ func (t *TogetherAIModel) AudioSpeech(modelName *string, audioContent *string, a
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (t *TogetherAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TogetherAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -773,18 +773,18 @@ func (t *TogetherAIModel) AudioSpeechWithSender(modelName *string, audioContent
 	return nil
 }
 
-func (t *TogetherAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (t *TogetherAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TogetherAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (t *TogetherAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TogetherAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (t *TogetherAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TogetherAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (t *TogetherAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
diff --git a/internal/entity/models/togetherai_test.go b/internal/entity/models/togetherai_test.go
index f9f5a7b79d..6ea82d6717 100644
--- a/internal/entity/models/togetherai_test.go
+++ b/internal/entity/models/togetherai_test.go
@@ -60,6 +60,7 @@ func TestTogetherAIFactory(t *testing.T) {
 }
 
 func TestTogetherAIChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTogetherAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -91,6 +92,7 @@ func TestTogetherAIChatHappyPath(t *testing.T) {
 	stop := []string{"END"}
 	effort := "high"
 	resp, err := newTogetherAIForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"openai/gpt-oss-20b",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -109,6 +111,7 @@ func TestTogetherAIChatHappyPath(t *testing.T) {
 }
 
 func TestTogetherAIChatForwardsReasoningEnabled(t *testing.T) {
+	ctx := t.Context()
 	srv := newTogetherAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "Qwen/Qwen3.5-9B" {
 			t.Errorf("model=%v", body["model"])
@@ -136,6 +139,7 @@ func TestTogetherAIChatForwardsReasoningEnabled(t *testing.T) {
 	apiKey := "test-key"
 	thinking := false
 	resp, err := newTogetherAIForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"Qwen/Qwen3.5-9B",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -151,14 +155,16 @@ func TestTogetherAIChatForwardsReasoningEnabled(t *testing.T) {
 }
 
 func TestTogetherAIChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newTogetherAIForTest("http://unused").ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newTogetherAIForTest("http://unused").ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestTogetherAIStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTogetherAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.URL.Path != "/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -182,6 +188,7 @@ func TestTogetherAIStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	err := newTogetherAIForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"meta-llama/Llama-3.3-70B-Instruct-Turbo",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -207,6 +214,7 @@ func TestTogetherAIStreamHappyPath(t *testing.T) {
 }
 
 func TestTogetherAIStreamStopsOnRootFinishReason(t *testing.T) {
+	ctx := t.Context()
 	srv := newTogetherAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w,
@@ -218,6 +226,7 @@ func TestTogetherAIStreamStopsOnRootFinishReason(t *testing.T) {
 	apiKey := "test-key"
 	var chunks []string
 	err := newTogetherAIForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"meta-llama/Llama-3.3-70B-Instruct-Turbo",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -237,6 +246,7 @@ func TestTogetherAIStreamStopsOnRootFinishReason(t *testing.T) {
 }
 
 func TestTogetherAIListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := newTogetherAIServer(t, func(t *testing.T, r *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s", r.Method)
@@ -253,28 +263,29 @@ func TestTogetherAIListModelsAndCheckConnection(t *testing.T) {
 
 	apiKey := "test-key"
 	model := newTogetherAIForTest(srv.URL)
-	models, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "openai/gpt-oss-20b,meta-llama/Llama-3.3-70B-Instruct-Turbo" {
 		t.Errorf("models=%v", models)
 	}
-	if err := model.CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := model.CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestTogetherAIUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newTogetherAIForTest("http://unused")
 	apiKey := "test-key"
 	// Rerank IS implemented; with nil documents it short-circuits to empty response (no error).
 	// It should NOT be blocked by APIConfigCheck.
-	if _, err := m.Rerank(nil, "", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
+	if _, err := m.Rerank(ctx, nil, "", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
 		t.Errorf("Rerank error=%v (expected no error for empty documents)", err)
 	}
 	// Balance IS a stub → "no such method"
-	if _, err := m.Balance(&APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Balance(ctx, &APIConfig{ApiKey: &apiKey}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance error=%v", err)
 	}
 }
diff --git a/internal/entity/models/tokenhub.go b/internal/entity/models/tokenhub.go
index a01b17d342..19cf73933d 100644
--- a/internal/entity/models/tokenhub.go
+++ b/internal/entity/models/tokenhub.go
@@ -59,7 +59,7 @@ func validateTokenHubChatRequest(modelName string, messages []Message) error {
 	return nil
 }
 
-func (t *TokenHubModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (t *TokenHubModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -113,7 +113,7 @@ func (t *TokenHubModel) ChatWithMessages(modelName string, messages []Message, a
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -184,7 +184,7 @@ func (t *TokenHubModel) ChatWithMessages(modelName string, messages []Message, a
 	return chatResponse, nil
 }
 
-func (t *TokenHubModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenHubModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -249,7 +249,7 @@ func (t *TokenHubModel) ChatStreamlyWithSender(modelName string, messages []Mess
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -315,7 +315,7 @@ func (t *TokenHubModel) ChatStreamlyWithSender(modelName string, messages []Mess
 	return sender(&endOfStream, nil)
 }
 
-func (t *TokenHubModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (t *TokenHubModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -343,7 +343,7 @@ func (t *TokenHubModel) Embed(modelName *string, texts []string, apiConfig *APIC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -395,35 +395,35 @@ func (t *TokenHubModel) Embed(modelName *string, texts []string, apiConfig *APIC
 	return embeddings, nil
 }
 
-func (t *TokenHubModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (t *TokenHubModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (t *TokenHubModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenHubModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (t *TokenHubModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenHubModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (t *TokenHubModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (t *TokenHubModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (t *TokenHubModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -434,7 +434,7 @@ func (t *TokenHubModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, t.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -490,19 +490,19 @@ func (t *TokenHubModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, e
 	return ParseListModel(modelList), nil
 }
 
-func (t *TokenHubModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (t *TokenHubModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := t.ListModels(apiConfig)
+func (t *TokenHubModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := t.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (t *TokenHubModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (t *TokenHubModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
 
-func (t *TokenHubModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (t *TokenHubModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s no such method", t.Name())
 }
diff --git a/internal/entity/models/tokenhub_test.go b/internal/entity/models/tokenhub_test.go
index ccdf0f8fd0..e91068d580 100644
--- a/internal/entity/models/tokenhub_test.go
+++ b/internal/entity/models/tokenhub_test.go
@@ -98,6 +98,7 @@ func TestTokenHubFactory(t *testing.T) {
 }
 
 func TestTokenHubChatWithMessagesForcesNonStreaming(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenHubServer(t, http.MethodPost, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["stream"] != false {
 			t.Errorf("stream=%v, want false", body["stream"])
@@ -116,6 +117,7 @@ func TestTokenHubChatWithMessagesForcesNonStreaming(t *testing.T) {
 	apiKey := "test-key"
 	stream := true
 	resp, err := newTokenHubForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -134,21 +136,24 @@ func TestTokenHubChatWithMessagesForcesNonStreaming(t *testing.T) {
 }
 
 func TestTokenHubChatRequiresAPIKey(t *testing.T) {
-	_, err := newTokenHubForTest("http://unused").ChatWithMessages("gpt-4o-mini", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
+	ctx := t.Context()
+	_, err := newTokenHubForTest("http://unused").ChatWithMessages(ctx, "gpt-4o-mini", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Fatalf("expected api-key error, got %v", err)
 	}
 }
 
 func TestTokenHubChatRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	_, err := newTokenHubForTest("http://unused").ChatWithMessages(" ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newTokenHubForTest("http://unused").ChatWithMessages(ctx, " ", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Fatalf("expected model-name error, got %v", err)
 	}
 }
 
 func TestTokenHubStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenHubSSEServer(t, "/chat/completions", strings.Join([]string{
 		`data: {"choices":[{"delta":{"reasoning_content":"thinking"}}]}`,
 		`data: {"choices":[{"delta":{"content":"hello"}}]}`,
@@ -161,6 +166,7 @@ func TestTokenHubStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	err := newTokenHubForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -188,9 +194,11 @@ func TestTokenHubStreamHappyPath(t *testing.T) {
 }
 
 func TestTokenHubStreamRejectsFalseStreamConfig(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
 	err := newTokenHubForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -204,8 +212,10 @@ func TestTokenHubStreamRejectsFalseStreamConfig(t *testing.T) {
 }
 
 func TestTokenHubStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newTokenHubForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -219,7 +229,9 @@ func TestTokenHubStreamRequiresSender(t *testing.T) {
 }
 
 func TestTokenHubStreamRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	err := newTokenHubForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"gpt-4o-mini",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{},
@@ -233,8 +245,10 @@ func TestTokenHubStreamRequiresAPIKey(t *testing.T) {
 }
 
 func TestTokenHubStreamRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newTokenHubForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		" ",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -248,6 +262,7 @@ func TestTokenHubStreamRequiresModelName(t *testing.T) {
 }
 
 func TestTokenHubEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenHubServer(t, http.MethodPost, "/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "text-embedding-3-small" {
 			t.Errorf("model=%v", body["model"])
@@ -269,7 +284,7 @@ func TestTokenHubEmbedHappyPath(t *testing.T) {
 
 	apiKey := "test-key"
 	model := "text-embedding-3-small"
-	embeddings, err := newTokenHubForTest(srv.URL).Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	embeddings, err := newTokenHubForTest(srv.URL).Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -279,20 +294,22 @@ func TestTokenHubEmbedHappyPath(t *testing.T) {
 }
 
 func TestTokenHubEmbedValidatesInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
-	if embeddings, err := newTokenHubForTest("http://unused").Embed(nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil || len(embeddings) != 0 {
+	if embeddings, err := newTokenHubForTest("http://unused").Embed(ctx, nil, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil || len(embeddings) != 0 {
 		t.Fatalf("empty input should return empty embeddings, got %#v err=%v", embeddings, err)
 	}
-	if _, err := newTokenHubForTest("http://unused").Embed(nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := newTokenHubForTest("http://unused").Embed(ctx, nil, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Fatalf("expected model-name error, got %v", err)
 	}
 	model := "text-embedding-3-small"
-	if _, err := newTokenHubForTest("http://unused").Embed(&model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := newTokenHubForTest("http://unused").Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Fatalf("expected api-key error, got %v", err)
 	}
 }
 
 func TestTokenHubListModelsHappyPathSkipsMalformedItems(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenHubServer(t, http.MethodGet, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []interface{}{
@@ -306,7 +323,7 @@ func TestTokenHubListModelsHappyPathSkipsMalformedItems(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newTokenHubForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newTokenHubForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -317,7 +334,8 @@ func TestTokenHubListModelsHappyPathSkipsMalformedItems(t *testing.T) {
 }
 
 func TestTokenHubListModelsValidatesResponseAndAPIKey(t *testing.T) {
-	if _, err := newTokenHubForTest("http://unused").ListModels(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	ctx := t.Context()
+	if _, err := newTokenHubForTest("http://unused").ListModels(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Fatalf("expected api-key error, got %v", err)
 	}
 
@@ -327,7 +345,7 @@ func TestTokenHubListModelsValidatesResponseAndAPIKey(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newTokenHubForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := newTokenHubForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "invalid models list format") {
 		t.Fatalf("expected invalid-format error, got %v", err)
 	}
diff --git a/internal/entity/models/tokenpony.go b/internal/entity/models/tokenpony.go
index 8c2fcead5e..63f304c755 100644
--- a/internal/entity/models/tokenpony.go
+++ b/internal/entity/models/tokenpony.go
@@ -52,7 +52,7 @@ func (t *TokenPonyModel) Name() string {
 }
 
 // ChatWithMessages sends a non-streaming chat request
-func (t *TokenPonyModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (t *TokenPonyModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -101,7 +101,7 @@ func (t *TokenPonyModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -162,7 +162,7 @@ func (t *TokenPonyModel) ChatWithMessages(modelName string, messages []Message,
 }
 
 // ChatStreamlyWithSender opens the SSE chat-completions
-func (t *TokenPonyModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenPonyModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -218,7 +218,7 @@ func (t *TokenPonyModel) ChatStreamlyWithSender(modelName string, messages []Mes
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -284,7 +284,7 @@ func (t *TokenPonyModel) ChatStreamlyWithSender(modelName string, messages []Mes
 	return nil
 }
 
-func (t *TokenPonyModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (t *TokenPonyModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := t.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -296,7 +296,7 @@ func (t *TokenPonyModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, t.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -332,51 +332,51 @@ func (t *TokenPonyModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 }
 
 // CheckConnection verifies the API key by calling ListModels.
-func (t *TokenPonyModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := t.ListModels(apiConfig)
+func (t *TokenPonyModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := t.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (t *TokenPonyModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (t *TokenPonyModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (t *TokenPonyModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (t *TokenPonyModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (t *TokenPonyModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenPonyModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (t *TokenPonyModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (t *TokenPonyModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (t *TokenPonyModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (t *TokenPonyModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (t *TokenPonyModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
 
-func (t *TokenPonyModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (t *TokenPonyModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", t.Name())
 }
diff --git a/internal/entity/models/tokenpony_test.go b/internal/entity/models/tokenpony_test.go
index 5f1094f40a..d8b3674faf 100644
--- a/internal/entity/models/tokenpony_test.go
+++ b/internal/entity/models/tokenpony_test.go
@@ -92,6 +92,7 @@ func TestTokenPonyFactory(t *testing.T) {
 }
 
 func TestTokenPonyChatHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "qwen3-32b" {
 			t.Errorf("model=%v", body["model"])
@@ -120,6 +121,7 @@ func TestTokenPonyChatHappyPath(t *testing.T) {
 	mt := 64
 	temp := 0.3
 	resp, err := newTokenPonyForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -138,6 +140,7 @@ func TestTokenPonyChatHappyPath(t *testing.T) {
 }
 
 func TestTokenPonyChatNoReasoning(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -149,6 +152,7 @@ func TestTokenPonyChatNoReasoning(t *testing.T) {
 
 	apiKey := "test-key"
 	resp, err := newTokenPonyForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-8b",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -164,7 +168,9 @@ func TestTokenPonyChatNoReasoning(t *testing.T) {
 }
 
 func TestTokenPonyChatRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	_, err := newTokenPonyForTest("http://unused").ChatWithMessages(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil)
@@ -174,15 +180,17 @@ func TestTokenPonyChatRequiresAPIKey(t *testing.T) {
 }
 
 func TestTokenPonyChatRequiresMessages(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	_, err := newTokenPonyForTest("http://unused").ChatWithMessages(
-		"qwen3-32b", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
+		ctx, "qwen3-32b", nil, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("expected messages-empty error, got %v", err)
 	}
 }
 
 func TestTokenPonyChatPropagatesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/chat/completions", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"bad key"}`))
@@ -191,6 +199,7 @@ func TestTokenPonyChatPropagatesHTTPError(t *testing.T) {
 
 	apiKey := "test-key"
 	_, err := newTokenPonyForTest(srv.URL).ChatWithMessages(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
@@ -200,6 +209,7 @@ func TestTokenPonyChatPropagatesHTTPError(t *testing.T) {
 }
 
 func TestTokenPonyStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonySSEServer(t, "/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"content":"Hello"}}]}`+"\n"+
@@ -212,6 +222,7 @@ func TestTokenPonyStreamHappyPath(t *testing.T) {
 	var chunks []string
 	var sawDone bool
 	err := newTokenPonyForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -238,6 +249,7 @@ func TestTokenPonyStreamHappyPath(t *testing.T) {
 }
 
 func TestTokenPonyStreamSplitsReasoning(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonySSEServer(t, "/chat/completions",
 		`data: {"choices":[{"index":0,"delta":{"role":"assistant"}}]}`+"\n"+
 			`data: {"choices":[{"index":0,"delta":{"reasoning_content":"step 1. "}}]}`+"\n"+
@@ -250,6 +262,7 @@ func TestTokenPonyStreamSplitsReasoning(t *testing.T) {
 	apiKey := "test-key"
 	var content, reasoning []string
 	err := newTokenPonyForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"deepseek-r1-0528",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -277,9 +290,11 @@ func TestTokenPonyStreamSplitsReasoning(t *testing.T) {
 }
 
 func TestTokenPonyStreamRejectsExplicitFalse(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	stream := false
 	err := newTokenPonyForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -292,8 +307,10 @@ func TestTokenPonyStreamRejectsExplicitFalse(t *testing.T) {
 }
 
 func TestTokenPonyStreamRequiresSender(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newTokenPonyForTest("http://unused").ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil, nil)
@@ -303,6 +320,7 @@ func TestTokenPonyStreamRequiresSender(t *testing.T) {
 }
 
 func TestTokenPonyStreamFailsWithoutTerminal(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonySSEServer(t, "/chat/completions",
 		`data: {"choices":[{"delta":{"content":"half"}}]}`+"\n",
 	)
@@ -310,6 +328,7 @@ func TestTokenPonyStreamFailsWithoutTerminal(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newTokenPonyForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -320,6 +339,7 @@ func TestTokenPonyStreamFailsWithoutTerminal(t *testing.T) {
 }
 
 func TestTokenPonyStreamRejectsMalformedFrame(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonySSEServer(t, "/chat/completions",
 		`data: {"choices":[{"delta":{"content":"ok"}}]}`+"\n"+
 			`data: {oops not json}`+"\n",
@@ -328,6 +348,7 @@ func TestTokenPonyStreamRejectsMalformedFrame(t *testing.T) {
 
 	apiKey := "test-key"
 	err := newTokenPonyForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -344,8 +365,10 @@ func TestTokenPonyStreamSurfacesUpstreamError(t *testing.T) {
 	)
 	defer srv.Close()
 
+	ctx := t.Context()
 	apiKey := "test-key"
 	err := newTokenPonyForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"qwen3-32b",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
@@ -359,6 +382,7 @@ func TestTokenPonyStreamSurfacesUpstreamError(t *testing.T) {
 }
 
 func TestTokenPonyListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -371,7 +395,7 @@ func TestTokenPonyListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newTokenPonyForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newTokenPonyForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -382,13 +406,15 @@ func TestTokenPonyListModelsHappyPath(t *testing.T) {
 }
 
 func TestTokenPonyListModelsRequiresAPIKey(t *testing.T) {
-	_, err := newTokenPonyForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newTokenPonyForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestTokenPonyCheckConnectionDelegatesToListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{{"id": "qwen3-32b"}},
@@ -397,12 +423,13 @@ func TestTokenPonyCheckConnectionDelegatesToListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newTokenPonyForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newTokenPonyForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 }
 
 func TestTokenPonyCheckConnectionPropagatesError(t *testing.T) {
+	ctx := t.Context()
 	srv := newTokenPonyServer(t, "/models", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = w.Write([]byte(`{"error":"bad key"}`))
@@ -410,40 +437,43 @@ func TestTokenPonyCheckConnectionPropagatesError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	err := newTokenPonyForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey})
+	err := newTokenPonyForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestTokenPonyBaseURLForRegionUnknown(t *testing.T) {
+	ctx := t.Context()
 	m := newTokenPonyForTest("http://unused")
 	apiKey := "test-key"
 	region := "missing"
-	_, err := m.ListModels(&APIConfig{ApiKey: &apiKey, Region: &region})
+	_, err := m.ListModels(ctx, &APIConfig{ApiKey: &apiKey, Region: &region})
 	if err == nil || !strings.Contains(err.Error(), "no base URL configured") {
 		t.Errorf("expected base-URL error, got %v", err)
 	}
 }
 
 func TestTokenPonyEmbedReturnsNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	model := "x"
-	_, err := newTokenPonyForTest("http://unused").Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := newTokenPonyForTest("http://unused").Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: want 'no such method', got %v", err)
 	}
 }
 
 func TestTokenPonyAudioOCRReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	m := newTokenPonyForTest("http://unused")
 	model := "x"
-	if _, err := m.TranscribeAudio(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.TranscribeAudio(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
-	if _, err := m.AudioSpeech(&model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.AudioSpeech(ctx, &model, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeech: %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, &model, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
diff --git a/internal/entity/models/tool_call_test.go b/internal/entity/models/tool_call_test.go
index 6bedbd911c..c13a536dbb 100644
--- a/internal/entity/models/tool_call_test.go
+++ b/internal/entity/models/tool_call_test.go
@@ -24,6 +24,7 @@ import (
 )
 
 func testNonStreamingToolCall(t *testing.T, modelName, path string, newDriver func(string) ModelDriver) {
+	ctx := t.Context()
 	t.Helper()
 	var requestBody map[string]interface{}
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -39,6 +40,7 @@ func testNonStreamingToolCall(t *testing.T, modelName, path string, newDriver fu
 
 	apiKey, toolChoice := "test-key", "required"
 	response, err := newDriver(server.URL).ChatWithMessages(
+		ctx,
 		modelName,
 		[]Message{
 			{Role: "assistant", ToolCalls: []map[string]interface{}{{"id": "call-1"}}},
@@ -65,6 +67,7 @@ func testNonStreamingToolCall(t *testing.T, modelName, path string, newDriver fu
 }
 
 func testStreamingToolCall(t *testing.T, modelName, path string, newDriver func(string) ModelDriver) {
+	ctx := t.Context()
 	t.Helper()
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != path {
@@ -86,6 +89,7 @@ data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ra
 		ToolCallsResult: &[]map[string]interface{}{{"id": "stale"}},
 	}
 	err := newDriver(server.URL).ChatStreamlyWithSender(
+		ctx,
 		modelName,
 		[]Message{{Role: "user", Content: "find ragflow"}},
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/entity/models/types.go b/internal/entity/models/types.go
index 6987cfd8b6..a16864243a 100644
--- a/internal/entity/models/types.go
+++ b/internal/entity/models/types.go
@@ -1,6 +1,7 @@
 package models
 
 import (
+	"context"
 	"encoding/json"
 	"ragflow/internal/common"
 )
@@ -30,33 +31,33 @@ type ModelDriver interface {
 	Name() string
 
 	// ChatWithMessages sends multiple messages synchronously
-	ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error)
+	ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error)
 	// ChatStreamlyWithSender sends multiple messages asynchronously
-	ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
+	ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
 	// Embed a list of texts into embeddings
-	Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error)
+	Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error)
 	// Rerank calculates similarity scores between query and texts
-	Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error)
+	Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error)
 	// TranscribeAudio transcribe audio
-	TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error)
-	TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
+	TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error)
+	TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
 	// AudioSpeech convert text to audio
-	AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error)
-	AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
+	AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error)
+	AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error
 	// OCRFile OCR file
-	OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error)
+	OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error)
 	// ParseFile parse file
-	ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error)
+	ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error)
 	// ListModels List supported models
-	ListModels(apiConfig *APIConfig) ([]ListModelResponse, error)
+	ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error)
 
-	Balance(apiConfig *APIConfig) (map[string]interface{}, error)
+	Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error)
 
-	CheckConnection(apiConfig *APIConfig) error
+	CheckConnection(ctx context.Context, apiConfig *APIConfig) error
 
-	ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error)
+	ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error)
 
-	ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error)
+	ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error)
 }
 
 type ChatResponse struct {
@@ -252,8 +253,8 @@ func NewRerankModel(driver ModelDriver, modelName *string, apiConfig *APIConfig)
 }
 
 // Rerank calculates similarity between query and texts
-func (r *RerankModel) Rerank(query string, texts []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
-	return r.ModelDriver.Rerank(r.ModelName, query, texts, apiConfig, rerankConfig, modelUsage)
+func (r *RerankModel) Rerank(ctx context.Context, query string, texts []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+	return r.ModelDriver.Rerank(ctx, r.ModelName, query, texts, apiConfig, rerankConfig, modelUsage)
 }
 
 // ToolConfig bundles tool-calling configuration for a ChatModel.
diff --git a/internal/entity/models/upstage.go b/internal/entity/models/upstage.go
index de0751d4d1..c22d2a4a46 100644
--- a/internal/entity/models/upstage.go
+++ b/internal/entity/models/upstage.go
@@ -52,7 +52,7 @@ func (u *UpstageModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (u *UpstageModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (u *UpstageModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := u.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -105,7 +105,7 @@ func (u *UpstageModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -168,7 +168,7 @@ func (u *UpstageModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (u *UpstageModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (u *UpstageModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := u.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -230,7 +230,7 @@ func (u *UpstageModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(context.Background(), "POST", url, bytes.NewBuffer(jsonData))
+	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return fmt.Errorf("failed to create request: %w", err)
 	}
@@ -312,7 +312,7 @@ type upstageEmbeddingResponse struct {
 }
 
 // Embed turns a list of texts into embedding vectors
-func (u *UpstageModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (u *UpstageModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := u.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -342,7 +342,7 @@ func (u *UpstageModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -398,7 +398,7 @@ func (u *UpstageModel) Embed(modelName *string, texts []string, apiConfig *APICo
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (u *UpstageModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (u *UpstageModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := u.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -410,7 +410,7 @@ func (u *UpstageModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	baseURL = strings.TrimSuffix(baseURL, "/")
 	url := fmt.Sprintf("%s/%s", baseURL, u.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -448,13 +448,13 @@ func (u *UpstageModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 }
 
 // Balance is not exposed by the Upstage API, so this returns "no such method".
-func (u *UpstageModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (u *UpstageModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (u *UpstageModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := u.ListModels(apiConfig)
+func (u *UpstageModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := u.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -463,42 +463,42 @@ func (u *UpstageModel) CheckConnection(apiConfig *APIConfig) error {
 
 // Rerank calculates similarity scores between query and documents. Upstage
 // does not expose a public rerank API, so this returns "no such method".
-func (u *UpstageModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (u *UpstageModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // TranscribeAudio transcribe audio
-func (u *UpstageModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (u *UpstageModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
 
-func (u *UpstageModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (u *UpstageModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", u.Name())
 }
 
 // AudioSpeech convert text to audio
-func (u *UpstageModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (u *UpstageModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
 
-func (u *UpstageModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (u *UpstageModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", u.Name())
 }
 
 // OCRFile OCR file
-func (u *UpstageModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (u *UpstageModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
 
 // ParseFile parse file
-func (u *UpstageModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (u *UpstageModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
 
-func (u *UpstageModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (u *UpstageModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
 
-func (u *UpstageModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (u *UpstageModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", u.Name())
 }
diff --git a/internal/entity/models/upstage_test.go b/internal/entity/models/upstage_test.go
index 8e81cec111..86d5df3cd0 100644
--- a/internal/entity/models/upstage_test.go
+++ b/internal/entity/models/upstage_test.go
@@ -23,6 +23,7 @@ func newUpstageForTest(baseURL string) *UpstageModel {
 // ---------- reasoning_effort / reasoning field ----------
 
 func TestUpstageChatPropagatesReasoningEffort(t *testing.T) {
+	ctx := t.Context()
 	// Per https://console.upstage.ai/api/docs/for-agents/raw, Upstage
 	// Solar models accept `reasoning_effort: minimal|low|medium|high`.
 	// ChatConfig.Effort is the canonical carrier; this test asserts it
@@ -38,7 +39,7 @@ func TestUpstageChatPropagatesReasoningEffort(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	effort := "high"
-	_, err := u.ChatWithMessages("solar-pro2",
+	_, err := u.ChatWithMessages(ctx, "solar-pro2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Effort: &effort}, nil)
@@ -51,6 +52,7 @@ func TestUpstageChatPropagatesReasoningEffort(t *testing.T) {
 }
 
 func TestUpstageChatOmitsReasoningEffortWhenUnset(t *testing.T) {
+	ctx := t.Context()
 	// If the caller does not opt in, the field must NOT be sent. Sending
 	// "minimal" by default would silently change behavior for downstream
 	// proxies that treat a present field differently from an absent one.
@@ -64,7 +66,7 @@ func TestUpstageChatOmitsReasoningEffortWhenUnset(t *testing.T) {
 
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
-	_, err := u.ChatWithMessages("solar-pro2",
+	_, err := u.ChatWithMessages(ctx, "solar-pro2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{}, // no Effort
@@ -79,6 +81,7 @@ func TestUpstageChatOmitsReasoningEffortWhenUnset(t *testing.T) {
 }
 
 func TestUpstageStreamPropagatesReasoningEffort(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, _ := io.ReadAll(r.Body)
@@ -94,7 +97,7 @@ func TestUpstageStreamPropagatesReasoningEffort(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	effort := "medium"
-	err := u.ChatStreamlyWithSender("solar-pro2",
+	err := u.ChatStreamlyWithSender(ctx, "solar-pro2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{Effort: &effort},
@@ -110,6 +113,7 @@ func TestUpstageStreamPropagatesReasoningEffort(t *testing.T) {
 }
 
 func TestUpstageChatExtractsReasoningField(t *testing.T) {
+	ctx := t.Context()
 	// Per the Upstage docs: when reasoning_effort is high|medium for
 	// solar-pro3 (or high for solar-pro2), the response's
 	// choices[0].message includes a `reasoning` field. The driver must
@@ -124,7 +128,7 @@ func TestUpstageChatExtractsReasoningField(t *testing.T) {
 
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := u.ChatWithMessages("solar-pro3",
+	resp, err := u.ChatWithMessages(ctx, "solar-pro3",
 		[]Message{{Role: "user", Content: "What is 15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -139,6 +143,7 @@ func TestUpstageChatExtractsReasoningField(t *testing.T) {
 }
 
 func TestUpstageChatHandlesAbsentReasoning(t *testing.T) {
+	ctx := t.Context()
 	// Models without reasoning (solar-mini, syn-pro) or low-effort
 	// requests return no `reasoning` field. The driver must leave
 	// ReasonContent empty without crashing.
@@ -149,7 +154,7 @@ func TestUpstageChatHandlesAbsentReasoning(t *testing.T) {
 
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
-	resp, err := u.ChatWithMessages("solar-mini",
+	resp, err := u.ChatWithMessages(ctx, "solar-mini",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
@@ -167,6 +172,7 @@ func TestUpstageChatHandlesAbsentReasoning(t *testing.T) {
 // https://console.upstage.ai/api/chat) round-trips through the request
 // body for both streaming and non-streaming paths.
 func TestUpstageRequestBodyMatchesSolarAPIShape(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		raw, _ := io.ReadAll(r.Body)
@@ -182,7 +188,7 @@ func TestUpstageRequestBodyMatchesSolarAPIShape(t *testing.T) {
 	topP := 0.9
 	stop := []string{"END"}
 	effort := "high"
-	_, err := u.ChatWithMessages("solar-pro2",
+	_, err := u.ChatWithMessages(ctx, "solar-pro2",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey},
 		&ChatConfig{MaxTokens: &mt, Temperature: &temp, TopP: &topP, Stop: &stop, Effort: &effort}, nil)
@@ -215,6 +221,7 @@ func TestUpstageRequestBodyMatchesSolarAPIShape(t *testing.T) {
 // ---------- Embed: duplicate / out-of-range / reorder ----------
 
 func TestUpstageEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	// A malformed upstream that repeats data[*].index would silently
 	// overwrite the earlier vector; the driver must fail loudly instead.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
@@ -227,13 +234,14 @@ func TestUpstageEmbedRejectsDuplicateIndex(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "solar-embedding-1-large-passage"
-	_, err := u.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := u.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index 0") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
 }
 
 func TestUpstageEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"data":[{"embedding":[1],"index":7}]}`)
 	}))
@@ -242,13 +250,14 @@ func TestUpstageEmbedRejectsOutOfRangeIndex(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "solar-embedding-1-large-passage"
-	_, err := u.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := u.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestUpstageEmbedHappyPathReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	// Upstream returns vectors in shuffled order; driver must realign.
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"data":[
@@ -261,7 +270,7 @@ func TestUpstageEmbedHappyPathReordersByIndex(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "solar-embedding-1-large-passage"
-	vecs, err := u.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := u.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -281,6 +290,7 @@ func TestUpstageEmbedHappyPathReordersByIndex(t *testing.T) {
 // reasoning_effort=high — both fields appear, sometimes in the same
 // chunk and sometimes separately.
 func TestUpstageStreamExtractsReasoningDelta(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		w.WriteHeader(http.StatusOK)
@@ -299,7 +309,7 @@ func TestUpstageStreamExtractsReasoningDelta(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	var contentChunks, reasoningChunks []string
-	err := u.ChatStreamlyWithSender("solar-pro3",
+	err := u.ChatStreamlyWithSender(ctx, "solar-pro3",
 		[]Message{{Role: "user", Content: "What is 15% of 80?"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(content *string, reason *string) error {
@@ -338,6 +348,7 @@ func TestUpstageStreamExtractsReasoningDelta(t *testing.T) {
 // present: reasoning is forwarded first so a UI consuming both can
 // render the chain-of-thought before the answer for that token.
 func TestUpstageStreamReasoningChunksArriveBeforeContent(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		w.WriteHeader(http.StatusOK)
@@ -354,7 +365,7 @@ func TestUpstageStreamReasoningChunksArriveBeforeContent(t *testing.T) {
 	u := newUpstageForTest(srv.URL)
 	apiKey := "test-key"
 	var seq []string
-	err := u.ChatStreamlyWithSender("solar-pro3",
+	err := u.ChatStreamlyWithSender(ctx, "solar-pro3",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(content *string, reason *string) error {
@@ -385,6 +396,7 @@ func TestUpstageStreamReasoningChunksArriveBeforeContent(t *testing.T) {
 // non-reasoning models (solar-mini, solar-pro2 with no reasoning_effort)
 // emit only delta.content. The driver must not regress on them.
 func TestUpstageStreamWithoutReasoningStillWorks(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		w.WriteHeader(http.StatusOK)
@@ -400,7 +412,7 @@ func TestUpstageStreamWithoutReasoningStillWorks(t *testing.T) {
 	apiKey := "test-key"
 	var content []string
 	var reasonCalled bool
-	err := u.ChatStreamlyWithSender("solar-mini",
+	err := u.ChatStreamlyWithSender(ctx, "solar-mini",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &apiKey}, nil, nil,
 		func(c *string, r *string) error {
diff --git a/internal/entity/models/vllm.go b/internal/entity/models/vllm.go
index ff8792bca0..6c497a0405 100644
--- a/internal/entity/models/vllm.go
+++ b/internal/entity/models/vllm.go
@@ -53,7 +53,7 @@ func (v *VllmModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (v *VllmModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (v *VllmModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -130,7 +130,7 @@ func (v *VllmModel) ChatWithMessages(modelName string, messages []Message, apiCo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -204,7 +204,7 @@ func (v *VllmModel) ChatWithMessages(modelName string, messages []Message, apiCo
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (v *VllmModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VllmModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -280,7 +280,7 @@ func (v *VllmModel) ChatStreamlyWithSender(modelName string, messages []Message,
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -360,7 +360,7 @@ type vllmEmbeddingResponse struct {
 }
 
 // Embed embeds a list of texts into embeddings
-func (v *VllmModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (v *VllmModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -400,7 +400,7 @@ func (v *VllmModel) Embed(modelName *string, texts []string, apiConfig *APIConfi
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -444,7 +444,7 @@ func (v *VllmModel) Embed(modelName *string, texts []string, apiConfig *APIConfi
 	return embeddings, nil
 }
 
-func (v *VllmModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (v *VllmModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -470,7 +470,7 @@ func (v *VllmModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -511,13 +511,13 @@ func (v *VllmModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error
 	return ParseListModel(modelList), nil
 }
 
-func (v *VllmModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (v *VllmModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection verifies that the configured vLLM base URL is reachable
-func (v *VllmModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := v.ListModels(apiConfig)
+func (v *VllmModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := v.ListModels(ctx, apiConfig)
 	return err
 }
 
@@ -550,7 +550,7 @@ type vllmRerankResponse struct {
 // Authorization header is sent only when APIConfig.ApiKey is non-empty,
 // matching the existing Embed/ListModels behaviour for this local
 // driver.
-func (v *VllmModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (v *VllmModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -593,7 +593,7 @@ func (v *VllmModel) Rerank(modelName *string, query string, documents []string,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -641,37 +641,37 @@ func (v *VllmModel) Rerank(modelName *string, query string, documents []string,
 }
 
 // TranscribeAudio transcribe audio
-func (v *VllmModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (v *VllmModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VllmModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VllmModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // AudioSpeech convert text to audio
-func (v *VllmModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (v *VllmModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VllmModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VllmModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // OCRFile OCR file
-func (v *VllmModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (v *VllmModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
 // ParseFile parse file
-func (v *VllmModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (v *VllmModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VllmModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (v *VllmModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VllmModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (v *VllmModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
diff --git a/internal/entity/models/vllm_rerank_test.go b/internal/entity/models/vllm_rerank_test.go
index 19203a1e76..d7d4f3db12 100644
--- a/internal/entity/models/vllm_rerank_test.go
+++ b/internal/entity/models/vllm_rerank_test.go
@@ -50,6 +50,7 @@ func newVllmModelForTest(baseURL string) *VllmModel {
 }
 
 func TestVllmRerankHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newVllmRerankServer(t, "Bearer test-key", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "BAAI/bge-reranker-v2-m3" {
 			t.Errorf("expected model=BAAI/bge-reranker-v2-m3, got %v", body["model"])
@@ -85,6 +86,7 @@ func TestVllmRerankHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "BAAI/bge-reranker-v2-m3"
 	resp, err := model.Rerank(
+		ctx,
 		&modelName,
 		"What is RAPTOR?",
 		[]string{"doc-zero", "doc-one", "doc-two"},
@@ -107,6 +109,7 @@ func TestVllmRerankHappyPath(t *testing.T) {
 }
 
 func TestVllmRerankTopNClamp(t *testing.T) {
+	ctx := t.Context()
 	srv := newVllmRerankServer(t, "Bearer test-key", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["top_n"] != float64(2) {
 			t.Errorf("expected top_n clamp to RerankConfig.TopN=2, got %v", body["top_n"])
@@ -119,6 +122,7 @@ func TestVllmRerankTopNClamp(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "BAAI/bge-reranker-v2-m3"
 	if _, err := model.Rerank(
+		ctx,
 		&modelName, "q",
 		[]string{"a", "b", "c", "d"},
 		&APIConfig{ApiKey: &apiKey},
@@ -130,10 +134,11 @@ func TestVllmRerankTopNClamp(t *testing.T) {
 }
 
 func TestVllmRerankEmptyDocuments(t *testing.T) {
+	ctx := t.Context()
 	model := newVllmModelForTest("http://unused")
 	apiKey := "test-key"
 	modelName := "BAAI/bge-reranker-v2-m3"
-	resp, err := model.Rerank(&modelName, "q", nil, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	resp, err := model.Rerank(ctx, &modelName, "q", nil, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err != nil {
 		t.Fatalf("expected nil error for empty documents, got %v", err)
 	}
@@ -146,6 +151,7 @@ func TestVllmRerankEmptyDocuments(t *testing.T) {
 // no APIConfig.ApiKey is configured. This diverges from the NVIDIA driver
 // which requires an API key.
 func TestVllmRerankWithoutAPIKey(t *testing.T) {
+	ctx := t.Context()
 	srv := newVllmRerankServer(t, "", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"results": []map[string]interface{}{
@@ -157,7 +163,7 @@ func TestVllmRerankWithoutAPIKey(t *testing.T) {
 
 	model := newVllmModelForTest(srv.URL)
 	modelName := "BAAI/bge-reranker-v2-m3"
-	resp, err := model.Rerank(&modelName, "q", []string{"a"}, &APIConfig{}, &RerankConfig{}, nil)
+	resp, err := model.Rerank(ctx, &modelName, "q", []string{"a"}, &APIConfig{}, &RerankConfig{}, nil)
 	if err != nil {
 		t.Fatalf("Rerank failed without api key: %v", err)
 	}
@@ -167,15 +173,17 @@ func TestVllmRerankWithoutAPIKey(t *testing.T) {
 }
 
 func TestVllmRerankRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	model := newVllmModelForTest("http://unused")
 	apiKey := "test-key"
-	_, err := model.Rerank(nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, nil, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestVllmRerankRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newVllmRerankServer(t, "Bearer test-key", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusInternalServerError)
 		_, _ = w.Write([]byte(`{"error":"boom"}`))
@@ -185,13 +193,14 @@ func TestVllmRerankRejectsHTTPError(t *testing.T) {
 	model := newVllmModelForTest(srv.URL)
 	apiKey := "test-key"
 	modelName := "BAAI/bge-reranker-v2-m3"
-	_, err := model.Rerank(&modelName, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, &modelName, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "vLLM rerank API error") {
 		t.Errorf("expected API error, got %v", err)
 	}
 }
 
 func TestVllmRerankRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newVllmRerankServer(t, "Bearer test-key", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"results": []map[string]interface{}{
@@ -204,7 +213,7 @@ func TestVllmRerankRejectsOutOfRangeIndex(t *testing.T) {
 	model := newVllmModelForTest(srv.URL)
 	apiKey := "test-key"
 	modelName := "BAAI/bge-reranker-v2-m3"
-	_, err := model.Rerank(&modelName, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
+	_, err := model.Rerank(ctx, &modelName, "q", []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err == nil || !strings.Contains(err.Error(), "unexpected rerank index") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
diff --git a/internal/entity/models/volcengine.go b/internal/entity/models/volcengine.go
index 1fb7b94e11..9920188792 100644
--- a/internal/entity/models/volcengine.go
+++ b/internal/entity/models/volcengine.go
@@ -73,7 +73,7 @@ func (v *VolcEngine) getAPIKey(apiConfig *APIConfig) string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (v *VolcEngine) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (v *VolcEngine) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -180,7 +180,7 @@ func (v *VolcEngine) ChatWithMessages(modelName string, messages []Message, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -264,7 +264,7 @@ func (v *VolcEngine) ChatWithMessages(modelName string, messages []Message, apiC
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (v *VolcEngine) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VolcEngine) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -386,7 +386,7 @@ func (v *VolcEngine) ChatStreamlyWithSender(modelName string, messages []Message
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -544,7 +544,7 @@ type volcenginePromptTokensDetails struct {
 }
 
 // Embed embeds a list of texts into embeddings
-func (v *VolcEngine) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (v *VolcEngine) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -591,7 +591,7 @@ func (v *VolcEngine) Embed(modelName *string, texts []string, apiConfig *APIConf
 		parsed, err := func() (volcengineEmbeddingResponse, error) {
 			var parsed volcengineEmbeddingResponse
 
-			ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+			ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 			defer cancel()
 
 			req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -636,35 +636,35 @@ func (v *VolcEngine) Embed(modelName *string, texts []string, apiConfig *APIConf
 }
 
 // Rerank calculates similarity scores between query and documents
-func (v *VolcEngine) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (v *VolcEngine) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", v.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (v *VolcEngine) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (v *VolcEngine) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VolcEngine) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VolcEngine) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // AudioSpeech convert text to audio
-func (v *VolcEngine) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (v *VolcEngine) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VolcEngine) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VolcEngine) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // OCRFile OCR file
-func (v *VolcEngine) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (v *VolcEngine) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
 // ParseFile parse file
-func (v *VolcEngine) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (v *VolcEngine) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
@@ -743,7 +743,7 @@ func inferVolcengineModelTypes(m volcengineModelData) []string {
 	return types
 }
 
-func (v *VolcEngine) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (v *VolcEngine) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -760,7 +760,7 @@ func (v *VolcEngine) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(resolvedBaseURL, "/"), modelsSuffix)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -825,11 +825,11 @@ func (v *VolcEngine) ListModels(apiConfig *APIConfig) ([]ListModelResponse, erro
 	return enrichedModels, nil
 }
 
-func (v *VolcEngine) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (v *VolcEngine) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VolcEngine) CheckConnection(apiConfig *APIConfig) error {
+func (v *VolcEngine) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -840,7 +840,7 @@ func (v *VolcEngine) CheckConnection(apiConfig *APIConfig) error {
 	}
 	url := fmt.Sprintf("%s/%s", resolvedBaseURL, v.baseModel.URLSuffix.Files)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -869,10 +869,10 @@ func (v *VolcEngine) CheckConnection(apiConfig *APIConfig) error {
 	return nil
 }
 
-func (v *VolcEngine) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (v *VolcEngine) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VolcEngine) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (v *VolcEngine) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
diff --git a/internal/entity/models/volcengine_test.go b/internal/entity/models/volcengine_test.go
index c7870d4ad4..dbde616be5 100644
--- a/internal/entity/models/volcengine_test.go
+++ b/internal/entity/models/volcengine_test.go
@@ -55,6 +55,7 @@ func TestVolcEngineConfigDeclaresModelsSuffix(t *testing.T) {
 }
 
 func TestVolcEngineListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newVolcEngineServer(t, func(t *testing.T, r *http.Request, w http.ResponseWriter) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s", r.Method)
@@ -77,7 +78,7 @@ func TestVolcEngineListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newVolcEngineForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newVolcEngineForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -87,23 +88,25 @@ func TestVolcEngineListModelsHappyPath(t *testing.T) {
 }
 
 func TestVolcEngineListModelsRejectsProviderError(t *testing.T) {
+	ctx := t.Context()
 	srv := newVolcEngineServer(t, func(t *testing.T, r *http.Request, w http.ResponseWriter) {
 		http.Error(w, "bad key", http.StatusUnauthorized)
 	})
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newVolcEngineForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := newVolcEngineForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "401 Unauthorized") || !strings.Contains(err.Error(), "bad key") {
 		t.Fatalf("expected provider error with status and body, got %v", err)
 	}
 }
 
 func TestVolcEngineListModelsRequiresModelsSuffix(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	model := NewVolcEngine(map[string]string{"default": "http://unused"}, URLSuffix{})
 
-	_, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "models URL suffix is not configured") {
 		t.Fatalf("expected missing models suffix error, got %v", err)
 	}
diff --git a/internal/entity/models/voyage.go b/internal/entity/models/voyage.go
index 247f0dded3..25d1a86db2 100644
--- a/internal/entity/models/voyage.go
+++ b/internal/entity/models/voyage.go
@@ -63,7 +63,7 @@ type voyageEmbeddingResponse struct {
 	Model  string                `json:"model"`
 }
 
-func (v *VoyageModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (v *VoyageModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -96,7 +96,7 @@ func (v *VoyageModel) Embed(modelName *string, texts []string, apiConfig *APICon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -167,7 +167,7 @@ type voyageRerankResponse struct {
 	Model string `json:"model"`
 }
 
-func (v *VoyageModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (v *VoyageModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := v.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -203,7 +203,7 @@ func (v *VoyageModel) Rerank(modelName *string, query string, documents []string
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -253,59 +253,59 @@ func (v *VoyageModel) Rerank(modelName *string, query string, documents []string
 	return rerankResponse, nil
 }
 
-func (v *VoyageModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (v *VoyageModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) CheckConnection(apiConfig *APIConfig) error {
+func (v *VoyageModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // ChatWithMessages is not exposed by the Voyage AI API.
-func (v *VoyageModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (v *VoyageModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VoyageModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
 // Balance is not exposed by the Voyage AI API.
-func (v *VoyageModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (v *VoyageModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
 // TranscribeAudio / AudioSpeech / OCRFile: Voyage does not host any of
 // these surfaces.
-func (v *VoyageModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (v *VoyageModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VoyageModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (v *VoyageModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (v *VoyageModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (v *VoyageModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
 // ParseFile parse file
-func (v *VoyageModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (v *VoyageModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (v *VoyageModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
 
-func (v *VoyageModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (v *VoyageModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", v.Name())
 }
diff --git a/internal/entity/models/voyage_test.go b/internal/entity/models/voyage_test.go
index e4f978fd82..3b0b255ff5 100644
--- a/internal/entity/models/voyage_test.go
+++ b/internal/entity/models/voyage_test.go
@@ -66,6 +66,7 @@ func TestVoyageNewModelWithCustomDefaultTransport(t *testing.T) {
 }
 
 func TestVoyageEmbedHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "voyage-3.5" {
 			t.Errorf("model=%v", body["model"])
@@ -85,7 +86,7 @@ func TestVoyageEmbedHappyPath(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	vecs, err := v.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := v.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -102,6 +103,7 @@ func TestVoyageEmbedHappyPath(t *testing.T) {
 // "dimensions" returns "Argument 'dimensions' is not supported by our
 // API"), so this name matters and must not regress.
 func TestVoyageEmbedPropagatesOutputDimension(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if got, ok := body["output_dimension"].(float64); !ok || got != 256 {
 			t.Errorf("output_dimension=%v want 256", body["output_dimension"])
@@ -120,7 +122,7 @@ func TestVoyageEmbedPropagatesOutputDimension(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"x"}, &APIConfig{ApiKey: &apiKey},
+	_, err := v.Embed(ctx, &model, []string{"x"}, &APIConfig{ApiKey: &apiKey},
 		&EmbeddingConfig{Dimension: 256}, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
@@ -131,6 +133,7 @@ func TestVoyageEmbedPropagatesOutputDimension(t *testing.T) {
 // would default the vector length, but only if we don't send the key
 // at all (sending output_dimension: 0 is a 400).
 func TestVoyageEmbedOmitsOutputDimensionWhenUnset(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if _, present := body["output_dimension"]; present {
 			t.Errorf("output_dimension must be absent when Dimension is unset, got %v", body["output_dimension"])
@@ -144,13 +147,14 @@ func TestVoyageEmbedOmitsOutputDimensionWhenUnset(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := v.Embed(ctx, &model, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 }
 
 func TestVoyageEmbedReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -165,7 +169,7 @@ func TestVoyageEmbedReordersByIndex(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	vecs, err := v.Embed(&model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := v.Embed(ctx, &model, []string{"a", "b", "c"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -177,6 +181,7 @@ func TestVoyageEmbedReordersByIndex(t *testing.T) {
 }
 
 func TestVoyageEmbedEmptyInputShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		t.Error("Embed([]) made an unexpected HTTP call")
 	}))
@@ -185,31 +190,34 @@ func TestVoyageEmbedEmptyInputShortCircuits(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	vecs, err := v.Embed(&model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	vecs, err := v.Embed(ctx, &model, []string{}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil || len(vecs) != 0 {
 		t.Errorf("Embed([])=(%v,%v)", vecs, err)
 	}
 }
 
 func TestVoyageEmbedRequiresAPIKey(t *testing.T) {
+	ctx := t.Context()
 	v := newVoyageForTest("http://unused")
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := v.Embed(ctx, &model, []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("expected api-key error, got %v", err)
 	}
 }
 
 func TestVoyageEmbedRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	v := newVoyageForTest("http://unused")
 	apiKey := "test-key"
-	_, err := v.Embed(nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := v.Embed(ctx, nil, []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestVoyageEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -223,13 +231,14 @@ func TestVoyageEmbedRejectsDuplicateIndex(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := v.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index 0") {
 		t.Errorf("expected duplicate error, got %v", err)
 	}
 }
 
 func TestVoyageEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -242,13 +251,14 @@ func TestVoyageEmbedRejectsOutOfRangeIndex(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := v.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestVoyageEmbedRejectsMissingSlot(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/embeddings", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -261,13 +271,14 @@ func TestVoyageEmbedRejectsMissingSlot(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	_, err := v.Embed(&model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := v.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding for input index 1") {
 		t.Errorf("expected missing-slot error, got %v", err)
 	}
 }
 
 func TestVoyageRerankHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		// Voyage's request key is top_k (not top_n).
 		if body["top_k"] != float64(3) {
@@ -291,7 +302,7 @@ func TestVoyageRerankHappyPath(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "rerank-2"
-	resp, err := v.Rerank(&model, "x", []string{"a", "b", "c"},
+	resp, err := v.Rerank(ctx, &model, "x", []string{"a", "b", "c"},
 		&APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 3}, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
@@ -308,6 +319,7 @@ func TestVoyageRerankHappyPath(t *testing.T) {
 }
 
 func TestVoyageRerankTopKDefaultsToLenDocuments(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/rerank", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["top_k"] != float64(4) {
 			t.Errorf("top_k=%v want 4 (len(documents))", body["top_k"])
@@ -319,7 +331,7 @@ func TestVoyageRerankTopKDefaultsToLenDocuments(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "rerank-2"
-	_, err := v.Rerank(&model, "x", []string{"a", "b", "c", "d"},
+	_, err := v.Rerank(ctx, &model, "x", []string{"a", "b", "c", "d"},
 		&APIConfig{ApiKey: &apiKey}, &RerankConfig{}, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
@@ -327,10 +339,11 @@ func TestVoyageRerankTopKDefaultsToLenDocuments(t *testing.T) {
 }
 
 func TestVoyageRerankEmptyDocuments(t *testing.T) {
+	ctx := t.Context()
 	v := newVoyageForTest("http://unused")
 	apiKey := "test-key"
 	model := "rerank-2"
-	resp, err := v.Rerank(&model, "x", nil,
+	resp, err := v.Rerank(ctx, &model, "x", nil,
 		&APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 0}, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
@@ -341,6 +354,7 @@ func TestVoyageRerankEmptyDocuments(t *testing.T) {
 }
 
 func TestVoyageRerankRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newVoyageServer(t, "/v1/rerank", func(t *testing.T, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"data": []map[string]interface{}{
@@ -353,7 +367,7 @@ func TestVoyageRerankRejectsOutOfRangeIndex(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "rerank-2"
-	_, err := v.Rerank(&model, "x", []string{"a", "b"},
+	_, err := v.Rerank(ctx, &model, "x", []string{"a", "b"},
 		&APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
@@ -361,6 +375,7 @@ func TestVoyageRerankRejectsOutOfRangeIndex(t *testing.T) {
 }
 
 func TestVoyageRerankRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	// A duplicate index would silently overwrite an earlier slot, which
 	// is the same failure mode Embed already guards against. Make sure
 	// Rerank fails loudly too.
@@ -377,7 +392,7 @@ func TestVoyageRerankRejectsDuplicateIndex(t *testing.T) {
 	v := newVoyageForTest(srv.URL)
 	apiKey := "test-key"
 	model := "rerank-2"
-	_, err := v.Rerank(&model, "x", []string{"a", "b"},
+	_, err := v.Rerank(ctx, &model, "x", []string{"a", "b"},
 		&APIConfig{ApiKey: &apiKey}, &RerankConfig{TopN: 2}, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate rerank index 0") {
 		t.Errorf("expected duplicate-index error, got %v", err)
@@ -389,6 +404,7 @@ func TestVoyageRerankRejectsDuplicateIndex(t *testing.T) {
 // (e.g. `.../v1//embeddings`). Rerank already trims, so Embed must
 // trim too; CodeRabbit flagged the inconsistency.
 func TestVoyageEmbedTrimsTrailingSlashInBaseURL(t *testing.T) {
+	ctx := t.Context()
 	var sawPath string
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		sawPath = r.URL.Path
@@ -404,7 +420,7 @@ func TestVoyageEmbedTrimsTrailingSlashInBaseURL(t *testing.T) {
 	)
 	apiKey := "test-key"
 	model := "voyage-3.5"
-	if _, err := v.Embed(&model, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
+	if _, err := v.Embed(ctx, &model, []string{"x"}, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 	if sawPath != "/v1/embeddings" {
diff --git a/internal/entity/models/xai.go b/internal/entity/models/xai.go
index 303e7e5d5c..a2c17417ec 100644
--- a/internal/entity/models/xai.go
+++ b/internal/entity/models/xai.go
@@ -63,7 +63,7 @@ func (x *XAIModel) Name() string {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response
-func (x *XAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (x *XAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -119,7 +119,7 @@ func (x *XAIModel) ChatWithMessages(modelName string, messages []Message, apiCon
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -190,7 +190,7 @@ func (x *XAIModel) ChatWithMessages(modelName string, messages []Message, apiCon
 }
 
 // ChatStreamlyWithSender sends messages and streams the response
-func (x *XAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -249,7 +249,7 @@ func (x *XAIModel) ChatStreamlyWithSender(modelName string, messages []Message,
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -326,12 +326,12 @@ func (x *XAIModel) ChatStreamlyWithSender(modelName string, messages []Message,
 
 // Embed embeds a list of texts into embeddings. xAI does not expose a
 // public embedding API yet, so this is left unimplemented.
-func (x *XAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (x *XAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
 // ListModels returns the list of model ids visible to the API key.
-func (x *XAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (x *XAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -347,7 +347,7 @@ func (x *XAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error)
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), modelsSuffix)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -387,13 +387,13 @@ func (x *XAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error)
 }
 
 // Balance is not exposed by the xAI API, so this returns "no such method".
-func (x *XAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (x *XAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method")
 }
 
 // CheckConnection runs a lightweight ListModels call to verify the API key.
-func (x *XAIModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := x.ListModels(apiConfig)
+func (x *XAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := x.ListModels(ctx, apiConfig)
 	if err != nil {
 		return err
 	}
@@ -401,12 +401,12 @@ func (x *XAIModel) CheckConnection(apiConfig *APIConfig) error {
 }
 
 // Rerank calculates similarity scores between query and documents
-func (x *XAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (x *XAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, Rerank not implemented", x.Name())
 }
 
 // TranscribeAudio transcribe audio
-func (x *XAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (x *XAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -535,12 +535,12 @@ func (x *XAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *A
 	return &ASRResponse{Text: result.Text}, nil
 }
 
-func (x *XAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
 // AudioSpeech convert text to audio
-func (x *XAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (x *XAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -602,24 +602,24 @@ func (x *XAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfi
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (x *XAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
 // OCRFile OCR file
-func (x *XAIModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (x *XAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
 // ParseFile parse file
-func (x *XAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (x *XAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (x *XAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (x *XAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
diff --git a/internal/entity/models/xai_test.go b/internal/entity/models/xai_test.go
index 44cc3fde04..25267917a2 100644
--- a/internal/entity/models/xai_test.go
+++ b/internal/entity/models/xai_test.go
@@ -43,6 +43,7 @@ func TestXAIConfigDeclaresModelsSuffix(t *testing.T) {
 }
 
 func TestXAIListModelsHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodGet {
 			t.Errorf("method=%s, want GET", r.Method)
@@ -69,7 +70,7 @@ func TestXAIListModelsHappyPath(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	models, err := newXAIForTest(srv.URL + "/").ListModels(&APIConfig{ApiKey: &apiKey})
+	models, err := newXAIForTest(srv.URL+"/").ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
@@ -79,26 +80,29 @@ func TestXAIListModelsHappyPath(t *testing.T) {
 }
 
 func TestXAIListModelsRequiresAPIKey(t *testing.T) {
-	_, err := newXAIForTest("http://unused").ListModels(&APIConfig{})
+	ctx := t.Context()
+	_, err := newXAIForTest("http://unused").ListModels(ctx, &APIConfig{})
 	if err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Fatalf("expected api key error, got %v", err)
 	}
 }
 
 func TestXAIListModelsRejectsProviderError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, "bad key", http.StatusUnauthorized)
 	}))
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newXAIForTest(srv.URL).ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := newXAIForTest(srv.URL).ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "401") || !strings.Contains(err.Error(), "bad key") {
 		t.Fatalf("expected provider error with status and body, got %v", err)
 	}
 }
 
 func TestXAICheckConnectionDelegatesToListModels(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/models" {
 			t.Errorf("path=%s, want /models", r.URL.Path)
@@ -108,12 +112,13 @@ func TestXAICheckConnectionDelegatesToListModels(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	if err := newXAIForTest(srv.URL).CheckConnection(&APIConfig{ApiKey: &apiKey}); err != nil {
+	if err := newXAIForTest(srv.URL).CheckConnection(ctx, &APIConfig{ApiKey: &apiKey}); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
 
 func TestXAIListModelsRequiresModelsSuffix(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		t.Errorf("ListModels should reject a missing models suffix before sending a request")
 	}))
@@ -122,7 +127,7 @@ func TestXAIListModelsRequiresModelsSuffix(t *testing.T) {
 	apiKey := "test-key"
 	model := NewXAIModel(map[string]string{"default": srv.URL}, URLSuffix{})
 
-	_, err := model.ListModels(&APIConfig{ApiKey: &apiKey})
+	_, err := model.ListModels(ctx, &APIConfig{ApiKey: &apiKey})
 	if err == nil || !strings.Contains(err.Error(), "models URL suffix is not configured") {
 		t.Fatalf("expected missing models suffix error, got %v", err)
 	}
diff --git a/internal/entity/models/xiaomi.go b/internal/entity/models/xiaomi.go
index e522952983..fd00c286c9 100644
--- a/internal/entity/models/xiaomi.go
+++ b/internal/entity/models/xiaomi.go
@@ -53,7 +53,7 @@ func (x *XiaomiModel) Name() string {
 	return "xiaomi"
 }
 
-func (x *XiaomiModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (x *XiaomiModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -139,7 +139,7 @@ func (x *XiaomiModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -224,7 +224,7 @@ func (x *XiaomiModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (x *XiaomiModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XiaomiModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -312,7 +312,7 @@ func (x *XiaomiModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -387,16 +387,16 @@ func (x *XiaomiModel) ChatStreamlyWithSender(modelName string, messages []Messag
 	return nil
 }
 
-func (x *XiaomiModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (x *XiaomiModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (x *XiaomiModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+func (x *XiaomiModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := x.newXiaomiASRRequest(ctx, modelName, file, apiConfig, asrConfig, false)
@@ -430,12 +430,12 @@ func (x *XiaomiModel) TranscribeAudio(modelName *string, file *string, apiConfig
 	return &ASRResponse{Text: result.Choices[0].Message.Content}, nil
 }
 
-func (x *XiaomiModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XiaomiModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := x.newXiaomiASRRequest(ctx, modelName, file, apiConfig, asrConfig, true)
@@ -563,9 +563,6 @@ func (x *XiaomiModel) newXiaomiASRRequest(ctx context.Context, modelName *string
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), strings.TrimPrefix(x.baseModel.URLSuffix.Chat, "/"))
 
-	if ctx == nil {
-		ctx = context.Background()
-	}
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return nil, fmt.Errorf("failed to create request: %w", err)
@@ -641,8 +638,8 @@ func readXiaomiASRStream(body io.Reader, sender func(*string, *string) error) er
 	return sender(&done, nil)
 }
 
-func (x *XiaomiModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+func (x *XiaomiModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := x.newXiaomiTTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, false)
@@ -667,12 +664,12 @@ func (x *XiaomiModel) AudioSpeech(modelName *string, audioContent *string, apiCo
 	return decodeXiaomiTTSResponse(body)
 }
 
-func (x *XiaomiModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XiaomiModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if sender == nil {
 		return fmt.Errorf("sender is required")
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := x.newXiaomiTTSRequest(ctx, modelName, audioContent, apiConfig, ttsConfig, true)
@@ -760,9 +757,6 @@ func (x *XiaomiModel) newXiaomiTTSRequest(ctx context.Context, modelName *string
 	}
 	url := fmt.Sprintf("%s/%s", strings.TrimSuffix(baseURL, "/"), strings.TrimPrefix(x.baseModel.URLSuffix.Chat, "/"))
 
-	if ctx == nil {
-		ctx = context.Background()
-	}
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
 	if err != nil {
 		return nil, fmt.Errorf("failed to create request: %w", err)
@@ -840,23 +834,23 @@ func decodeXiaomiAudioData(data string) ([]byte, error) {
 	return base64.StdEncoding.DecodeString(data)
 }
 
-func (x *XiaomiModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (x *XiaomiModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (x *XiaomiModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (x *XiaomiModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (x *XiaomiModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) CheckConnection(apiConfig *APIConfig) error {
+func (x *XiaomiModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -864,10 +858,10 @@ func (x *XiaomiModel) CheckConnection(apiConfig *APIConfig) error {
 	return err
 }
 
-func (x *XiaomiModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (x *XiaomiModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
 
-func (x *XiaomiModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (x *XiaomiModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("no such method %s", x.Name())
 }
diff --git a/internal/entity/models/xiaomi_test.go b/internal/entity/models/xiaomi_test.go
index 03b1caca01..31637cd684 100644
--- a/internal/entity/models/xiaomi_test.go
+++ b/internal/entity/models/xiaomi_test.go
@@ -111,7 +111,8 @@ func TestXiaomiChatHappyPath(t *testing.T) {
 	apiKey := "test-key"
 	maxTokens := 1024
 	thinking := false
-	resp, err := newXiaomiForTest(srv.URL).ChatWithMessages(
+	ctx := t.Context()
+	resp, err := newXiaomiForTest(srv.URL).ChatWithMessages(ctx,
 		"mimo-v2.5-pro",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -129,6 +130,7 @@ func TestXiaomiChatHappyPath(t *testing.T) {
 }
 
 func TestXiaomiUsesEmptyRegionBaseURLOverride(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -145,7 +147,7 @@ func TestXiaomiUsesEmptyRegionBaseURLOverride(t *testing.T) {
 		map[string]string{"default": srv.URL},
 		URLSuffix{Chat: "v1/chat/completions"},
 	)
-	resp, err := m.ChatWithMessages("mimo-v2.5-pro", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := m.ChatWithMessages(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("ChatWithMessages: %v", err)
 	}
@@ -155,6 +157,7 @@ func TestXiaomiUsesEmptyRegionBaseURLOverride(t *testing.T) {
 }
 
 func TestXiaomiAPIConfigBaseURLOverridesRegionMap(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/override/chat", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -172,7 +175,7 @@ func TestXiaomiAPIConfigBaseURLOverridesRegionMap(t *testing.T) {
 		map[string]string{"default": "http://unused"},
 		URLSuffix{Chat: "override/chat"},
 	)
-	resp, err := m.ChatWithMessages("mimo-v2.5-pro", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey, BaseURL: &baseURL}, nil, nil)
+	resp, err := m.ChatWithMessages(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "ping"}}, &APIConfig{ApiKey: &apiKey, BaseURL: &baseURL}, nil, nil)
 	if err != nil {
 		t.Fatalf("ChatWithMessages: %v", err)
 	}
@@ -182,6 +185,7 @@ func TestXiaomiAPIConfigBaseURLOverridesRegionMap(t *testing.T) {
 }
 
 func TestXiaomiChatExtractsReasoningContent(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"choices": []map[string]interface{}{{
@@ -195,7 +199,7 @@ func TestXiaomiChatExtractsReasoningContent(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	resp, err := newXiaomiForTest(srv.URL).ChatWithMessages("mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := newXiaomiForTest(srv.URL).ChatWithMessages(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("ChatWithMessages: %v", err)
 	}
@@ -205,20 +209,22 @@ func TestXiaomiChatExtractsReasoningContent(t *testing.T) {
 }
 
 func TestXiaomiChatRequiresInputs(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	m := newXiaomiForTest("http://unused")
-	if _, err := m.ChatWithMessages("mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
+	if _, err := m.ChatWithMessages(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("api key guard: %v", err)
 	}
-	if _, err := m.ChatWithMessages("", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
+	if _, err := m.ChatWithMessages(ctx, "", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("model guard: %v", err)
 	}
-	if _, err := m.ChatWithMessages("mimo-v2.5-pro", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "messages is empty") {
+	if _, err := m.ChatWithMessages(ctx, "mimo-v2.5-pro", nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err == nil || !strings.Contains(err.Error(), "messages is empty") {
 		t.Errorf("messages guard: %v", err)
 	}
 }
 
 func TestXiaomiChatRejectsHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		w.WriteHeader(http.StatusUnauthorized)
 		_, _ = io.WriteString(w, `{"error":"unauthorized"}`)
@@ -226,13 +232,14 @@ func TestXiaomiChatRejectsHTTPError(t *testing.T) {
 	defer srv.Close()
 
 	apiKey := "test-key"
-	_, err := newXiaomiForTest(srv.URL).ChatWithMessages("mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newXiaomiForTest(srv.URL).ChatWithMessages(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "401") {
 		t.Errorf("expected 401 propagated, got %v", err)
 	}
 }
 
 func TestXiaomiStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, body map[string]interface{}, w http.ResponseWriter) {
 		if body["stream"] != true {
 			t.Errorf("stream=%v want true", body["stream"])
@@ -250,6 +257,7 @@ func TestXiaomiStreamHappyPath(t *testing.T) {
 	var content, reasoning []string
 	var sawDone bool
 	err := newXiaomiForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"mimo-v2.5-pro",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -284,6 +292,7 @@ func TestXiaomiStreamHappyPath(t *testing.T) {
 }
 
 func TestXiaomiStreamHandlesCRLFFrames(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w,
@@ -296,6 +305,7 @@ func TestXiaomiStreamHandlesCRLFFrames(t *testing.T) {
 	apiKey := "test-key"
 	var content []string
 	err := newXiaomiForTest(srv.URL).ChatStreamlyWithSender(
+		ctx,
 		"mimo-v2.5-pro",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{ApiKey: &apiKey},
@@ -317,6 +327,7 @@ func TestXiaomiStreamHandlesCRLFFrames(t *testing.T) {
 }
 
 func TestXiaomiStreamRejectsMalformedFrame(t *testing.T) {
+	ctx := t.Context()
 	srv := newXiaomiServer(t, "/v1/chat/completions", func(t *testing.T, _ *http.Request, _ map[string]interface{}, w http.ResponseWriter) {
 		w.Header().Set("Content-Type", "text/event-stream")
 		_, _ = io.WriteString(w, "data: {bad json}\n\n")
@@ -325,37 +336,38 @@ func TestXiaomiStreamRejectsMalformedFrame(t *testing.T) {
 
 	apiKey := "test-key"
 	// Malformed SSE frames are silently skipped; the stream completes and sends [DONE].
-	err := newXiaomiForTest(srv.URL).ChatStreamlyWithSender("mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })
+	err := newXiaomiForTest(srv.URL).ChatStreamlyWithSender(ctx, "mimo-v2.5-pro", []Message{{Role: "user", Content: "x"}}, &APIConfig{ApiKey: &apiKey}, nil, nil, func(*string, *string) error { return nil })
 	if err != nil {
 		t.Errorf("expected no error on malformed frame, got %v", err)
 	}
 }
 
 func TestXiaomiUnsupportedMethods(t *testing.T) {
+	ctx := t.Context()
 	m := newXiaomiForTest("http://unused")
 	model := "mimo-v2.5-pro"
 	apiKey := "test-key"
 	cfg := &APIConfig{ApiKey: &apiKey}
 
-	if _, err := m.Embed(&model, []string{"x"}, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Embed(ctx, &model, []string{"x"}, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Embed: %v", err)
 	}
-	if _, err := m.Rerank(&model, "q", []string{"d"}, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.Rerank(ctx, &model, "q", []string{"d"}, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Rerank: %v", err)
 	}
 	// CheckConnection IS implemented — verifies API config and base URL are reachable.
-	if err := m.CheckConnection(cfg); err != nil {
+	if err := m.CheckConnection(ctx, cfg); err != nil {
 		t.Errorf("CheckConnection: %v", err)
 	}
 	// TranscribeAudio IS implemented; with nil file it returns input validation error.
-	if _, err := m.TranscribeAudio(&model, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "file is missing") {
+	if _, err := m.TranscribeAudio(ctx, &model, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "file is missing") {
 		t.Errorf("TranscribeAudio: %v", err)
 	}
 	// AudioSpeech IS implemented; with nil content it returns input validation error.
-	if _, err := m.AudioSpeech(&model, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "audio content is empty") {
+	if _, err := m.AudioSpeech(ctx, &model, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "audio content is empty") {
 		t.Errorf("AudioSpeech: %v", err)
 	}
-	if _, err := m.OCRFile(&model, nil, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := m.OCRFile(ctx, &model, nil, nil, cfg, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: %v", err)
 	}
 }
diff --git a/internal/entity/models/xinference.go b/internal/entity/models/xinference.go
index 79c256669b..3672545ac4 100644
--- a/internal/entity/models/xinference.go
+++ b/internal/entity/models/xinference.go
@@ -143,7 +143,7 @@ func buildXinferenceChatBody(modelName string, messages []Message, stream bool,
 }
 
 // ChatWithMessages sends multiple messages with roles and returns the response.
-func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (x *XinferenceModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -165,7 +165,7 @@ func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message,
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -213,7 +213,7 @@ func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message,
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender.
-func (x *XinferenceModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XinferenceModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -342,7 +342,7 @@ type xinferenceEmbeddingResponse struct {
 }
 
 // Embed POSTs the input texts to the tenant's Xinference
-func (x *XinferenceModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (x *XinferenceModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -377,7 +377,7 @@ func (x *XinferenceModel) Embed(modelName *string, texts []string, apiConfig *AP
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -450,7 +450,7 @@ type xinferenceRerankResponse struct {
 // in the API's ranking order. Caller may sort by Index to recover
 // original input order. Xinference rerank models are launched with
 // --model-type rerank and exposed under the OpenAI-compatible base URL.
-func (x *XinferenceModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (x *XinferenceModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -489,7 +489,7 @@ func (x *XinferenceModel) Rerank(modelName *string, query string, documents []st
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -541,7 +541,7 @@ func (x *XinferenceModel) Rerank(modelName *string, query string, documents []st
 	return &rerankResponse, nil
 }
 
-func (x *XinferenceModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (x *XinferenceModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -650,11 +650,11 @@ func (x *XinferenceModel) TranscribeAudio(modelName *string, file *string, apiCo
 	}, nil
 }
 
-func (x *XinferenceModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XinferenceModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XinferenceModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (x *XinferenceModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -716,21 +716,21 @@ func (x *XinferenceModel) AudioSpeech(modelName *string, audioContent *string, a
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (x *XinferenceModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XinferenceModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XinferenceModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (x *XinferenceModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XinferenceModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (x *XinferenceModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
 // ListModels returns the model IDs exposed by Xinference's OpenAI-compatible
 // /v1/models endpoint.
-func (x *XinferenceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (x *XinferenceModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -742,7 +742,7 @@ func (x *XinferenceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	baseURL = normalizeXinferenceBaseURL(baseURL)
 	url := fmt.Sprintf("%s/%s", baseURL, x.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -777,19 +777,19 @@ func (x *XinferenceModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse,
 	return ParseListModel(ModelList{Models: result.Data}), nil
 }
 
-func (x *XinferenceModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (x *XinferenceModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XinferenceModel) CheckConnection(apiConfig *APIConfig) error {
-	_, err := x.ListModels(apiConfig)
+func (x *XinferenceModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
+	_, err := x.ListModels(ctx, apiConfig)
 	return err
 }
 
-func (x *XinferenceModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (x *XinferenceModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XinferenceModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (x *XinferenceModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
diff --git a/internal/entity/models/xinference_test.go b/internal/entity/models/xinference_test.go
index 4ff820c900..872c9564bc 100644
--- a/internal/entity/models/xinference_test.go
+++ b/internal/entity/models/xinference_test.go
@@ -66,6 +66,7 @@ func TestXinferenceFactoryRoute(t *testing.T) {
 }
 
 func TestXinferenceChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T) {
+	ctx := t.Context()
 	var seen map[string]interface{}
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/chat/completions" {
@@ -90,7 +91,7 @@ func TestXinferenceChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T)
 	x := newXinferenceForTest(srv.URL)
 	maxTokens := 32
 	temp := 0.2
-	resp, err := x.ChatWithMessages("qwen2.5-instruct",
+	resp, err := x.ChatWithMessages(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "ping"}},
 		&APIConfig{},
 		&ChatConfig{MaxTokens: &maxTokens, Temperature: &temp}, nil)
@@ -112,6 +113,7 @@ func TestXinferenceChatHappyPathNormalizesBaseURLAndOmitsEmptyAuth(t *testing.T)
 }
 
 func TestXinferenceChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if got := r.Header.Get("Authorization"); got != "Bearer sk-test" {
 			t.Errorf("Authorization=%q, want Bearer sk-test", got)
@@ -122,7 +124,7 @@ func TestXinferenceChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL + "/v1")
 	key := "sk-test"
-	_, err := x.ChatWithMessages("qwen2.5-instruct",
+	_, err := x.ChatWithMessages(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{ApiKey: &key}, nil, nil)
 	if err != nil {
@@ -131,6 +133,7 @@ func TestXinferenceChatSendsAuthHeaderWhenKeyProvided(t *testing.T) {
 }
 
 func TestXinferenceChatExtractsReasoningFields(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
 		_, _ = io.WriteString(w, `{"choices":[{"message":{
 			"content":"12",
@@ -140,7 +143,7 @@ func TestXinferenceChatExtractsReasoningFields(t *testing.T) {
 	defer srv.Close()
 
 	x := newXinferenceForTest(srv.URL)
-	resp, err := x.ChatWithMessages("qwen3",
+	resp, err := x.ChatWithMessages(ctx, "qwen3",
 		[]Message{{Role: "user", Content: "15% of 80?"}},
 		&APIConfig{}, nil, nil)
 	if err != nil {
@@ -152,6 +155,7 @@ func TestXinferenceChatExtractsReasoningFields(t *testing.T) {
 }
 
 func TestXinferenceStreamHappyPath(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/chat/completions" {
 			t.Errorf("path=%s", r.URL.Path)
@@ -176,7 +180,7 @@ func TestXinferenceStreamHappyPath(t *testing.T) {
 	var content []string
 	var reasoning []string
 	var sawDone bool
-	err := x.ChatStreamlyWithSender("qwen2.5-instruct",
+	err := x.ChatStreamlyWithSender(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "hi"}},
 		&APIConfig{}, nil, nil,
 		func(c *string, r *string) error {
@@ -206,9 +210,10 @@ func TestXinferenceStreamHappyPath(t *testing.T) {
 }
 
 func TestXinferenceStreamRejectsFalseStreamConfig(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
 	stream := false
-	err := x.ChatStreamlyWithSender("qwen2.5-instruct",
+	err := x.ChatStreamlyWithSender(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{},
 		&ChatConfig{Stream: &stream},
@@ -220,6 +225,7 @@ func TestXinferenceStreamRejectsFalseStreamConfig(t *testing.T) {
 }
 
 func TestXinferenceStreamCancelsOnIdle(t *testing.T) {
+	ctx := t.Context()
 	withXinferenceIdleTimeout(t, 200*time.Millisecond)
 
 	hold := make(chan struct{})
@@ -239,7 +245,7 @@ func TestXinferenceStreamCancelsOnIdle(t *testing.T) {
 	t.Cleanup(func() { close(hold) })
 
 	x := newXinferenceForTest(srv.URL)
-	err := x.ChatStreamlyWithSender("qwen2.5-instruct",
+	err := x.ChatStreamlyWithSender(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil,
 		nil,
@@ -250,6 +256,7 @@ func TestXinferenceStreamCancelsOnIdle(t *testing.T) {
 }
 
 func TestXinferenceListModelsAndCheckConnection(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/v1/models" {
 			t.Errorf("path=%s, want /v1/models", r.URL.Path)
@@ -264,14 +271,14 @@ func TestXinferenceListModelsAndCheckConnection(t *testing.T) {
 	x := newXinferenceForTest(srv.URL)
 	key := "sk-test"
 	apiConfig := &APIConfig{ApiKey: &key}
-	models, err := x.ListModels(apiConfig)
+	models, err := x.ListModels(ctx, apiConfig)
 	if err != nil {
 		t.Fatalf("ListModels: %v", err)
 	}
 	if joinModelNames(models, ",") != "qwen2.5-instruct,custom-chat" {
 		t.Errorf("models=%v", models)
 	}
-	if err := x.CheckConnection(apiConfig); err != nil {
+	if err = x.CheckConnection(ctx, apiConfig); err != nil {
 		t.Fatalf("CheckConnection: %v", err)
 	}
 }
@@ -306,6 +313,7 @@ func newXinferenceEmbedServer(t *testing.T, handler func(t *testing.T, body map[
 }
 
 func TestXinferenceEmbedHappyPathAndOmitsEmptyAuth(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "bge-m3" {
 			t.Errorf("model=%v, want bge-m3", body["model"])
@@ -322,7 +330,7 @@ func TestXinferenceEmbedHappyPathAndOmitsEmptyAuth(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	got, err := x.Embed(&model, []string{"hello", "world"}, &APIConfig{}, nil, nil)
+	got, err := x.Embed(ctx, &model, []string{"hello", "world"}, &APIConfig{}, nil, nil)
 	if err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
@@ -338,6 +346,7 @@ func TestXinferenceEmbedHappyPathAndOmitsEmptyAuth(t *testing.T) {
 }
 
 func TestXinferenceEmbedSendsAuthWhenKeyConfigured(t *testing.T) {
+	ctx := t.Context()
 	gotAuth := ""
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		gotAuth = r.Header.Get("Authorization")
@@ -348,7 +357,7 @@ func TestXinferenceEmbedSendsAuthWhenKeyConfigured(t *testing.T) {
 	x := newXinferenceForTest(srv.URL)
 	key := "sk-test"
 	model := "bge-m3"
-	if _, err := x.Embed(&model, []string{"x"}, &APIConfig{ApiKey: &key}, nil, nil); err != nil {
+	if _, err := x.Embed(ctx, &model, []string{"x"}, &APIConfig{ApiKey: &key}, nil, nil); err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 	if gotAuth != "Bearer sk-test" {
@@ -357,6 +366,7 @@ func TestXinferenceEmbedSendsAuthWhenKeyConfigured(t *testing.T) {
 }
 
 func TestXinferenceEmbedNormalizesBaseURLWithV1Suffix(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_, _ = io.WriteString(w, `{"data":[{"index":0,"embedding":[0.1]}]}`)
 	})
@@ -367,12 +377,13 @@ func TestXinferenceEmbedNormalizesBaseURLWithV1Suffix(t *testing.T) {
 		URLSuffix{Chat: "v1/chat/completions", Embedding: "v1/embeddings", Models: "v1/models"},
 	)
 	model := "bge-m3"
-	if _, err := x.Embed(&model, []string{"x"}, &APIConfig{}, nil, nil); err != nil {
+	if _, err := x.Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil); err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 }
 
 func TestXinferenceEmbedForwardsDimension(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["dimensions"] != float64(384) {
 			t.Errorf("dimensions=%v, want 384", body["dimensions"])
@@ -383,12 +394,13 @@ func TestXinferenceEmbedForwardsDimension(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	if _, err := x.Embed(&model, []string{"x"}, &APIConfig{}, &EmbeddingConfig{Dimension: 384}, nil); err != nil {
+	if _, err := x.Embed(ctx, &model, []string{"x"}, &APIConfig{}, &EmbeddingConfig{Dimension: 384}, nil); err != nil {
 		t.Fatalf("Embed: %v", err)
 	}
 }
 
 func TestXinferenceEmbedRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_, _ = io.WriteString(w, `{"data":[{"index":0,"embedding":[0.1]},{"index":0,"embedding":[0.2]}]}`)
 	})
@@ -396,13 +408,14 @@ func TestXinferenceEmbedRejectsDuplicateIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	_, err := x.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate embedding index") {
 		t.Errorf("expected duplicate-index error, got %v", err)
 	}
 }
 
 func TestXinferenceEmbedRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_, _ = io.WriteString(w, `{"data":[{"index":5,"embedding":[0.1]}]}`)
 	})
@@ -410,13 +423,14 @@ func TestXinferenceEmbedRejectsOutOfRangeIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	_, err := x.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("expected out-of-range error, got %v", err)
 	}
 }
 
 func TestXinferenceEmbedRejectsMissingIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceEmbedServer(t, func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		// Two inputs requested but only one returned — index 1 is missing.
 		_, _ = io.WriteString(w, `{"data":[{"index":0,"embedding":[0.1]}]}`)
@@ -425,16 +439,17 @@ func TestXinferenceEmbedRejectsMissingIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	_, err := x.Embed(&model, []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, &model, []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "missing embedding") {
 		t.Errorf("expected missing-embedding error, got %v", err)
 	}
 }
 
 func TestXinferenceEmbedEmptyTextsShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
 	model := "bge-m3"
-	got, err := x.Embed(&model, nil, &APIConfig{}, nil, nil)
+	got, err := x.Embed(ctx, &model, nil, &APIConfig{}, nil, nil)
 	if err != nil {
 		t.Fatalf("expected nil error for empty inputs, got %v", err)
 	}
@@ -444,14 +459,16 @@ func TestXinferenceEmbedEmptyTextsShortCircuits(t *testing.T) {
 }
 
 func TestXinferenceEmbedRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
-	_, err := x.Embed(nil, []string{"x"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, nil, []string{"x"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("expected model-name error, got %v", err)
 	}
 }
 
 func TestXinferenceEmbedSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusInternalServerError)
 		_, _ = io.WriteString(w, `{"error":"model not loaded"}`)
@@ -460,27 +477,29 @@ func TestXinferenceEmbedSurfacesHTTPError(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-m3"
-	_, err := x.Embed(&model, []string{"x"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "Xinference embeddings API error") {
 		t.Errorf("expected API error, got %v", err)
 	}
 }
 
 func TestXinferenceEmbedRejectsMissingEmbeddingSuffix(t *testing.T) {
+	ctx := t.Context()
 	x := NewXinferenceModel(
 		map[string]string{"default": "http://unused"},
 		URLSuffix{Chat: "v1/chat/completions"}, // no Embedding suffix
 	)
 	model := "bge-m3"
-	_, err := x.Embed(&model, []string{"x"}, &APIConfig{}, nil, nil)
+	_, err := x.Embed(ctx, &model, []string{"x"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no embedding URL suffix configured") {
 		t.Errorf("expected missing-suffix error, got %v", err)
 	}
 }
 
 func TestXinferenceMissingBaseURLFailsClearly(t *testing.T) {
+	ctx := t.Context()
 	x := NewXinferenceModel(map[string]string{}, URLSuffix{Chat: "v1/chat/completions"})
-	_, err := x.ChatWithMessages("qwen2.5-instruct",
+	_, err := x.ChatWithMessages(ctx, "qwen2.5-instruct",
 		[]Message{{Role: "user", Content: "x"}},
 		&APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "base URL") {
@@ -489,29 +508,30 @@ func TestXinferenceMissingBaseURLFailsClearly(t *testing.T) {
 }
 
 func TestXinferenceUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
 	model := "qwen2.5-instruct"
 
-	if _, err := x.Balance(&APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := x.Balance(ctx, &APIConfig{}); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("Balance: expected no such method, got %v", err)
 	}
 	// TranscribeAudio IS implemented; it validates inputs after APIConfigCheck.
 	// The test passes nil file, which should yield an input-validation error,
 	// not an api-key error.
-	if _, err := x.TranscribeAudio(&model, nil, &APIConfig{}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
+	if _, err := x.TranscribeAudio(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("TranscribeAudio: expected input validation error (file missing), got %v", err)
 	}
-	if err := x.TranscribeAudioWithSender(&model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := x.TranscribeAudioWithSender(ctx, &model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("TranscribeAudioWithSender: expected no such method, got %v", err)
 	}
 	// AudioSpeech IS implemented; it validates inputs after APIConfigCheck.
-	if _, err := x.AudioSpeech(&model, nil, &APIConfig{}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
+	if _, err := x.AudioSpeech(ctx, &model, nil, &APIConfig{}, nil, nil); err == nil || strings.Contains(err.Error(), "api key is required") {
 		t.Errorf("AudioSpeech: expected input validation error (text missing), got %v", err)
 	}
-	if err := x.AudioSpeechWithSender(&model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if err := x.AudioSpeechWithSender(ctx, &model, nil, &APIConfig{}, nil, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("AudioSpeechWithSender: expected no such method, got %v", err)
 	}
-	if _, err := x.OCRFile(&model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
+	if _, err := x.OCRFile(ctx, &model, nil, nil, &APIConfig{}, nil, nil); err == nil || !strings.Contains(err.Error(), "no such method") {
 		t.Errorf("OCRFile: expected no such method, got %v", err)
 	}
 }
@@ -546,6 +566,7 @@ func newXinferenceRerankServer(t *testing.T, expectedAuth string, handler func(t
 }
 
 func TestXinferenceRerankHappyPathReordersByIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceRerankServer(t, "", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if body["model"] != "bge-reranker-v2-m3" {
 			t.Errorf("model=%v", body["model"])
@@ -568,7 +589,7 @@ func TestXinferenceRerankHappyPathReordersByIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-reranker-v2-m3"
-	resp, err := x.Rerank(&model, "capital of France",
+	resp, err := x.Rerank(ctx, &model, "capital of France",
 		[]string{"Paris is the capital of France.", "Eiffel Tower.", "Berlin is the capital of Germany."},
 		&APIConfig{}, nil, nil,
 	)
@@ -587,6 +608,7 @@ func TestXinferenceRerankHappyPathReordersByIndex(t *testing.T) {
 }
 
 func TestXinferenceRerankNormalizesV1BaseURL(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceRerankServer(t, "Bearer test-key", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{"results": []map[string]interface{}{}})
 	})
@@ -598,13 +620,14 @@ func TestXinferenceRerankNormalizesV1BaseURL(t *testing.T) {
 	)
 	apiKey := "test-key"
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
 }
 
 func TestXinferenceRerankRespectsTopNConfig(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceRerankServer(t, "", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		if got := body["top_n"].(float64); got != 2 {
 			t.Errorf("top_n=%v want 2", got)
@@ -615,16 +638,17 @@ func TestXinferenceRerankRespectsTopNConfig(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a", "b", "c", "d"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a", "b", "c", "d"}, &APIConfig{}, &RerankConfig{TopN: 2}, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
 }
 
 func TestXinferenceRerankEmptyDocumentsShortCircuits(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
 	model := "bge-reranker-v2-m3"
-	resp, err := x.Rerank(&model, "q", nil, &APIConfig{}, nil, nil)
+	resp, err := x.Rerank(ctx, &model, "q", nil, &APIConfig{}, nil, nil)
 	if err != nil {
 		t.Fatalf("Rerank: %v", err)
 	}
@@ -634,14 +658,16 @@ func TestXinferenceRerankEmptyDocumentsShortCircuits(t *testing.T) {
 }
 
 func TestXinferenceRerankRequiresModelName(t *testing.T) {
+	ctx := t.Context()
 	x := newXinferenceForTest("http://unused")
-	_, err := x.Rerank(nil, "q", []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := x.Rerank(ctx, nil, "q", []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "model name is required") {
 		t.Errorf("err=%v", err)
 	}
 }
 
 func TestXinferenceRerankRejectsOutOfRangeIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceRerankServer(t, "", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"results": []map[string]interface{}{{"index": 5, "relevance_score": 0.1}},
@@ -651,13 +677,14 @@ func TestXinferenceRerankRejectsOutOfRangeIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "out of range") {
 		t.Errorf("err=%v", err)
 	}
 }
 
 func TestXinferenceRerankRejectsDuplicateIndex(t *testing.T) {
+	ctx := t.Context()
 	srv := newXinferenceRerankServer(t, "", func(t *testing.T, body map[string]interface{}, w http.ResponseWriter) {
 		_ = json.NewEncoder(w).Encode(map[string]interface{}{
 			"results": []map[string]interface{}{
@@ -670,13 +697,14 @@ func TestXinferenceRerankRejectsDuplicateIndex(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a", "b"}, &APIConfig{}, nil, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a", "b"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "duplicate") {
 		t.Errorf("err=%v", err)
 	}
 }
 
 func TestXinferenceRerankSurfacesHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusInternalServerError)
 		_, _ = w.Write([]byte(`{"error":"model not loaded"}`))
@@ -685,19 +713,20 @@ func TestXinferenceRerankSurfacesHTTPError(t *testing.T) {
 
 	x := newXinferenceForTest(srv.URL)
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "Xinference rerank API error") {
 		t.Errorf("err=%v", err)
 	}
 }
 
 func TestXinferenceRerankRejectsMissingRerankSuffix(t *testing.T) {
+	ctx := t.Context()
 	x := NewXinferenceModel(
 		map[string]string{"default": "http://unused"},
 		URLSuffix{Chat: "v1/chat/completions"},
 	)
 	model := "bge-reranker-v2-m3"
-	_, err := x.Rerank(&model, "q", []string{"a"}, &APIConfig{}, nil, nil)
+	_, err := x.Rerank(ctx, &model, "q", []string{"a"}, &APIConfig{}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "no rerank URL suffix configured") {
 		t.Errorf("err=%v", err)
 	}
diff --git a/internal/entity/models/xunfei.go b/internal/entity/models/xunfei.go
index 80bbe35d78..b7a6104c35 100644
--- a/internal/entity/models/xunfei.go
+++ b/internal/entity/models/xunfei.go
@@ -48,7 +48,7 @@ func (x *XunFeiModel) Name() string {
 	return "XunFei Spark"
 }
 
-func (x *XunFeiModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (x *XunFeiModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -113,7 +113,7 @@ func (x *XunFeiModel) ChatWithMessages(modelName string, messages []Message, api
 		return nil, fmt.Errorf("failed to marshal request body: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -180,7 +180,7 @@ func (x *XunFeiModel) ChatWithMessages(modelName string, messages []Message, api
 	return chatResponse, nil
 }
 
-func (x *XunFeiModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XunFeiModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, modelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -250,7 +250,7 @@ func (x *XunFeiModel) ChatStreamlyWithSender(modelName string, messages []Messag
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -321,39 +321,39 @@ func (x *XunFeiModel) ChatStreamlyWithSender(modelName string, messages []Messag
 	return nil
 }
 
-func (x *XunFeiModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (x *XunFeiModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (x *XunFeiModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (x *XunFeiModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XunFeiModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (x *XunFeiModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (x *XunFeiModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (x *XunFeiModel) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (x *XunFeiModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (x *XunFeiModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -372,7 +372,7 @@ func (x *XunFeiModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, bytes.NewBuffer(jsonData))
@@ -411,18 +411,18 @@ func (x *XunFeiModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, err
 	return ParseListModel(modelList), nil
 }
 
-func (x *XunFeiModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (x *XunFeiModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) CheckConnection(apiConfig *APIConfig) error {
+func (x *XunFeiModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	return fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (x *XunFeiModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
 
-func (x *XunFeiModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (x *XunFeiModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", x.Name())
 }
diff --git a/internal/entity/models/xunfei_test.go b/internal/entity/models/xunfei_test.go
index a1ae98ee46..781d719f13 100644
--- a/internal/entity/models/xunfei_test.go
+++ b/internal/entity/models/xunfei_test.go
@@ -3,6 +3,7 @@ package models
 import "testing"
 
 func TestXunFeiUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
+	ctx := t.Context()
 	driver := NewXunFeiModel(map[string]string{"default": "http://unused"}, URLSuffix{}).
 		NewInstance(map[string]string{"default": "http://unused"})
 	modelName := "spark"
@@ -13,48 +14,48 @@ func TestXunFeiUnsupportedMethodsReturnNoSuchMethod(t *testing.T) {
 		call func() error
 	}{
 		{"Embed", func() error {
-			_, err := driver.Embed(&modelName, []string{text}, &APIConfig{}, nil, nil)
+			_, err := driver.Embed(ctx, &modelName, []string{text}, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"Rerank", func() error {
-			_, err := driver.Rerank(&modelName, text, []string{text}, &APIConfig{}, nil, nil)
+			_, err := driver.Rerank(ctx, &modelName, text, []string{text}, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"TranscribeAudio", func() error {
-			_, err := driver.TranscribeAudio(&modelName, &text, &APIConfig{}, nil, nil)
+			_, err := driver.TranscribeAudio(ctx, &modelName, &text, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"TranscribeAudioWithSender", func() error {
-			return driver.TranscribeAudioWithSender(&modelName, &text, &APIConfig{}, nil, nil, nil)
+			return driver.TranscribeAudioWithSender(ctx, &modelName, &text, &APIConfig{}, nil, nil, nil)
 		}},
 		{"AudioSpeech", func() error {
-			_, err := driver.AudioSpeech(&modelName, &text, &APIConfig{}, nil, nil)
+			_, err := driver.AudioSpeech(ctx, &modelName, &text, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"AudioSpeechWithSender", func() error {
-			return driver.AudioSpeechWithSender(&modelName, &text, &APIConfig{}, nil, nil, nil)
+			return driver.AudioSpeechWithSender(ctx, &modelName, &text, &APIConfig{}, nil, nil, nil)
 		}},
 		{"OCRFile", func() error {
-			_, err := driver.OCRFile(&modelName, nil, &text, &APIConfig{}, nil, nil)
+			_, err := driver.OCRFile(ctx, &modelName, nil, &text, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"ParseFile", func() error {
-			_, err := driver.ParseFile(&modelName, nil, &text, &APIConfig{}, nil, nil)
+			_, err := driver.ParseFile(ctx, &modelName, nil, &text, &APIConfig{}, nil, nil)
 			return err
 		}},
 		{"Balance", func() error {
-			_, err := driver.Balance(&APIConfig{})
+			_, err := driver.Balance(ctx, &APIConfig{})
 			return err
 		}},
 		{"CheckConnection", func() error {
-			return driver.CheckConnection(&APIConfig{})
+			return driver.CheckConnection(ctx, &APIConfig{})
 		}},
 		{"ListTasks", func() error {
-			_, err := driver.ListTasks(&APIConfig{})
+			_, err := driver.ListTasks(ctx, &APIConfig{})
 			return err
 		}},
 		{"ShowTask", func() error {
-			_, err := driver.ShowTask("task-id", &APIConfig{})
+			_, err := driver.ShowTask(ctx, "task-id", &APIConfig{})
 			return err
 		}},
 	}
diff --git a/internal/entity/models/zhipu-ai.go b/internal/entity/models/zhipu-ai.go
index a40e7ab2a3..8f9b2ecfbd 100644
--- a/internal/entity/models/zhipu-ai.go
+++ b/internal/entity/models/zhipu-ai.go
@@ -87,7 +87,7 @@ type ZhipuChatResponse struct {
 }
 
 // ChatWithMessages sends multiple messages with roles and returns response
-func (z *ZhipuAIModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
+func (z *ZhipuAIModel) ChatWithMessages(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -171,7 +171,7 @@ func (z *ZhipuAIModel) ChatWithMessages(modelName string, messages []Message, ap
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -247,7 +247,7 @@ func (z *ZhipuAIModel) ChatWithMessages(modelName string, messages []Message, ap
 }
 
 // ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
-func (z *ZhipuAIModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (z *ZhipuAIModel) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -337,7 +337,7 @@ func (z *ZhipuAIModel) ChatStreamlyWithSender(modelName string, messages []Messa
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -507,7 +507,7 @@ type zhipuUsage struct {
 }
 
 // Embed embeddings a list of texts into embeddings
-func (z *ZhipuAIModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
+func (z *ZhipuAIModel) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -539,7 +539,7 @@ func (z *ZhipuAIModel) Embed(modelName *string, texts []string, apiConfig *APICo
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -583,7 +583,7 @@ func (z *ZhipuAIModel) Embed(modelName *string, texts []string, apiConfig *APICo
 	return embeddings, nil
 }
 
-func (z *ZhipuAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, error) {
+func (z *ZhipuAIModel) ListModels(ctx context.Context, apiConfig *APIConfig) ([]ListModelResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -595,7 +595,7 @@ func (z *ZhipuAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 
 	url := fmt.Sprintf("%s/%s", baseURL, z.baseModel.URLSuffix.Models)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -630,11 +630,11 @@ func (z *ZhipuAIModel) ListModels(apiConfig *APIConfig) ([]ListModelResponse, er
 	return ParseListModel(modelList), nil
 }
 
-func (z *ZhipuAIModel) Balance(apiConfig *APIConfig) (map[string]interface{}, error) {
+func (z *ZhipuAIModel) Balance(ctx context.Context, apiConfig *APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("%s, no such method", z.Name())
 }
 
-func (z *ZhipuAIModel) CheckConnection(apiConfig *APIConfig) error {
+func (z *ZhipuAIModel) CheckConnection(ctx context.Context, apiConfig *APIConfig) error {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -646,7 +646,7 @@ func (z *ZhipuAIModel) CheckConnection(apiConfig *APIConfig) error {
 
 	url := fmt.Sprintf("%s/%s", baseURL, z.baseModel.URLSuffix.Files)
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "GET", url, nil)
@@ -710,7 +710,7 @@ type zhipuOCRResponse struct {
 // Rerank calculates similarity scores between query and documents using
 // the ZhipuAI /rerank endpoint (e.g. glm-rerank). The result is one
 // score per input text, in the same order the documents were given.
-func (z *ZhipuAIModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
+func (z *ZhipuAIModel) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -748,7 +748,7 @@ func (z *ZhipuAIModel) Rerank(modelName *string, query string, documents []strin
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), nonStreamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, nonStreamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -792,7 +792,7 @@ func (z *ZhipuAIModel) Rerank(modelName *string, query string, documents []strin
 }
 
 // TranscribeAudio transcribe audio
-func (z *ZhipuAIModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
+func (z *ZhipuAIModel) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -844,7 +844,7 @@ func (z *ZhipuAIModel) TranscribeAudio(modelName *string, file *string, apiConfi
 		return nil, fmt.Errorf("failed to close multipart writer: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, &body)
@@ -919,12 +919,12 @@ func writeZhipuASRField(writer *multipart.Writer, key string, value interface{})
 	}
 }
 
-func (z *ZhipuAIModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (z *ZhipuAIModel) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("%s, no such method", z.Name())
 }
 
 // AudioSpeech convert text to audio
-func (z *ZhipuAIModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
+func (z *ZhipuAIModel) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -939,7 +939,7 @@ func (z *ZhipuAIModel) AudioSpeech(modelName *string, audioContent *string, apiC
 		return nil, fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -966,7 +966,7 @@ func (z *ZhipuAIModel) AudioSpeech(modelName *string, audioContent *string, apiC
 	return &TTSResponse{Audio: body}, nil
 }
 
-func (z *ZhipuAIModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (z *ZhipuAIModel) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return err
 	}
@@ -984,7 +984,7 @@ func (z *ZhipuAIModel) AudioSpeechWithSender(modelName *string, audioContent *st
 		return fmt.Errorf("failed to marshal request: %w", err)
 	}
 
-	ctx, cancel := context.WithTimeout(context.Background(), streamCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, streamCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(jsonData))
@@ -1064,7 +1064,7 @@ func (z *ZhipuAIModel) buildTTSRequest(modelName *string, audioContent *string,
 }
 
 // OCRFile OCR file
-func (z *ZhipuAIModel) OCRFile(modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
+func (z *ZhipuAIModel) OCRFile(ctx context.Context, modelName *string, content []byte, fileURL *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
 	if err := z.baseModel.APIConfigCheck(apiConfig); err != nil {
 		return nil, err
 	}
@@ -1108,7 +1108,7 @@ func (z *ZhipuAIModel) OCRFile(modelName *string, content []byte, fileURL *strin
 	}
 
 	url := fmt.Sprintf("%s/%s", baseURL, strings.TrimPrefix(z.baseModel.URLSuffix.OCR, "/"))
-	ctx, cancel := context.WithTimeout(context.Background(), longOpCallTimeout)
+	ctx, cancel := context.WithTimeout(ctx, longOpCallTimeout)
 	defer cancel()
 
 	req, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewBuffer(jsonData))
@@ -1147,14 +1147,14 @@ func (z *ZhipuAIModel) OCRFile(modelName *string, content []byte, fileURL *strin
 }
 
 // ParseFile parse file
-func (z *ZhipuAIModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
+func (z *ZhipuAIModel) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", z.Name())
 }
 
-func (z *ZhipuAIModel) ListTasks(apiConfig *APIConfig) ([]ListTaskStatus, error) {
+func (z *ZhipuAIModel) ListTasks(ctx context.Context, apiConfig *APIConfig) ([]ListTaskStatus, error) {
 	return nil, fmt.Errorf("%s, no such method", z.Name())
 }
 
-func (z *ZhipuAIModel) ShowTask(taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
+func (z *ZhipuAIModel) ShowTask(ctx context.Context, taskID string, apiConfig *APIConfig) (*TaskResponse, error) {
 	return nil, fmt.Errorf("%s, no such method", z.Name())
 }
diff --git a/internal/entity/models/zhipu-ai_test.go b/internal/entity/models/zhipu-ai_test.go
index 4e4c0fbafd..dad8fa7f46 100644
--- a/internal/entity/models/zhipu-ai_test.go
+++ b/internal/entity/models/zhipu-ai_test.go
@@ -9,6 +9,7 @@ import (
 )
 
 func TestZhipuAIOCRFileSendsLayoutParsingRequest(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-ocr"
 	fileURL := "https://example.com/doc.png"
@@ -52,7 +53,7 @@ func TestZhipuAIOCRFileSendsLayoutParsingRequest(t *testing.T) {
 	defer server.Close()
 
 	model := NewZhipuAIModel(map[string]string{"default": server.URL}, URLSuffix{OCR: "layout_parsing"})
-	resp, err := model.OCRFile(&modelName, nil, &fileURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	resp, err := model.OCRFile(ctx, &modelName, nil, &fileURL, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err != nil {
 		t.Fatalf("OCRFile returned error: %v", err)
 	}
@@ -62,6 +63,7 @@ func TestZhipuAIOCRFileSendsLayoutParsingRequest(t *testing.T) {
 }
 
 func TestZhipuAIOCRFileEncodesContent(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-ocr"
 	content := []byte("sample image bytes")
@@ -81,12 +83,13 @@ func TestZhipuAIOCRFileEncodesContent(t *testing.T) {
 	defer server.Close()
 
 	model := NewZhipuAIModel(map[string]string{"default": server.URL}, URLSuffix{OCR: "layout_parsing"})
-	if _, err := model.OCRFile(&modelName, content, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
+	if _, err := model.OCRFile(ctx, &modelName, content, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
 		t.Fatalf("OCRFile returned error: %v", err)
 	}
 }
 
 func TestZhipuAIOCRFileDetectsPDFContent(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-ocr"
 	content := []byte("%PDF-1.7 sample")
@@ -106,12 +109,13 @@ func TestZhipuAIOCRFileDetectsPDFContent(t *testing.T) {
 	defer server.Close()
 
 	model := NewZhipuAIModel(map[string]string{"default": server.URL}, URLSuffix{OCR: "layout_parsing"})
-	if _, err := model.OCRFile(&modelName, content, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
+	if _, err := model.OCRFile(ctx, &modelName, content, nil, &APIConfig{ApiKey: &apiKey}, nil, nil); err != nil {
 		t.Fatalf("OCRFile returned error: %v", err)
 	}
 }
 
 func TestZhipuAIOCRFileValidation(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-ocr"
 	fileURL := "https://example.com/doc.png"
@@ -149,7 +153,7 @@ func TestZhipuAIOCRFileValidation(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			_, err := model.OCRFile(tt.modelName, nil, tt.fileURL, tt.apiConfig, nil, nil)
+			_, err := model.OCRFile(ctx, tt.modelName, nil, tt.fileURL, tt.apiConfig, nil, nil)
 			if err == nil || !strings.Contains(err.Error(), tt.want) {
 				t.Fatalf("error = %v, want containing %q", err, tt.want)
 			}
diff --git a/internal/entity/models/zhipu_ai_test.go b/internal/entity/models/zhipu_ai_test.go
index 586f3dc116..b0ae7e685a 100644
--- a/internal/entity/models/zhipu_ai_test.go
+++ b/internal/entity/models/zhipu_ai_test.go
@@ -33,6 +33,7 @@ func writeZhipuAITestAudio(t *testing.T) string {
 }
 
 func TestZhipuAITranscribeAudio(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s", r.Method)
@@ -92,6 +93,7 @@ func TestZhipuAITranscribeAudio(t *testing.T) {
 	modelName := "glm-asr-2512"
 	file := writeZhipuAITestAudio(t)
 	resp, err := newZhipuAIForTest(srv.URL).TranscribeAudio(
+		ctx,
 		&modelName,
 		&file,
 		&APIConfig{ApiKey: &apiKey},
@@ -115,6 +117,7 @@ func TestZhipuAITranscribeAudio(t *testing.T) {
 }
 
 func TestZhipuAITranscribeAudioValidation(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-asr-2512"
 	file := "speech.mp3"
@@ -133,7 +136,7 @@ func TestZhipuAITranscribeAudioValidation(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			_, err := newZhipuAIForTest("http://unused").TranscribeAudio(tt.modelName, tt.file, tt.apiConfig, nil, nil)
+			_, err := newZhipuAIForTest("http://unused").TranscribeAudio(ctx, tt.modelName, tt.file, tt.apiConfig, nil, nil)
 			if err == nil || !strings.Contains(err.Error(), tt.want) {
 				t.Fatalf("error=%v, want %q", err, tt.want)
 			}
@@ -142,10 +145,12 @@ func TestZhipuAITranscribeAudioValidation(t *testing.T) {
 }
 
 func TestZhipuAITranscribeAudioRequiresASRSuffix(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-asr-2512"
 	file := writeZhipuAITestAudio(t)
 	_, err := NewZhipuAIModel(map[string]string{"default": "http://unused"}, URLSuffix{}).TranscribeAudio(
+		ctx,
 		&modelName,
 		&file,
 		&APIConfig{ApiKey: &apiKey},
@@ -158,6 +163,7 @@ func TestZhipuAITranscribeAudioRequiresASRSuffix(t *testing.T) {
 }
 
 func TestZhipuAITranscribeAudioHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusBadRequest)
 		_, _ = io.WriteString(w, `{"error":"bad request"}`)
@@ -167,13 +173,14 @@ func TestZhipuAITranscribeAudioHTTPError(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "glm-asr-2512"
 	file := writeZhipuAITestAudio(t)
-	_, err := newZhipuAIForTest(srv.URL).TranscribeAudio(&modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newZhipuAIForTest(srv.URL).TranscribeAudio(ctx, &modelName, &file, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "ZhipuAI ASR API error: 400 Bad Request") {
 		t.Fatalf("error=%v", err)
 	}
 }
 
 func TestZhipuAIAudioSpeech(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s", r.Method)
@@ -221,6 +228,7 @@ func TestZhipuAIAudioSpeech(t *testing.T) {
 	modelName := "glm-tts"
 	content := "hello"
 	resp, err := newZhipuAIForTest(srv.URL).AudioSpeech(
+		ctx,
 		&modelName,
 		&content,
 		&APIConfig{ApiKey: &apiKey},
@@ -242,6 +250,7 @@ func TestZhipuAIAudioSpeech(t *testing.T) {
 }
 
 func TestZhipuAIAudioSpeechWithSender(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		var body map[string]interface{}
 		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
@@ -261,6 +270,7 @@ func TestZhipuAIAudioSpeechWithSender(t *testing.T) {
 	content := "hello"
 	var chunks []string
 	err := newZhipuAIForTest(srv.URL).AudioSpeechWithSender(
+		ctx,
 		&modelName,
 		&content,
 		&APIConfig{ApiKey: &apiKey},
@@ -285,6 +295,7 @@ func TestZhipuAIAudioSpeechWithSender(t *testing.T) {
 }
 
 func TestZhipuAIAudioSpeechValidation(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-tts"
 	content := "hello"
@@ -303,7 +314,7 @@ func TestZhipuAIAudioSpeechValidation(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			_, err := newZhipuAIForTest("http://unused").AudioSpeech(tt.modelName, tt.audioContent, tt.apiConfig, nil, nil)
+			_, err := newZhipuAIForTest("http://unused").AudioSpeech(ctx, tt.modelName, tt.audioContent, tt.apiConfig, nil, nil)
 			if err == nil || !strings.Contains(err.Error(), tt.want) {
 				t.Fatalf("error=%v, want %q", err, tt.want)
 			}
@@ -312,10 +323,12 @@ func TestZhipuAIAudioSpeechValidation(t *testing.T) {
 }
 
 func TestZhipuAIAudioSpeechRequiresTTSSuffix(t *testing.T) {
+	ctx := t.Context()
 	apiKey := "test-key"
 	modelName := "glm-tts"
 	content := "hello"
 	_, err := NewZhipuAIModel(map[string]string{"default": "http://unused"}, URLSuffix{}).AudioSpeech(
+		ctx,
 		&modelName,
 		&content,
 		&APIConfig{ApiKey: &apiKey},
@@ -328,6 +341,7 @@ func TestZhipuAIAudioSpeechRequiresTTSSuffix(t *testing.T) {
 }
 
 func TestZhipuAIAudioSpeechHTTPError(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusBadRequest)
 		_, _ = io.WriteString(w, `{"error":"bad request"}`)
@@ -337,13 +351,14 @@ func TestZhipuAIAudioSpeechHTTPError(t *testing.T) {
 	apiKey := "test-key"
 	modelName := "glm-tts"
 	content := "hello"
-	_, err := newZhipuAIForTest(srv.URL).AudioSpeech(&modelName, &content, &APIConfig{ApiKey: &apiKey}, nil, nil)
+	_, err := newZhipuAIForTest(srv.URL).AudioSpeech(ctx, &modelName, &content, &APIConfig{ApiKey: &apiKey}, nil, nil)
 	if err == nil || !strings.Contains(err.Error(), "ZhipuAI TTS API error: 400 Bad Request") {
 		t.Fatalf("error=%v", err)
 	}
 }
 
 func TestZhipuAIChatStreamlyWithSenderCollectsToolCalls(t *testing.T) {
+	ctx := t.Context()
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.Method != http.MethodPost {
 			t.Errorf("method=%s", r.Method)
@@ -402,6 +417,7 @@ func TestZhipuAIChatStreamlyWithSenderCollectsToolCalls(t *testing.T) {
 		map[string]string{"default": srv.URL},
 		URLSuffix{Chat: "chat/completions"},
 	).ChatStreamlyWithSender(
+		ctx,
 		"glm-4",
 		[]Message{{Role: "user", Content: "what is marigold"}},
 		&APIConfig{ApiKey: &apiKey},
diff --git a/internal/handler/auth.go b/internal/handler/auth.go
index 55d9ebbe4c..0f41486573 100644
--- a/internal/handler/auth.go
+++ b/internal/handler/auth.go
@@ -17,6 +17,7 @@
 package handler
 
 import (
+	"context"
 	"fmt"
 	"net/http"
 	"ragflow/internal/common"
@@ -37,10 +38,10 @@ type AuthHandler struct {
 // so the test suite can swap in a stub without spinning up the
 // full UserService (which requires a live Redis + JWT secret).
 type userTokenResolver interface {
-	GetUserByToken(authorization string) (*entity.User, common.ErrorCode, error)
-	GetUserByAPIToken(token string) (*entity.User, common.ErrorCode, error)
-	GetUserByBetaAPIToken(token string) (*entity.User, common.ErrorCode, error)
-	GetAPITokenByBeta(authorization string) (*entity.APIToken, error)
+	GetUserByToken(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error)
+	GetUserByAPIToken(ctx context.Context, token string) (*entity.User, common.ErrorCode, error)
+	GetUserByBetaAPIToken(ctx context.Context, token string) (*entity.User, common.ErrorCode, error)
+	GetAPITokenByBeta(ctx context.Context, authorization string) (*entity.APIToken, error)
 }
 
 // NewAuthHandler create auth handler
@@ -69,6 +70,7 @@ func NewAuthHandler() *AuthHandler {
 // regular user token must keep working here too.
 func (h *AuthHandler) BetaAuthMiddleware() gin.HandlerFunc {
 	return func(c *gin.Context) {
+		ctx := c.Request.Context()
 		auth := c.GetHeader("Authorization")
 		if auth == "" {
 			if cookie, err := c.Cookie(oauthAuthCookie); err == nil {
@@ -82,13 +84,13 @@ func (h *AuthHandler) BetaAuthMiddleware() gin.HandlerFunc {
 			return
 		}
 		// AUTH_JWT
-		if u, code, err := h.userService.GetUserByToken(auth); err == nil && code == common.CodeSuccess {
+		if u, code, err := h.userService.GetUserByToken(ctx, auth); err == nil && code == common.CodeSuccess {
 			c.Set("user", u)
 			c.Next()
 			return
 		}
 		// Then try a regular API token (non-beta public bot flow).
-		if u, code, err := h.userService.GetUserByAPIToken(auth); err == nil && code == common.CodeSuccess {
+		if u, code, err := h.userService.GetUserByAPIToken(ctx, auth); err == nil && code == common.CodeSuccess {
 			c.Set("user", u)
 			c.Set("auth_via_api_token", true)
 			c.Next()
@@ -101,9 +103,9 @@ func (h *AuthHandler) BetaAuthMiddleware() gin.HandlerFunc {
 		// Mirrors the python
 		// `APIToken.query(beta=token).dialog_id` lookup in
 		// bot_api.py:agent_bot_logs.
-		if u, code, err := h.userService.GetUserByBetaAPIToken(auth); err == nil && code == common.CodeSuccess {
+		if u, code, err := h.userService.GetUserByBetaAPIToken(ctx, auth); err == nil && code == common.CodeSuccess {
 			c.Set("user", u)
-			if tok, terr := h.userService.GetAPITokenByBeta(auth); terr == nil && tok != nil && tok.DialogID != nil {
+			if tok, terr := h.userService.GetAPITokenByBeta(ctx, auth); terr == nil && tok != nil && tok.DialogID != nil {
 				// tok.DialogID is *string (nullable in the schema), but
 				// downstream handlers (GetAgentbotLogs, GetAgentLogs)
 				// read "agent_id" with agentID.(string) — they cannot
@@ -126,6 +128,7 @@ func (h *AuthHandler) BetaAuthMiddleware() gin.HandlerFunc {
 // Validates that the user is authenticated and is a superuser (admin)
 func (h *AuthHandler) AuthMiddleware() gin.HandlerFunc {
 	return func(c *gin.Context) {
+		ctx := c.Request.Context()
 		token := c.GetHeader("Authorization")
 		if token == "" {
 			common.ResponseWithHttpCodeData(c, http.StatusUnauthorized, 401, nil, "Missing Authorization header")
@@ -136,9 +139,9 @@ func (h *AuthHandler) AuthMiddleware() gin.HandlerFunc {
 		authViaAPIToken := false
 
 		// Get user by access token
-		user, code, err := h.userService.GetUserByToken(token)
+		user, code, err := h.userService.GetUserByToken(ctx, token)
 		if err != nil {
-			user, code, err = h.userService.GetUserByAPIToken(token)
+			user, code, err = h.userService.GetUserByAPIToken(ctx, token)
 			if err != nil {
 				common.ResponseWithHttpCodeData(c, http.StatusUnauthorized, code, nil, "Invalid access token")
 				c.Abort()
diff --git a/internal/handler/bot_test.go b/internal/handler/bot_test.go
index 7e12236a58..c0664e1c6e 100644
--- a/internal/handler/bot_test.go
+++ b/internal/handler/bot_test.go
@@ -784,7 +784,7 @@ func TestBotRoutes_RequireAuth(t *testing.T) {
 func TestBotMiddleware_NonBearerRegularToken(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	stub := &stubUserTokenResolver{
-		getUserByTokenFn: func(auth string) (*entity.User, common.ErrorCode, error) {
+		getUserByTokenFn: func(ctx context.Context, auth string) (*entity.User, common.ErrorCode, error) {
 			if auth != "raw-access-token-abc" {
 				t.Errorf("GetUserByToken called with %q, want raw-access-token-abc", auth)
 			}
@@ -820,36 +820,36 @@ func TestBotMiddleware_NonBearerRegularToken(t *testing.T) {
 // return safe defaults (CodeUnauthorized so the middleware
 // short-circuits to 401).
 type stubUserTokenResolver struct {
-	getUserByTokenFn        func(authorization string) (*entity.User, common.ErrorCode, error)
-	getUserByAPITokenFn     func(token string) (*entity.User, common.ErrorCode, error)
-	getUserByBetaAPITokenFn func(token string) (*entity.User, common.ErrorCode, error)
-	getAPITokenByBetaFn     func(authorization string) (*entity.APIToken, error)
+	getUserByTokenFn        func(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error)
+	getUserByAPITokenFn     func(ctx context.Context, token string) (*entity.User, common.ErrorCode, error)
+	getUserByBetaAPITokenFn func(ctx context.Context, token string) (*entity.User, common.ErrorCode, error)
+	getAPITokenByBetaFn     func(ctx context.Context, authorization string) (*entity.APIToken, error)
 }
 
-func (s *stubUserTokenResolver) GetUserByToken(authorization string) (*entity.User, common.ErrorCode, error) {
+func (s *stubUserTokenResolver) GetUserByToken(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error) {
 	if s.getUserByTokenFn != nil {
-		return s.getUserByTokenFn(authorization)
+		return s.getUserByTokenFn(ctx, authorization)
 	}
 	return nil, common.CodeUnauthorized, errors.New("not stubbed")
 }
 
-func (s *stubUserTokenResolver) GetUserByAPIToken(token string) (*entity.User, common.ErrorCode, error) {
+func (s *stubUserTokenResolver) GetUserByAPIToken(ctx context.Context, token string) (*entity.User, common.ErrorCode, error) {
 	if s.getUserByAPITokenFn != nil {
-		return s.getUserByAPITokenFn(token)
+		return s.getUserByAPITokenFn(ctx, token)
 	}
 	return nil, common.CodeUnauthorized, errors.New("not stubbed")
 }
 
-func (s *stubUserTokenResolver) GetUserByBetaAPIToken(token string) (*entity.User, common.ErrorCode, error) {
+func (s *stubUserTokenResolver) GetUserByBetaAPIToken(ctx context.Context, token string) (*entity.User, common.ErrorCode, error) {
 	if s.getUserByBetaAPITokenFn != nil {
-		return s.getUserByBetaAPITokenFn(token)
+		return s.getUserByBetaAPITokenFn(ctx, token)
 	}
 	return nil, common.CodeUnauthorized, errors.New("not stubbed")
 }
 
-func (s *stubUserTokenResolver) GetAPITokenByBeta(authorization string) (*entity.APIToken, error) {
+func (s *stubUserTokenResolver) GetAPITokenByBeta(ctx context.Context, authorization string) (*entity.APIToken, error) {
 	if s.getAPITokenByBetaFn != nil {
-		return s.getAPITokenByBetaFn(authorization)
+		return s.getAPITokenByBetaFn(ctx, authorization)
 	}
 	return nil, errors.New("not stubbed")
 }
@@ -870,7 +870,7 @@ func (s *stubUserTokenResolver) GetAPITokenByBeta(authorization string) (*entity
 func TestBotRoutes_NoRegularAuthRequired(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	stub := &stubUserTokenResolver{
-		getUserByTokenFn: func(auth string) (*entity.User, common.ErrorCode, error) {
+		getUserByTokenFn: func(ctx context.Context, auth string) (*entity.User, common.ErrorCode, error) {
 			return &entity.User{ID: "u-regular"}, common.CodeSuccess, nil
 		},
 	}
@@ -955,7 +955,7 @@ func TestBotRoutes_NoRegularAuthRequired(t *testing.T) {
 func TestDownloadAttachment_Unauth(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	stub := &stubUserTokenResolver{
-		getUserByTokenFn: func(auth string) (*entity.User, common.ErrorCode, error) {
+		getUserByTokenFn: func(ctx context.Context, auth string) (*entity.User, common.ErrorCode, error) {
 			return nil, common.CodeUnauthorized, errors.New("invalid token")
 		},
 	}
@@ -968,13 +968,14 @@ func TestDownloadAttachment_Unauth(t *testing.T) {
 	// resolve via GetUserByToken. Both branches must reject
 	// with 401.
 	g.Use(func(c *gin.Context) {
+		ctx := c.Request.Context()
 		auth := c.GetHeader("Authorization")
 		if auth == "" {
 			common.ResponseWithCodeData(c, common.CodeUnauthorized, nil, "Authorization required")
 			c.Abort()
 			return
 		}
-		if u, code, err := stub.GetUserByToken(auth); err != nil || code != common.CodeSuccess {
+		if u, code, err := stub.GetUserByToken(ctx, auth); err != nil || code != common.CodeSuccess {
 			common.ResponseWithCodeData(c, common.CodeUnauthorized, nil, "Invalid auth credentials")
 			c.Abort()
 			return
diff --git a/internal/handler/chat_audio.go b/internal/handler/chat_audio.go
index 62000e6637..2197e51a5f 100644
--- a/internal/handler/chat_audio.go
+++ b/internal/handler/chat_audio.go
@@ -58,6 +58,7 @@ func (h *ChatHandler) ChatAudioSpeech(c *gin.Context) {
 		return
 	}
 
+	ctx := c.Request.Context()
 	c.Request.Body = http.MaxBytesReader(c.Writer, c.Request.Body, chatAudioSpeechMaxBodyBytes)
 
 	var req chatAudioSpeechRequest
@@ -84,7 +85,7 @@ func (h *ChatHandler) ChatAudioSpeech(c *gin.Context) {
 		if seg == "" {
 			continue
 		}
-		resp, err := driver.AudioSpeech(&modelName, &seg, apiConfig, &modelModule.TTSConfig{Format: "mp3"}, nil)
+		resp, err := driver.AudioSpeech(ctx, &modelName, &seg, apiConfig, &modelModule.TTSConfig{Format: "mp3"}, nil)
 		if err != nil {
 			common.Warn("chat TTS synthesis failed",
 				zap.Int("segmentIndex", i),
@@ -154,6 +155,7 @@ func (h *ChatHandler) ChatAudioTranscription(c *gin.Context) {
 		return
 	}
 
+	ctx := c.Request.Context()
 	// Cap the body before any multipart parsing so an oversized upload is
 	// rejected instead of being drained into memory or spooled to disk.
 	c.Request.Body = http.MaxBytesReader(c.Writer, c.Request.Body, chatAudioUploadMaxBytes)
@@ -234,7 +236,7 @@ func (h *ChatHandler) ChatAudioTranscription(c *gin.Context) {
 			return nil
 		}
 
-		if err = driver.TranscribeAudioWithSender(&modelName, &tmpPath, apiConfig, &modelModule.ASRConfig{}, nil, sender); err != nil {
+		if err = driver.TranscribeAudioWithSender(ctx, &modelName, &tmpPath, apiConfig, &modelModule.ASRConfig{}, nil, sender); err != nil {
 			errEvent := map[string]interface{}{"event": "error", "text": err.Error()}
 			data, _ := json.Marshal(errEvent)
 			_, _ = c.Writer.WriteString(fmt.Sprintf("data: %s\n\n", data))
@@ -252,7 +254,7 @@ func (h *ChatHandler) ChatAudioTranscription(c *gin.Context) {
 		return
 	}
 
-	resp, err := driver.TranscribeAudio(&modelName, &tmpPath, apiConfig, &modelModule.ASRConfig{}, nil)
+	resp, err := driver.TranscribeAudio(ctx, &modelName, &tmpPath, apiConfig, &modelModule.ASRConfig{}, nil)
 	if err != nil {
 		common.ErrorWithCode(c, common.CodeServerError, err.Error())
 		return
diff --git a/internal/handler/chat_recommendation.go b/internal/handler/chat_recommendation.go
index b47ed6609e..5edfb868de 100644
--- a/internal/handler/chat_recommendation.go
+++ b/internal/handler/chat_recommendation.go
@@ -56,7 +56,8 @@ func (h *ChatHandler) Recommendation(c *gin.Context) {
 		common.ResponseWithCodeData(c, common.CodeArgumentError, nil, "question is required")
 		return
 	}
-	questions, err := service.GenerateRelatedQuestions(user.ID, req.Question, req.SearchID, h.searchSvc, h.tenantSvc, h.llm)
+	ctx := c.Request.Context()
+	questions, err := service.GenerateRelatedQuestions(ctx, user.ID, req.Question, req.SearchID, h.searchSvc, h.tenantSvc, h.llm)
 	if err != nil {
 		jsonInternalError(c, err)
 		return
diff --git a/internal/handler/chunk.go b/internal/handler/chunk.go
index 5c9d61f4ed..d21d105c7b 100644
--- a/internal/handler/chunk.go
+++ b/internal/handler/chunk.go
@@ -16,6 +16,7 @@
 package handler
 
 import (
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -39,7 +40,7 @@ type chunkService interface {
 	UpdateChunk(req *service.UpdateChunkRequest, userID string) error
 	RemoveChunks(req *service.RemoveChunksRequest, userID string) (int64, error)
 	Parse(userID, datasetID string, req *service.ParseFileRequest) (map[string]interface{}, common.ErrorCode, error)
-	AddChunk(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error)
+	AddChunk(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error)
 	StopParsing(userID, datasetID string, req service.StopParsingRequest) (*service.StopParsingResponse, common.ErrorCode, error)
 }
 
@@ -698,6 +699,7 @@ func addChunkResponseMessage(code common.ErrorCode, err error) string {
 }
 
 func (h *ChunkHandler) AddChunk(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -756,7 +758,7 @@ func (h *ChunkHandler) AddChunk(c *gin.Context) {
 		ImageBase64:       imageBase64,
 	}
 
-	resp, err := h.chunkService.AddChunk(&req, userID)
+	resp, err := h.chunkService.AddChunk(ctx, &req, userID)
 	if err != nil {
 		var codedErr service.ErrorCoder
 		if errors.As(err, &codedErr) {
diff --git a/internal/handler/chunk_test.go b/internal/handler/chunk_test.go
index 828a1abe3a..8a0516475b 100644
--- a/internal/handler/chunk_test.go
+++ b/internal/handler/chunk_test.go
@@ -1,6 +1,7 @@
 package handler
 
 import (
+	"context"
 	"encoding/json"
 	"errors"
 	"net/http"
@@ -20,7 +21,7 @@ import (
 // Only the methods actually called by the test are set; others panic.
 type mockChunkSvc struct {
 	retrievalTestFn func(req *service.RetrievalTestRequest, userID string) (*service.RetrievalTestResponse, error)
-	addChunkFn      func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error)
+	addChunkFn      func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error)
 	listFn          func(req *service.ListChunksRequest, userID string) (*service.ListChunksResponse, error)
 	switchChunksFn  func(userID, datasetID, documentID string, availableInt int, chunkIDs []string) error
 	updateChunkFn   func(req *service.UpdateChunkRequest, userID string) error
@@ -82,9 +83,9 @@ func (m *mockChunkSvc) StopParsing(userID, datasetID string, req service.StopPar
 func (m *mockChunkSvc) Parse(string, string, *service.ParseFileRequest) (map[string]interface{}, common.ErrorCode, error) {
 	panic("not implemented")
 }
-func (m *mockChunkSvc) AddChunk(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+func (m *mockChunkSvc) AddChunk(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 	if m.addChunkFn != nil {
-		return m.addChunkFn(req, userID)
+		return m.addChunkFn(ctx, req, userID)
 	}
 	return &service.AddChunkResponse{Chunk: map[string]interface{}{"id": "chunk-1"}}, nil
 }
@@ -641,7 +642,7 @@ func (e addChunkTestError) Code() common.ErrorCode { return e.code }
 
 func TestChunkHandlerAddChunkSuccess(t *testing.T) {
 	mock := &mockChunkSvc{
-		addChunkFn: func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+		addChunkFn: func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 			if userID != "user1" {
 				t.Fatalf("userID = %q, want user1", userID)
 			}
@@ -679,7 +680,7 @@ func TestChunkHandlerAddChunkSuccess(t *testing.T) {
 
 func TestChunkHandlerAddChunkPathIDsOverrideBody(t *testing.T) {
 	mock := &mockChunkSvc{
-		addChunkFn: func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+		addChunkFn: func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 			if req.DatasetID != "kb1" || req.DocumentID != "doc1" {
 				t.Fatalf("path IDs were not preserved: %#v", req)
 			}
@@ -710,7 +711,7 @@ func TestChunkHandlerAddChunkPathIDsOverrideBody(t *testing.T) {
 
 func TestChunkHandlerAddChunkCodedError(t *testing.T) {
 	mock := &mockChunkSvc{
-		addChunkFn: func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+		addChunkFn: func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 			return nil, addChunkTestError{code: common.CodeDataError, msg: "`content` is required"}
 		},
 	}
@@ -761,7 +762,7 @@ func TestChunkHandlerAddChunkValidatesListFields(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			mock := &mockChunkSvc{
-				addChunkFn: func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+				addChunkFn: func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 					t.Fatal("service should not be called for invalid request")
 					return nil, nil
 				},
@@ -792,7 +793,7 @@ func TestChunkHandlerAddChunkValidatesListFields(t *testing.T) {
 
 func TestChunkHandlerAddChunkHidesServerErrorDetails(t *testing.T) {
 	mock := &mockChunkSvc{
-		addChunkFn: func(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+		addChunkFn: func(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 			return nil, addChunkTestError{code: common.CodeServerError, msg: "encode chunk embedding: provider secret"}
 		},
 	}
diff --git a/internal/handler/dataset.go b/internal/handler/dataset.go
index 099d0985f2..1ebbd2b432 100644
--- a/internal/handler/dataset.go
+++ b/internal/handler/dataset.go
@@ -708,7 +708,8 @@ func (h *DatasetsHandler) CheckEmbedding(c *gin.Context) {
 		return
 	}
 
-	data, code, err := h.datasetsService.CheckEmbedding(userID, datasetID, &req)
+	ctx := c.Request.Context()
+	data, code, err := h.datasetsService.CheckEmbedding(ctx, userID, datasetID, &req)
 	if err != nil {
 		if code == common.CodeNotEffective {
 			common.ResponseWithCodeData(c, code, data, err.Error())
diff --git a/internal/handler/providers.go b/internal/handler/providers.go
index 53eca76eed..b24daaba75 100644
--- a/internal/handler/providers.go
+++ b/internal/handler/providers.go
@@ -190,6 +190,7 @@ func (h *ProviderHandler) CreateProviderInstance(c *gin.Context) {
 		return
 	}
 
+	ctx := c.Request.Context()
 	var req CreateProviderInstanceRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		common.ErrorWithCode(c, common.CodeBadRequest, err.Error())
@@ -211,7 +212,7 @@ func (h *ProviderHandler) CreateProviderInstance(c *gin.Context) {
 		return
 	}
 
-	_, err := h.modelProviderService.CreateProviderInstance(providerName, req.InstanceName, req.APIKey, req.BaseURL, req.Region, userID, req.ModelInfo)
+	_, err := h.modelProviderService.CreateProviderInstance(ctx, providerName, req.InstanceName, req.APIKey, req.BaseURL, req.Region, userID, req.ModelInfo)
 	if err != nil {
 		common.ErrorWithCode(c, common.CodeServerError, err.Error())
 		return
@@ -276,9 +277,10 @@ func (h *ProviderHandler) ShowInstanceBalance(c *gin.Context) {
 	}
 
 	userID := c.GetString("user_id")
+	ctx := c.Request.Context()
 
 	// Get tenant ID from user
-	balance, errorCode, err := h.modelProviderService.ShowInstanceBalance(providerName, instanceName, userID)
+	balance, errorCode, err := h.modelProviderService.ShowInstanceBalance(ctx, providerName, instanceName, userID)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -294,6 +296,7 @@ func (h *ProviderHandler) CheckConnection(c *gin.Context) {
 		return
 	}
 
+	ctx := c.Request.Context()
 	var req service.CheckConnectionRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, err.Error())
@@ -301,7 +304,7 @@ func (h *ProviderHandler) CheckConnection(c *gin.Context) {
 	}
 
 	userID := c.GetString("user_id")
-	errCode, err := h.modelProviderService.CheckConnection(providerName, req.APIKey, req.Region, req.BaseURL, req.InstanceID, userID, service.ListModelNames(req.ModelInfo))
+	errCode, err := h.modelProviderService.CheckConnection(ctx, providerName, req.APIKey, req.Region, req.BaseURL, req.InstanceID, userID, service.ListModelNames(req.ModelInfo))
 	if err != nil {
 		common.ErrorWithCode(c, errCode, err.Error())
 		return
@@ -311,6 +314,7 @@ func (h *ProviderHandler) CheckConnection(c *gin.Context) {
 }
 
 func (h *ProviderHandler) CheckInstanceConnection(c *gin.Context) {
+	ctx := c.Request.Context()
 	providerName := c.Param("provider_name")
 	if providerName == "" {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, "Provider name is required")
@@ -336,7 +340,7 @@ func (h *ProviderHandler) CheckInstanceConnection(c *gin.Context) {
 	baseURL, _ := instanceInfo["base_url"].(string)
 	instanceID, _ := instanceInfo["id"].(string)
 
-	errorCode, err := h.modelProviderService.CheckConnection(providerName, apikey, region, baseURL, instanceID, userID, nil)
+	errorCode, err := h.modelProviderService.CheckConnection(ctx, providerName, apikey, region, baseURL, instanceID, userID, nil)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -346,6 +350,7 @@ func (h *ProviderHandler) CheckInstanceConnection(c *gin.Context) {
 }
 
 func (h *ProviderHandler) ListTasks(c *gin.Context) {
+	ctx := c.Request.Context()
 	providerName := c.Param("provider_name")
 	if providerName == "" {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, "Provider name is required")
@@ -361,7 +366,7 @@ func (h *ProviderHandler) ListTasks(c *gin.Context) {
 	userID := c.GetString("user_id")
 
 	// Get tenant ID from user
-	listTaskResponse, errorCode, err := h.modelProviderService.ListTasks(providerName, instanceName, userID)
+	listTaskResponse, errorCode, err := h.modelProviderService.ListTasks(ctx, providerName, instanceName, userID)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -371,6 +376,7 @@ func (h *ProviderHandler) ListTasks(c *gin.Context) {
 }
 
 func (h *ProviderHandler) ShowTask(c *gin.Context) {
+	ctx := c.Request.Context()
 	providerName := c.Param("provider_name")
 	if providerName == "" {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, "Provider name is required")
@@ -392,7 +398,7 @@ func (h *ProviderHandler) ShowTask(c *gin.Context) {
 	userID := c.GetString("user_id")
 
 	// Get tenant ID from user
-	taskResponse, errorCode, err := h.modelProviderService.ShowTask(providerName, instanceName, taskID, userID)
+	taskResponse, errorCode, err := h.modelProviderService.ShowTask(ctx, providerName, instanceName, taskID, userID)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -411,6 +417,7 @@ type AlterProviderInstanceRequest struct {
 }
 
 func (h *ProviderHandler) AlterProviderInstance(c *gin.Context) {
+	ctx := c.Request.Context()
 	providerName := c.Param("provider_name")
 	if providerName == "" {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, "Provider name is required")
@@ -440,7 +447,7 @@ func (h *ProviderHandler) AlterProviderInstance(c *gin.Context) {
 		verify = *req.Verify
 	}
 
-	code, err := h.modelProviderService.AlterProviderInstance(userID, providerName, instanceName, req.InstanceName, req.APIKey, req.BaseURL, req.Region, req.ModelInfo, verify)
+	code, err := h.modelProviderService.AlterProviderInstance(ctx, userID, providerName, instanceName, req.InstanceName, req.APIKey, req.BaseURL, req.Region, req.ModelInfo, verify)
 	if err != nil {
 		common.ErrorWithCode(c, code, err.Error())
 		return
@@ -477,6 +484,7 @@ func (h *ProviderHandler) DropProviderInstance(c *gin.Context) {
 }
 
 func (h *ProviderHandler) ListInstanceModels(c *gin.Context) {
+	ctx := c.Request.Context()
 	providerName := c.Param("provider_name")
 	if providerName == "" {
 		common.ResponseWithHttpCodeData(c, http.StatusBadRequest, 400, nil, "Provider name is required")
@@ -498,7 +506,7 @@ func (h *ProviderHandler) ListInstanceModels(c *gin.Context) {
 	if keywords == "true" {
 		// list supported models
 
-		modelList, err := h.modelProviderService.ListSupportedModels(providerName, instanceName, c.GetString("user_id"))
+		modelList, err := h.modelProviderService.ListSupportedModels(ctx, providerName, instanceName, c.GetString("user_id"))
 		if err != nil {
 			common.ErrorWithCode(c, common.CodeServerError, err.Error())
 			return
@@ -694,6 +702,7 @@ type ChatToModelRequest struct {
 }
 
 func (h *ProviderHandler) ChatToModel(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req ChatToModelRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -799,6 +808,7 @@ func (h *ProviderHandler) ChatToModel(c *gin.Context) {
 
 		// Stream response using sender function (the best performance, no channel)
 		errorCode, err := h.modelProviderService.ChatToModelStreamWithSender(
+			ctx,
 			req.ProviderName,
 			req.InstanceName,
 			req.ModelName,
@@ -830,6 +840,7 @@ func (h *ProviderHandler) ChatToModel(c *gin.Context) {
 		messages[i] = models.Message{Role: role, Content: content}
 	}
 	response, errorCode, err = h.modelProviderService.ChatToModelWithMessages(
+		ctx,
 		req.ProviderName,
 		req.InstanceName,
 		req.ModelName,
@@ -864,6 +875,7 @@ type EmbedTextRequest struct {
 }
 
 func (h *ProviderHandler) EmbedText(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req EmbedTextRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -909,7 +921,7 @@ func (h *ProviderHandler) EmbedText(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.EmbedText(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Texts, &apiConfig, &embeddingConfig)
+	response, errorCode, err = h.modelProviderService.EmbedText(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Texts, &apiConfig, &embeddingConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -929,6 +941,7 @@ type RerankDocumentRequest struct {
 }
 
 func (h *ProviderHandler) RerankDocument(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req RerankDocumentRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -974,7 +987,7 @@ func (h *ProviderHandler) RerankDocument(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.RerankDocument(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Query, req.Documents, &apiConfig, &rerankConfig)
+	response, errorCode, err = h.modelProviderService.RerankDocument(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Query, req.Documents, &apiConfig, &rerankConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -995,6 +1008,7 @@ type TranscribeAudioRequest struct {
 }
 
 func (h *ProviderHandler) TranscribeAudio(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req TranscribeAudioRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -1070,7 +1084,7 @@ func (h *ProviderHandler) TranscribeAudio(c *gin.Context) {
 		}
 
 		// Stream response using sender function ( the best performance, no channel)
-		errorCode, err := h.modelProviderService.TranscribeAudioStream(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.File, &apiConfig, &asrConfig, sender)
+		errorCode, err := h.modelProviderService.TranscribeAudioStream(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.File, &apiConfig, &asrConfig, sender)
 		if errorCode != common.CodeSuccess {
 			c.SSEvent("error", err.Error())
 		}
@@ -1082,7 +1096,7 @@ func (h *ProviderHandler) TranscribeAudio(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.TranscribeAudio(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.File, &apiConfig, &asrConfig)
+	response, errorCode, err = h.modelProviderService.TranscribeAudio(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.File, &apiConfig, &asrConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -1102,6 +1116,7 @@ type AudioSpeechRequest struct {
 }
 
 func (h *ProviderHandler) AudioSpeech(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req AudioSpeechRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -1177,7 +1192,7 @@ func (h *ProviderHandler) AudioSpeech(c *gin.Context) {
 		}
 
 		// Stream response using sender function ( the best performance, no channel)
-		errorCode, err := h.modelProviderService.AudioSpeechStream(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Text, &apiConfig, &ttsConfig, sender)
+		errorCode, err := h.modelProviderService.AudioSpeechStream(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Text, &apiConfig, &ttsConfig, sender)
 		if errorCode != common.CodeSuccess {
 			c.SSEvent("error", err.Error())
 		}
@@ -1189,7 +1204,7 @@ func (h *ProviderHandler) AudioSpeech(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.AudioSpeech(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Text, &apiConfig, &ttsConfig)
+	response, errorCode, err = h.modelProviderService.AudioSpeech(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Text, &apiConfig, &ttsConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -1208,6 +1223,7 @@ type OCRFileRequest struct {
 }
 
 func (h *ProviderHandler) OCRFile(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req OCRFileRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -1251,7 +1267,7 @@ func (h *ProviderHandler) OCRFile(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.OCRFile(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Content, req.URL, &apiConfig, &OCRConfig)
+	response, errorCode, err = h.modelProviderService.OCRFile(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Content, req.URL, &apiConfig, &OCRConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
@@ -1270,6 +1286,7 @@ type ParseFileRequest struct {
 }
 
 func (h *ProviderHandler) ParseFile(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req ParseFileRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		println("JSON bind error: %v (type: %T)", err, err)
@@ -1313,7 +1330,7 @@ func (h *ProviderHandler) ParseFile(c *gin.Context) {
 	var errorCode common.ErrorCode
 	var err error
 
-	response, errorCode, err = h.modelProviderService.ParseFile(req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Content, req.URL, &apiConfig, &parseFileConfig)
+	response, errorCode, err = h.modelProviderService.ParseFile(ctx, req.ProviderName, req.InstanceName, req.ModelName, req.ModelID, userID, req.Content, req.URL, &apiConfig, &parseFileConfig)
 	if err != nil {
 		common.ErrorWithCode(c, errorCode, err.Error())
 		return
diff --git a/internal/handler/searchbot.go b/internal/handler/searchbot.go
index 933b65ee12..1af77240d9 100644
--- a/internal/handler/searchbot.go
+++ b/internal/handler/searchbot.go
@@ -131,6 +131,7 @@ func (h *SearchBotHandler) SetAskService(svc *service.AskService) { h.askSvc = s
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/searchbots/related_questions [post]
 func (h *SearchBotHandler) Handle(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -148,7 +149,7 @@ func (h *SearchBotHandler) Handle(c *gin.Context) {
 		return
 	}
 
-	questions, err := service.GenerateRelatedQuestions(user.ID, req.Question, req.SearchID, h.searchSvc, h.tenantSvc, h.llm)
+	questions, err := service.GenerateRelatedQuestions(ctx, user.ID, req.Question, req.SearchID, h.searchSvc, h.tenantSvc, h.llm)
 	if err != nil {
 		common.Warn("searchbot related questions failed", zap.String("error", err.Error()))
 		common.ResponseWithCodeData(c, common.CodeOperatingError, nil, "LLM call failed")
@@ -378,10 +379,11 @@ func (h *SearchBotHandler) MindMap(c *gin.Context) {
 	common.SuccessWithData(c, mindMap, "success")
 }
 
-// SearchbotDetail returns the public share-page bootstrap payload for a
+// SearchBotDetail returns the public share-page bootstrap payload for a
 // search app. The route is mounted under apiNoAuth but still requires a beta
 // token, matching Python's AUTH_BETA flow.
-func (h *SearchBotHandler) SearchbotDetail(c *gin.Context) {
+func (h *SearchBotHandler) SearchBotDetail(c *gin.Context) {
+	ctx := c.Request.Context()
 	searchID := strings.TrimSpace(c.Query("search_id"))
 	if searchID == "" {
 		common.ResponseWithCodeData(c, common.CodeArgumentError, nil, "search_id is required")
@@ -389,7 +391,7 @@ func (h *SearchBotHandler) SearchbotDetail(c *gin.Context) {
 	}
 
 	userSvc := service.NewUserService()
-	user, code, err := userSvc.GetUserByBetaAPIToken(c.GetHeader("Authorization"))
+	user, code, err := userSvc.GetUserByBetaAPIToken(ctx, c.GetHeader("Authorization"))
 	if err != nil {
 		common.ResponseWithCodeData(c, code, nil, "Authentication error: API key is invalid!")
 		return
diff --git a/internal/handler/searchbot_detail_test.go b/internal/handler/searchbot_detail_test.go
index 479db34c16..39a7db7883 100644
--- a/internal/handler/searchbot_detail_test.go
+++ b/internal/handler/searchbot_detail_test.go
@@ -58,7 +58,7 @@ func newSearchbotDetailRouter() *gin.Engine {
 	gin.SetMode(gin.TestMode)
 	h := NewSearchBotHandler(service.NewSearchService(), nil, nil, nil)
 	r := gin.New()
-	r.GET("/api/v1/searchbots/detail", h.SearchbotDetail)
+	r.GET("/api/v1/searchbots/detail", h.SearchBotDetail)
 	return r
 }
 
diff --git a/internal/handler/user.go b/internal/handler/user.go
index a2030e2a89..7985eadc1f 100644
--- a/internal/handler/user.go
+++ b/internal/handler/user.go
@@ -56,6 +56,7 @@ func NewUserHandler(userService *service.UserService) *UserHandler {
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/users [post]
 func (h *UserHandler) Register(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req service.RegisterRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		common.ResponseWithCodeData(c, common.CodeBadRequest, false, err.Error())
@@ -90,7 +91,7 @@ func (h *UserHandler) Register(c *gin.Context) {
 	c.Header("Access-Control-Allow-Headers", "*")
 	c.Header("Access-Control-Expose-Headers", "Authorization")
 
-	profile := h.userService.GetUserProfile(user)
+	profile := h.userService.GetUserProfile(ctx, user)
 	common.SuccessWithData(c, profile, fmt.Sprintf("%s, welcome aboard!", req.Nickname))
 }
 
@@ -104,6 +105,7 @@ func (h *UserHandler) Register(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/users/login [post]
 func (h *UserHandler) Login(c *gin.Context) {
+	ctx := c.Request.Context()
 	startAt := time.Now()
 	operationLog := &common.OperationLog{
 		EventTime:  startAt,
@@ -170,7 +172,7 @@ func (h *UserHandler) Login(c *gin.Context) {
 	c.Header("Access-Control-Allow-Headers", "*")
 	c.Header("Access-Control-Expose-Headers", "Authorization")
 
-	profile := h.userService.GetUserProfile(user)
+	profile := h.userService.GetUserProfile(ctx, user)
 	common.SuccessWithData(c, profile, "Welcome back!")
 }
 
@@ -184,6 +186,7 @@ func (h *UserHandler) Login(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /v1/user/login [post]
 func (h *UserHandler) LoginByEmail(c *gin.Context) {
+	ctx := c.Request.Context()
 	startAt := time.Now()
 	operationLog := &common.OperationLog{
 		EventTime:  startAt,
@@ -221,7 +224,7 @@ func (h *UserHandler) LoginByEmail(c *gin.Context) {
 		return
 	}
 
-	user, code, err := h.userService.LoginByEmail(&req)
+	user, code, err := h.userService.LoginByEmail(ctx, &req)
 	if err != nil {
 		common.ResponseWithCodeData(c, code, false, err.Error())
 		operationLog.ErrorCode = uint16(code)
@@ -255,7 +258,7 @@ func (h *UserHandler) LoginByEmail(c *gin.Context) {
 	c.Header("Access-Control-Allow-Headers", "*")
 	c.Header("Access-Control-Expose-Headers", "Authorization")
 
-	profile := h.userService.GetUserProfile(user)
+	profile := h.userService.GetUserProfile(ctx, user)
 	common.SuccessWithData(c, profile, "Welcome back!")
 }
 
@@ -269,6 +272,7 @@ func (h *UserHandler) LoginByEmail(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/users/{id} [get]
 func (h *UserHandler) GetUserByID(c *gin.Context) {
+	ctx := c.Request.Context()
 	idStr := c.Param("id")
 	id, err := strconv.ParseUint(idStr, 10, 32)
 	if err != nil {
@@ -276,7 +280,7 @@ func (h *UserHandler) GetUserByID(c *gin.Context) {
 		return
 	}
 
-	user, code, err := h.userService.GetUserByID(uint(id))
+	user, code, err := h.userService.GetUserByID(ctx, uint(id))
 	if err != nil {
 		common.ResponseWithCodeData(c, code, false, err.Error())
 		return
@@ -295,6 +299,7 @@ func (h *UserHandler) GetUserByID(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /v1/user/logout [post]
 func (h *UserHandler) Logout(c *gin.Context) {
+	ctx := c.Request.Context()
 	startAt := time.Now()
 	operationLog := &common.OperationLog{
 		EventTime:  startAt,
@@ -335,7 +340,7 @@ func (h *UserHandler) Logout(c *gin.Context) {
 	}
 
 	// Get user by access token
-	user, code, err := h.userService.GetUserByToken(token)
+	user, code, err := h.userService.GetUserByToken(ctx, token)
 	if err != nil {
 		common.ResponseWithHttpCodeData(c, http.StatusUnauthorized, code, nil, "Invalid access token")
 		c.Abort()
@@ -369,6 +374,7 @@ func (h *UserHandler) Logout(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /v1/user/info [get]
 func (h *UserHandler) Info(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -376,7 +382,7 @@ func (h *UserHandler) Info(c *gin.Context) {
 	}
 
 	// Get user profile
-	profile := h.userService.GetUserProfile(user)
+	profile := h.userService.GetUserProfile(ctx, user)
 
 	common.SuccessWithData(c, profile, "success")
 }
@@ -392,6 +398,7 @@ func (h *UserHandler) Info(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/users/me [patch]
 func (h *UserHandler) Setting(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -406,7 +413,7 @@ func (h *UserHandler) Setting(c *gin.Context) {
 	}
 
 	// Update user settings
-	code, err := h.userService.UpdateUserSettings(user, &req)
+	code, err := h.userService.UpdateUserSettings(ctx, user, &req)
 	if err != nil {
 		if code == common.CodeExceptionError {
 			common.ResponseWithCodeData(c, common.CodeExceptionError, false, err.Error())
@@ -430,6 +437,7 @@ func (h *UserHandler) Setting(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /v1/user/setting/password [post]
 func (h *UserHandler) ChangePassword(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -444,7 +452,7 @@ func (h *UserHandler) ChangePassword(c *gin.Context) {
 	}
 
 	// Change password
-	code, err := h.userService.ChangePassword(user, &req)
+	code, err := h.userService.ChangePassword(ctx, user, &req)
 	if err != nil {
 		common.ResponseWithCodeData(c, code, false, err.Error())
 		return
@@ -482,6 +490,7 @@ func (h *UserHandler) GetLoginChannels(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /v1/user/set_tenant_info [post]
 func (h *UserHandler) SetTenantInfo(c *gin.Context) {
+	ctx := c.Request.Context()
 	user, errorCode, errorMessage := GetUser(c)
 	if errorCode != common.CodeSuccess {
 		common.ErrorWithCode(c, errorCode, errorMessage)
@@ -531,7 +540,7 @@ func (h *UserHandler) SetTenantInfo(c *gin.Context) {
 		req.TTSID = &value
 	}
 
-	code, err := h.userService.SetTenantInfo(user.ID, &req)
+	code, err := h.userService.SetTenantInfo(ctx, user.ID, &req)
 	if err != nil {
 		common.ResponseWithCodeData(c, code, nil, err.Error())
 		return
@@ -684,6 +693,7 @@ func (h *UserHandler) ForgotVerifyOTP(c *gin.Context) {
 // @Success 200 {object} map[string]interface{}
 // @Router /api/v1/auth/password/reset [post]
 func (h *UserHandler) ForgotResetPassword(c *gin.Context) {
+	ctx := c.Request.Context()
 	var req service.ForgotResetPasswordRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		common.ResponseWithCodeData(c, common.CodeArgumentError, false, err.Error())
@@ -714,7 +724,7 @@ func (h *UserHandler) ForgotResetPassword(c *gin.Context) {
 	// already in the Authorization header). Mirror the Python contract
 	// `user.to_safe_dict(for_self=True)` by stripping those fields before
 	// writing. PR #15290 review.
-	profile := h.userService.GetUserProfile(user)
+	profile := h.userService.GetUserProfile(ctx, user)
 	delete(profile, "password")
 	delete(profile, "access_token")
 	common.SuccessWithData(c, profile, "Password reset successful. Logged in.")
diff --git a/internal/ingestion/component/docx_vision_dispatch.go b/internal/ingestion/component/docx_vision_dispatch.go
index 0cd81882f5..622bb968e5 100644
--- a/internal/ingestion/component/docx_vision_dispatch.go
+++ b/internal/ingestion/component/docx_vision_dispatch.go
@@ -27,6 +27,7 @@
 package component
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -71,6 +72,7 @@ var (
 // It returns (result, handled, error). handled is true when the
 // dispatched result was modified.
 func maybeDispatchDOCXVision(
+	ctx context.Context,
 	fileType utility.FileType,
 	dispatched parserDispatchResult,
 	inputs map[string]any,
@@ -125,7 +127,7 @@ func maybeDispatchDOCXVision(
 			}
 
 			messages := buildVisionMessages(prompt, imageB64)
-			resp, err := visionChatInvoker(driver, modelName, messages, apiConfig)
+			resp, err := visionChatInvoker(ctx, driver, modelName, messages, apiConfig)
 			if err != nil {
 				return
 			}
@@ -290,11 +292,12 @@ func extractDOCXVisionAnswer(resp *modelModule.ChatResponse) string {
 }
 
 func defaultVisionChatInvoker(
+	ctx context.Context,
 	driver modelModule.ModelDriver,
 	modelName string,
 	messages []modelModule.Message,
 	apiConfig *modelModule.APIConfig,
 ) (*modelModule.ChatResponse, error) {
 	vision := true
-	return driver.ChatWithMessages(modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
+	return driver.ChatWithMessages(ctx, modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
 }
diff --git a/internal/ingestion/component/markdown_vision_dispatch.go b/internal/ingestion/component/markdown_vision_dispatch.go
index 88a3294da8..59f3793d6b 100644
--- a/internal/ingestion/component/markdown_vision_dispatch.go
+++ b/internal/ingestion/component/markdown_vision_dispatch.go
@@ -28,6 +28,7 @@
 package component
 
 import (
+	"context"
 	"fmt"
 	"strings"
 	"sync"
@@ -53,6 +54,7 @@ var (
 // path produces JSON items with doc_type_kwd == "image" and an
 // "image" base64 field. It returns (result, handled, error).
 func maybeDispatchMarkdownVision(
+	ctx context.Context,
 	fileType utility.FileType,
 	dispatched parserDispatchResult,
 	inputs map[string]any,
@@ -122,7 +124,7 @@ func maybeDispatchMarkdownVision(
 			}
 
 			messages := buildVisionMessages(prompt, item.imageB64)
-			resp, err := visionChatInvoker(driver, modelName, messages, apiConfig)
+			resp, err := visionChatInvoker(ctx, driver, modelName, messages, apiConfig)
 			if err != nil {
 				return
 			}
diff --git a/internal/ingestion/component/media_dispatch.go b/internal/ingestion/component/media_dispatch.go
index 09a4c9b378..5285220a82 100644
--- a/internal/ingestion/component/media_dispatch.go
+++ b/internal/ingestion/component/media_dispatch.go
@@ -52,6 +52,7 @@ import (
 // Video dispatch: IMAGE2TEXT vision chat ---
 
 func maybeDispatchVideo(
+	ctx context.Context,
 	fileType utility.FileType,
 	filename string,
 	binary []byte,
@@ -94,7 +95,7 @@ func maybeDispatchVideo(
 		},
 	}}
 	vision := true
-	resp, err := driver.ChatWithMessages(modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
+	resp, err := driver.ChatWithMessages(ctx, modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
 	if err != nil {
 		return parserDispatchResult{}, true,
 			fmt.Errorf("Parser: video describe: %w", err)
@@ -124,6 +125,7 @@ func maybeDispatchVideo(
 //   4. Returns combined text
 
 func maybeDispatchImage(
+	ctx context.Context,
 	fileType utility.FileType,
 	filename string,
 	binary []byte,
@@ -221,7 +223,7 @@ func maybeDispatchImage(
 		},
 	}}
 	vision := true
-	resp, err := driver.ChatWithMessages(modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
+	resp, err := driver.ChatWithMessages(ctx, modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
 	if err != nil {
 		if ocrText != "" {
 			return parserDispatchResult{
@@ -262,6 +264,7 @@ func maybeDispatchImage(
 //   - Returns the transcription as text
 
 func maybeDispatchAudio(
+	ctx context.Context,
 	fileType utility.FileType,
 	filename string,
 	binary []byte,
@@ -294,7 +297,7 @@ func maybeDispatchAudio(
 	}
 	defer os.Remove(tmpFile)
 
-	resp, err := driver.TranscribeAudio(&modelName, &tmpFile, apiConfig, nil, nil)
+	resp, err := driver.TranscribeAudio(ctx, &modelName, &tmpFile, apiConfig, nil, nil)
 	if err != nil {
 		return parserDispatchResult{}, true,
 			fmt.Errorf("Parser: audio transcription: %w", err)
diff --git a/internal/ingestion/component/parser.go b/internal/ingestion/component/parser.go
index 432c8f3939..9b8e24cdf9 100644
--- a/internal/ingestion/component/parser.go
+++ b/internal/ingestion/component/parser.go
@@ -377,7 +377,7 @@ func (c *ParserComponent) Invoke(ctx context.Context, inputs map[string]any) (ma
 		}
 	}
 
-	dispatched, handledVision, visionErr := maybeDispatchPDFVision(fileTypeExt, filename, binary, inputs, c.Setups)
+	dispatched, handledVision, visionErr := maybeDispatchPDFVision(ctx, fileTypeExt, filename, binary, inputs, c.Setups)
 	if visionErr != nil {
 		return nil, visionErr
 	}
@@ -386,7 +386,7 @@ func (c *ParserComponent) Invoke(ctx context.Context, inputs map[string]any) (ma
 	if !handledVision {
 		// Video dispatch: IMAGE2TEXT vision chat.
 		// Mirrors Python's _video().
-		dispatched, handledMedia, visionErr = maybeDispatchVideo(fileTypeExt, filename, binary, inputs, c.Setups)
+		dispatched, handledMedia, visionErr = maybeDispatchVideo(ctx, fileTypeExt, filename, binary, inputs, c.Setups)
 		if visionErr != nil {
 			return nil, visionErr
 		}
@@ -395,7 +395,7 @@ func (c *ParserComponent) Invoke(ctx context.Context, inputs map[string]any) (ma
 	if !handledVision && !handledMedia {
 		// Image/Picture dispatch: OCR + IMAGE2TEXT vision describe.
 		// Mirrors Python's rag/app/picture.py:chunk() image branch.
-		dispatched, handledImage, visionErr = maybeDispatchImage(fileTypeExt, filename, binary, inputs, c.Setups)
+		dispatched, handledImage, visionErr = maybeDispatchImage(ctx, fileTypeExt, filename, binary, inputs, c.Setups)
 		if visionErr != nil {
 			return nil, visionErr
 		}
@@ -404,25 +404,25 @@ func (c *ParserComponent) Invoke(ctx context.Context, inputs map[string]any) (ma
 	if !handledVision && !handledMedia && !handledImage {
 		// Audio dispatch: SPEECH2TEXT transcription.
 		// Mirrors Python's rag/app/audio.py:chunk().
-		dispatched, handledAudio, visionErr = maybeDispatchAudio(fileTypeExt, filename, binary, inputs, c.Setups)
+		dispatched, handledAudio, visionErr = maybeDispatchAudio(ctx, fileTypeExt, filename, binary, inputs, c.Setups)
 		if visionErr != nil {
 			return nil, visionErr
 		}
 	}
 	if !handledVision && !handledMedia && !handledImage && !handledAudio {
-		dispatched = dispatchParse(fileTypeExt, filename, binary, c.Setups)
+		dispatched = dispatchParse(ctx, fileTypeExt, filename, binary, c.Setups)
 		dispatched = hydrateEmptyDispatchPayload(dispatched, binary)
 
 		// DOCX vision figure enhancement: enrich the markdown
 		// with LLM-generated descriptions of embedded images.
 		// Mirrors Python's vision_figure_parser_docx_wrapper_naive.
-		dispatched, _, _ = maybeDispatchDOCXVision(fileTypeExt, dispatched, inputs, c.Setups)
+		dispatched, _, _ = maybeDispatchDOCXVision(ctx, fileTypeExt, dispatched, inputs, c.Setups)
 
 		// Markdown vision figure enhancement: enrich parsed
 		// markdown JSON items with LLM-generated descriptions of
 		// referenced images (![alt](url)). Mirrors Python's
 		// enhance_media_sections_with_vision in _markdown.
-		dispatched, _, _ = maybeDispatchMarkdownVision(fileTypeExt, dispatched, inputs)
+		dispatched, _, _ = maybeDispatchMarkdownVision(ctx, fileTypeExt, dispatched, inputs)
 	}
 	// Known/supported families must fail loudly when dispatch or
 	// parsing breaks. Only unknown families keep the raw-text fallback.
diff --git a/internal/ingestion/component/parser_dispatch.go b/internal/ingestion/component/parser_dispatch.go
index e3a0ded7fb..981eb87f9b 100644
--- a/internal/ingestion/component/parser_dispatch.go
+++ b/internal/ingestion/component/parser_dispatch.go
@@ -24,6 +24,7 @@
 package component
 
 import (
+	"context"
 	"fmt"
 	"maps"
 	"strings"
@@ -138,7 +139,7 @@ func resolveOutputFormat(family string, setups map[string]schema.ParserSetup, al
 // re-reading setups. lib_type is no longer threaded through: the
 // Python dispatcher picks a single backend per family and the Go
 // constructors mirror that.
-func dispatchParse(fileType utility.FileType, filename string, data []byte, setups map[string]schema.ParserSetup) parserDispatchResult {
+func dispatchParse(ctx context.Context, fileType utility.FileType, filename string, data []byte, setups map[string]schema.ParserSetup) parserDispatchResult {
 	if fileType == utility.FileTypeOTHER {
 		// Unknown / unset family. The component treats the bytes
 		// as text pages; splitIntoPages handles it. We return no
@@ -159,7 +160,7 @@ func dispatchParse(fileType utility.FileType, filename string, data []byte, setu
 	}
 	configureParserFromSetups(p, fileType, setups)
 
-	res := p.ParseWithResult(filename, data)
+	res := p.ParseWithResult(ctx, filename, data)
 	if res.Err != nil {
 		return parserDispatchResult{Err: fmt.Errorf("Parser: %q: %w", fileType, res.Err)}
 	}
diff --git a/internal/ingestion/component/parser_dispatch_pdf_vision_cgo_test.go b/internal/ingestion/component/parser_dispatch_pdf_vision_cgo_test.go
index 52874131e1..b697bec671 100644
--- a/internal/ingestion/component/parser_dispatch_pdf_vision_cgo_test.go
+++ b/internal/ingestion/component/parser_dispatch_pdf_vision_cgo_test.go
@@ -35,7 +35,7 @@ func TestDispatch_PDFVisionJSON_RealPDFFixture(t *testing.T) {
 	}
 
 	var callCount atomic.Int32
-	pdfVisionChatInvoker = func(_ modelModule.ModelDriver, modelName string, messages []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
+	pdfVisionChatInvoker = func(ctx context.Context, _ modelModule.ModelDriver, modelName string, messages []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
 		if modelName != "fixture-vlm" {
 			t.Fatalf("modelName = %q, want fixture-vlm", modelName)
 		}
diff --git a/internal/ingestion/component/parser_dispatch_test.go b/internal/ingestion/component/parser_dispatch_test.go
index 350383e09d..197ad40692 100644
--- a/internal/ingestion/component/parser_dispatch_test.go
+++ b/internal/ingestion/component/parser_dispatch_test.go
@@ -430,7 +430,7 @@ func TestDispatch_PDFVisionJSON_UsesTenantAwareModel(t *testing.T) {
 		}
 		return nil, "resolved-vlm", nil, nil
 	}
-	pdfVisionChatInvoker = func(_ modelModule.ModelDriver, modelName string, messages []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
+	pdfVisionChatInvoker = func(ctx context.Context, _ modelModule.ModelDriver, modelName string, messages []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
 		if modelName != "resolved-vlm" {
 			return nil, fmt.Errorf("modelName = %q, want resolved-vlm", modelName)
 		}
@@ -508,7 +508,7 @@ func TestDispatch_PDFVisionJSON_PreservesEmptyPages(t *testing.T) {
 		return nil, "resolved-vlm", nil, nil
 	}
 	call := 0
-	pdfVisionChatInvoker = func(_ modelModule.ModelDriver, _ string, _ []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
+	pdfVisionChatInvoker = func(ctx context.Context, _ modelModule.ModelDriver, _ string, _ []modelModule.Message, _ *modelModule.APIConfig) (*modelModule.ChatResponse, error) {
 		call++
 		answer := ""
 		if call == 1 {
@@ -595,49 +595,49 @@ type mineruTestDriver struct{}
 
 func (d *mineruTestDriver) NewInstance(baseURL map[string]string) modelModule.ModelDriver { return d }
 func (d *mineruTestDriver) Name() string                                                  { return "mineru" }
-func (d *mineruTestDriver) ChatWithMessages(modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, chatModelConfig *modelModule.ChatConfig, usage *common.ModelUsage) (*modelModule.ChatResponse, error) {
+func (d *mineruTestDriver) ChatWithMessages(ctx context.Context, modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, chatModelConfig *modelModule.ChatConfig, usage *common.ModelUsage) (*modelModule.ChatResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) ChatStreamlyWithSender(modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *mineruTestDriver) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) Embed(modelName *string, texts []string, apiConfig *modelModule.APIConfig, embeddingConfig *modelModule.EmbeddingConfig, usage *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
+func (d *mineruTestDriver) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *modelModule.APIConfig, embeddingConfig *modelModule.EmbeddingConfig, usage *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) Rerank(modelName *string, query string, documents []string, apiConfig *modelModule.APIConfig, rerankConfig *modelModule.RerankConfig, usage *common.ModelUsage) (*modelModule.RerankResponse, error) {
+func (d *mineruTestDriver) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *modelModule.APIConfig, rerankConfig *modelModule.RerankConfig, usage *common.ModelUsage) (*modelModule.RerankResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) TranscribeAudio(modelName *string, file *string, apiConfig *modelModule.APIConfig, asrConfig *modelModule.ASRConfig, usage *common.ModelUsage) (*modelModule.ASRResponse, error) {
+func (d *mineruTestDriver) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *modelModule.APIConfig, asrConfig *modelModule.ASRConfig, usage *common.ModelUsage) (*modelModule.ASRResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *modelModule.APIConfig, asrConfig *modelModule.ASRConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *mineruTestDriver) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *modelModule.APIConfig, asrConfig *modelModule.ASRConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) AudioSpeech(modelName *string, audioContent *string, apiConfig *modelModule.APIConfig, ttsConfig *modelModule.TTSConfig, usage *common.ModelUsage) (*modelModule.TTSResponse, error) {
+func (d *mineruTestDriver) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *modelModule.APIConfig, ttsConfig *modelModule.TTSConfig, usage *common.ModelUsage) (*modelModule.TTSResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *modelModule.APIConfig, ttsConfig *modelModule.TTSConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *mineruTestDriver) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *modelModule.APIConfig, ttsConfig *modelModule.TTSConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) OCRFile(modelName *string, content []byte, url *string, apiConfig *modelModule.APIConfig, ocrConfig *modelModule.OCRConfig, usage *common.ModelUsage) (*modelModule.OCRFileResponse, error) {
+func (d *mineruTestDriver) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *modelModule.APIConfig, ocrConfig *modelModule.OCRConfig, usage *common.ModelUsage) (*modelModule.OCRFileResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) ParseFile(modelName *string, content []byte, url *string, apiConfig *modelModule.APIConfig, parseFileConfig *modelModule.ParseFileConfig, usage *common.ModelUsage) (*modelModule.ParseFileResponse, error) {
+func (d *mineruTestDriver) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *modelModule.APIConfig, parseFileConfig *modelModule.ParseFileConfig, usage *common.ModelUsage) (*modelModule.ParseFileResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) ListModels(apiConfig *modelModule.APIConfig) ([]modelModule.ListModelResponse, error) {
+func (d *mineruTestDriver) ListModels(ctx context.Context, apiConfig *modelModule.APIConfig) ([]modelModule.ListModelResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) Balance(apiConfig *modelModule.APIConfig) (map[string]interface{}, error) {
+func (d *mineruTestDriver) Balance(ctx context.Context, apiConfig *modelModule.APIConfig) (map[string]interface{}, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) CheckConnection(apiConfig *modelModule.APIConfig) error {
+func (d *mineruTestDriver) CheckConnection(ctx context.Context, apiConfig *modelModule.APIConfig) error {
 	return fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) ListTasks(apiConfig *modelModule.APIConfig) ([]modelModule.ListTaskStatus, error) {
+func (d *mineruTestDriver) ListTasks(ctx context.Context, apiConfig *modelModule.APIConfig) ([]modelModule.ListTaskStatus, error) {
 	return nil, fmt.Errorf("not implemented")
 }
-func (d *mineruTestDriver) ShowTask(taskID string, apiConfig *modelModule.APIConfig) (*modelModule.TaskResponse, error) {
+func (d *mineruTestDriver) ShowTask(ctx context.Context, taskID string, apiConfig *modelModule.APIConfig) (*modelModule.TaskResponse, error) {
 	return nil, fmt.Errorf("not implemented")
 }
 
diff --git a/internal/ingestion/component/pdf_vision_dispatch.go b/internal/ingestion/component/pdf_vision_dispatch.go
index b47d1f8db8..8354a833c9 100644
--- a/internal/ingestion/component/pdf_vision_dispatch.go
+++ b/internal/ingestion/component/pdf_vision_dispatch.go
@@ -43,6 +43,7 @@ var (
 )
 
 func maybeDispatchPDFVision(
+	ctx context.Context,
 	fileType utility.FileType,
 	filename string,
 	binary []byte,
@@ -86,7 +87,7 @@ func maybeDispatchPDFVision(
 		return parserDispatchResult{}, true, fmt.Errorf(
 			`Parser: pdf parse_method %q requires tenant_id to resolve IMAGE2TEXT model`, modelID)
 	}
-	res, err := dispatchPDFVision(filename, binary, tenantID, modelID, setup)
+	res, err := dispatchPDFVision(ctx, filename, binary, tenantID, modelID, setup)
 	if err != nil {
 		return parserDispatchResult{}, true, err
 	}
@@ -373,6 +374,7 @@ func isNamedPDFParseMethod(raw string) bool {
 }
 
 func dispatchPDFVision(
+	ctx context.Context,
 	filename string,
 	binary []byte,
 	tenantID string,
@@ -396,7 +398,7 @@ func dispatchPDFVision(
 	markdownParts := make([]string, 0, len(renderedPages))
 	for _, page := range renderedPages {
 		prompt := renderPDFVisionPrompt(promptTemplate, page.PageNumber)
-		resp, err := pdfVisionChatInvoker(driver, resolvedModelName, buildPDFVisionMessages(prompt, page.ImageURL), apiConfig)
+		resp, err := pdfVisionChatInvoker(ctx, driver, resolvedModelName, buildPDFVisionMessages(prompt, page.ImageURL), apiConfig)
 		if err != nil {
 			return parserDispatchResult{}, fmt.Errorf("Parser: pdf vision page %d: %w", page.PageNumber, err)
 		}
@@ -472,13 +474,14 @@ func defaultPDFVisionModelResolver(
 }
 
 func defaultPDFVisionChatInvoker(
+	ctx context.Context,
 	driver modelModule.ModelDriver,
 	modelName string,
 	messages []modelModule.Message,
 	apiConfig *modelModule.APIConfig,
 ) (*modelModule.ChatResponse, error) {
 	vision := true
-	return driver.ChatWithMessages(modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
+	return driver.ChatWithMessages(ctx, modelName, messages, apiConfig, &modelModule.ChatConfig{Vision: &vision}, nil)
 }
 
 func loadPDFVisionPrompt(name string) (string, error) {
diff --git a/internal/ingestion/component/tokenizer.go b/internal/ingestion/component/tokenizer.go
index bc6e9488a3..dcae6b5ea9 100644
--- a/internal/ingestion/component/tokenizer.go
+++ b/internal/ingestion/component/tokenizer.go
@@ -153,7 +153,7 @@ type EmbeddingResult struct {
 // Embedder is the testability seam for the embedding branch.
 type Embedder interface {
 	MaxTokens() int
-	Encode(texts []string) ([]EmbeddingResult, error)
+	Encode(ctx context.Context, texts []string) ([]EmbeddingResult, error)
 }
 
 // EmbedderResolver resolves the embedder for one tokenizer invocation.
@@ -487,7 +487,7 @@ func encodeWithTimeout(ctx context.Context, embedder Embedder, texts []string) (
 		encErr  error
 	)
 	timeoutErr := runtime.WithTimeout(ctx, tokenizerTimeout(), func(timeoutCtx context.Context) error {
-		results, encErr = embedder.Encode(texts)
+		results, encErr = embedder.Encode(ctx, texts)
 		return encErr
 	})
 	if timeoutErr != nil {
diff --git a/internal/ingestion/component/tokenizer_unit_test.go b/internal/ingestion/component/tokenizer_unit_test.go
index d106979896..403799740f 100644
--- a/internal/ingestion/component/tokenizer_unit_test.go
+++ b/internal/ingestion/component/tokenizer_unit_test.go
@@ -53,7 +53,7 @@ func (s *stubEmbedder) MaxTokens() int {
 	return s.maxTokens
 }
 
-func (s *stubEmbedder) Encode(texts []string) ([]EmbeddingResult, error) {
+func (s *stubEmbedder) Encode(ctx context.Context, texts []string) ([]EmbeddingResult, error) {
 	s.calls.Add(1)
 	copied := append([]string(nil), texts...)
 	s.callInputs = append(s.callInputs, copied)
diff --git a/internal/ingestion/pipeline/template_integration_test.go b/internal/ingestion/pipeline/template_integration_test.go
index 7dc5215b7a..0b25a3d0a5 100644
--- a/internal/ingestion/pipeline/template_integration_test.go
+++ b/internal/ingestion/pipeline/template_integration_test.go
@@ -46,7 +46,7 @@ type fixedEmbedder struct{}
 
 func (fixedEmbedder) MaxTokens() int { return 0 }
 
-func (fixedEmbedder) Encode(texts []string) ([]componentpkg.EmbeddingResult, error) {
+func (fixedEmbedder) Encode(ctx context.Context, texts []string) ([]componentpkg.EmbeddingResult, error) {
 	out := make([]componentpkg.EmbeddingResult, 0, len(texts))
 	for _, text := range texts {
 		out = append(out, componentpkg.EmbeddingResult{
diff --git a/internal/ingestion/task/embedder.go b/internal/ingestion/task/embedder.go
index 99c9423049..80ca34f202 100644
--- a/internal/ingestion/task/embedder.go
+++ b/internal/ingestion/task/embedder.go
@@ -17,6 +17,7 @@
 package task
 
 import (
+	"context"
 	"fmt"
 	"strings"
 
@@ -37,12 +38,12 @@ func (e *embedder) MaxTokens() int {
 	return e.model.MaxTokens
 }
 
-func (e *embedder) Encode(texts []string) ([]componentpkg.EmbeddingResult, error) {
+func (e *embedder) Encode(ctx context.Context, texts []string) ([]componentpkg.EmbeddingResult, error) {
 	if e.model.ModelDriver == nil {
 		return nil, fmt.Errorf("embedder: embedding model driver is nil for model %v", e.model.ModelName)
 	}
 	config := &models.EmbeddingConfig{Dimension: 0}
-	embeds, err := e.model.ModelDriver.Embed(e.model.ModelName, texts, e.model.APIConfig, config, nil)
+	embeds, err := e.model.ModelDriver.Embed(ctx, e.model.ModelName, texts, e.model.APIConfig, config, nil)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/ingestion/task/embedder_test.go b/internal/ingestion/task/embedder_test.go
index 61618c9156..df2803e789 100644
--- a/internal/ingestion/task/embedder_test.go
+++ b/internal/ingestion/task/embedder_test.go
@@ -17,6 +17,7 @@
 package task
 
 import (
+	"context"
 	"ragflow/internal/common"
 	"testing"
 
@@ -74,7 +75,7 @@ type stubDriver struct {
 	capturedTexts []string
 }
 
-func (d *stubDriver) Embed(modelName *string, texts []string, apiConfig *models.APIConfig, embeddingConfig *models.EmbeddingConfig, usage *common.ModelUsage) ([]models.EmbeddingData, error) {
+func (d *stubDriver) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *models.APIConfig, embeddingConfig *models.EmbeddingConfig, usage *common.ModelUsage) ([]models.EmbeddingData, error) {
 	d.capturedTexts = texts
 	result := make([]models.EmbeddingData, len(texts))
 	for i := range texts {
@@ -88,46 +89,48 @@ func (d *stubDriver) Embed(modelName *string, texts []string, apiConfig *models.
 }
 func (d *stubDriver) NewInstance(baseURL map[string]string) models.ModelDriver { return d }
 func (d *stubDriver) Name() string                                             { return "stub" }
-func (d *stubDriver) ChatWithMessages(modelName string, messages []models.Message, apiConfig *models.APIConfig, chatModelConfig *models.ChatConfig, usage *common.ModelUsage) (*models.ChatResponse, error) {
+func (d *stubDriver) ChatWithMessages(ctx context.Context, modelName string, messages []models.Message, apiConfig *models.APIConfig, chatModelConfig *models.ChatConfig, usage *common.ModelUsage) (*models.ChatResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) ChatStreamlyWithSender(modelName string, messages []models.Message, apiConfig *models.APIConfig, modelConfig *models.ChatConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *stubDriver) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []models.Message, apiConfig *models.APIConfig, modelConfig *models.ChatConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return nil
 }
-func (d *stubDriver) Rerank(modelName *string, query string, documents []string, apiConfig *models.APIConfig, rerankConfig *models.RerankConfig, usage *common.ModelUsage) (*models.RerankResponse, error) {
+func (d *stubDriver) Rerank(ctx context.Context, modelName *string, query string, documents []string, apiConfig *models.APIConfig, rerankConfig *models.RerankConfig, usage *common.ModelUsage) (*models.RerankResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) TranscribeAudio(modelName *string, file *string, apiConfig *models.APIConfig, asrConfig *models.ASRConfig, usage *common.ModelUsage) (*models.ASRResponse, error) {
+func (d *stubDriver) TranscribeAudio(ctx context.Context, modelName *string, file *string, apiConfig *models.APIConfig, asrConfig *models.ASRConfig, usage *common.ModelUsage) (*models.ASRResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *models.APIConfig, asrConfig *models.ASRConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *stubDriver) TranscribeAudioWithSender(ctx context.Context, modelName *string, file *string, apiConfig *models.APIConfig, asrConfig *models.ASRConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return nil
 }
-func (d *stubDriver) AudioSpeech(modelName *string, audioContent *string, apiConfig *models.APIConfig, ttsConfig *models.TTSConfig, usage *common.ModelUsage) (*models.TTSResponse, error) {
+func (d *stubDriver) AudioSpeech(ctx context.Context, modelName *string, audioContent *string, apiConfig *models.APIConfig, ttsConfig *models.TTSConfig, usage *common.ModelUsage) (*models.TTSResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *models.APIConfig, ttsConfig *models.TTSConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *stubDriver) AudioSpeechWithSender(ctx context.Context, modelName *string, audioContent *string, apiConfig *models.APIConfig, ttsConfig *models.TTSConfig, usage *common.ModelUsage, sender func(*string, *string) error) error {
 	return nil
 }
-func (d *stubDriver) OCRFile(modelName *string, content []byte, url *string, apiConfig *models.APIConfig, ocrConfig *models.OCRConfig, usage *common.ModelUsage) (*models.OCRFileResponse, error) {
+func (d *stubDriver) OCRFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *models.APIConfig, ocrConfig *models.OCRConfig, usage *common.ModelUsage) (*models.OCRFileResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) ParseFile(modelName *string, content []byte, url *string, apiConfig *models.APIConfig, parseFileConfig *models.ParseFileConfig, usage *common.ModelUsage) (*models.ParseFileResponse, error) {
+func (d *stubDriver) ParseFile(ctx context.Context, modelName *string, content []byte, url *string, apiConfig *models.APIConfig, parseFileConfig *models.ParseFileConfig, usage *common.ModelUsage) (*models.ParseFileResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) ListModels(apiConfig *models.APIConfig) ([]models.ListModelResponse, error) {
+func (d *stubDriver) ListModels(ctx context.Context, apiConfig *models.APIConfig) ([]models.ListModelResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) Balance(apiConfig *models.APIConfig) (map[string]interface{}, error) {
+func (d *stubDriver) Balance(ctx context.Context, apiConfig *models.APIConfig) (map[string]interface{}, error) {
 	return nil, nil
 }
-func (d *stubDriver) CheckConnection(apiConfig *models.APIConfig) error { return nil }
-func (d *stubDriver) ListTasks(apiConfig *models.APIConfig) ([]models.ListTaskStatus, error) {
+func (d *stubDriver) CheckConnection(ctx context.Context, apiConfig *models.APIConfig) error {
+	return nil
+}
+func (d *stubDriver) ListTasks(ctx context.Context, apiConfig *models.APIConfig) ([]models.ListTaskStatus, error) {
 	return nil, nil
 }
-func (d *stubDriver) ShowTask(taskID string, apiConfig *models.APIConfig) (*models.TaskResponse, error) {
+func (d *stubDriver) ShowTask(ctx context.Context, taskID string, apiConfig *models.APIConfig) (*models.TaskResponse, error) {
 	return nil, nil
 }
-func (d *stubDriver) ToolCall(name string, arguments map[string]interface{}) (string, error) {
+func (d *stubDriver) ToolCall(ctx context.Context, name string, arguments map[string]interface{}) (string, error) {
 	return "", nil
 }
diff --git a/internal/parser/parser/audio_parser.go b/internal/parser/parser/audio_parser.go
index 495c9f1b99..fab1c63cf3 100644
--- a/internal/parser/parser/audio_parser.go
+++ b/internal/parser/parser/audio_parser.go
@@ -22,6 +22,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 	"path/filepath"
 	"strings"
@@ -70,7 +71,7 @@ func (p *AudioParser) ConfigureFromSetup(setup map[string]any) {
 // file extension against the audio extension whitelist. The actual
 // speech-to-text transcription happens via maybeDispatchAudio at the
 // component layer (mirrors Python's LLMBundle.transcription call).
-func (p *AudioParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *AudioParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	ext := strings.ToLower(filepath.Ext(filename))
 	if len(ext) > 1 && ext[0] == '.' {
 		ext = ext[1:]
diff --git a/internal/parser/parser/audio_parser_test.go b/internal/parser/parser/audio_parser_test.go
index 0910b9ad3a..19f9dbdba4 100644
--- a/internal/parser/parser/audio_parser_test.go
+++ b/internal/parser/parser/audio_parser_test.go
@@ -22,6 +22,7 @@ import (
 )
 
 func TestAudioParser_ValidExtension(t *testing.T) {
+	ctx := t.Context()
 	p := NewAudioParser()
 	p.ConfigureFromSetup(map[string]any{
 		"output_format": "text",
@@ -30,7 +31,7 @@ func TestAudioParser_ValidExtension(t *testing.T) {
 		},
 	})
 
-	result := p.ParseWithResult("test.mp3", []byte{1, 2, 3, 4})
+	result := p.ParseWithResult(ctx, "test.mp3", []byte{1, 2, 3, 4})
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
@@ -43,9 +44,10 @@ func TestAudioParser_ValidExtension(t *testing.T) {
 }
 
 func TestAudioParser_AllExtensions(t *testing.T) {
+	ctx := t.Context()
 	for _, ext := range []string{"da", "wave", "wav", "mp3", "aac", "flac", "ogg", "aiff", "au", "midi", "wma", "realaudio", "vqf", "oggvorbis", "ape"} {
 		p := NewAudioParser()
-		result := p.ParseWithResult("audio."+ext, []byte{})
+		result := p.ParseWithResult(ctx, "audio."+ext, []byte{})
 		if result.Err != nil {
 			t.Errorf("extension .%s rejected: %v", ext, result.Err)
 		}
@@ -53,8 +55,9 @@ func TestAudioParser_AllExtensions(t *testing.T) {
 }
 
 func TestAudioParser_InvalidExtension(t *testing.T) {
+	ctx := t.Context()
 	p := NewAudioParser()
-	result := p.ParseWithResult("notaudio.txt", []byte{})
+	result := p.ParseWithResult(ctx, "notaudio.txt", []byte{})
 	if result.Err == nil {
 		t.Fatal("expected error for .txt extension")
 	}
@@ -76,19 +79,21 @@ func TestAudioParser_ConfigureVLM(t *testing.T) {
 }
 
 func TestAudioParser_ConfigureNilSafe(t *testing.T) {
+	ctx := t.Context()
 	p := NewAudioParser()
 	p.ConfigureFromSetup(nil)
 	p.ConfigureFromSetup(map[string]any{})
 	// Should not panic.
-	result := p.ParseWithResult("test.wav", []byte{})
+	result := p.ParseWithResult(ctx, "test.wav", []byte{})
 	if result.Err != nil {
 		t.Errorf("unexpected error: %v", result.Err)
 	}
 }
 
 func TestAudioParser_DefaultOutputFormat(t *testing.T) {
+	ctx := t.Context()
 	p := NewAudioParser()
-	result := p.ParseWithResult("test.flac", []byte{})
+	result := p.ParseWithResult(ctx, "test.flac", []byte{})
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
diff --git a/internal/parser/parser/csv_parser.go b/internal/parser/parser/csv_parser.go
index aa527fc5ec..c37e5d88f6 100644
--- a/internal/parser/parser/csv_parser.go
+++ b/internal/parser/parser/csv_parser.go
@@ -31,6 +31,7 @@
 package parser
 
 import (
+	"context"
 	"encoding/csv"
 	"fmt"
 	"html"
@@ -113,7 +114,7 @@ func (p *CSVParser) ConfigureFromSetup(setup map[string]any) {
 // Python's RAGFlowExcelParser.html().
 // When TCADP parse_method is configured, the file is dispatched to
 // the Tencent Cloud Document Parsing API.
-func (p *CSVParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *CSVParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	method := normalizeXLSXParseMethod(p.ParseMethod)
 	switch method {
 	case "tcadp":
diff --git a/internal/parser/parser/doc_parser.go b/internal/parser/parser/doc_parser.go
index f0ef78435b..2aabc61b6a 100644
--- a/internal/parser/parser/doc_parser.go
+++ b/internal/parser/parser/doc_parser.go
@@ -19,6 +19,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 
 	officeOxide "github.com/yfedoseev/office_oxide/go"
@@ -39,7 +40,7 @@ func (p *DOCParser) String() string {
 // the Go side uses office_oxide which supports DOC via PlainText.
 // OutputFormat="text" — the python side falls back to text for
 // legacy DOC files since structured extraction is unreliable.
-func (p *DOCParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *DOCParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	doc, err := officeOxide.OpenFromBytes(data, "doc")
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("doc open: %w", err)}
diff --git a/internal/parser/parser/docx_parser.go b/internal/parser/parser/docx_parser.go
index 11781f729a..7e3619ebea 100644
--- a/internal/parser/parser/docx_parser.go
+++ b/internal/parser/parser/docx_parser.go
@@ -19,6 +19,7 @@
 package parser
 
 import (
+	"context"
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
@@ -66,7 +67,7 @@ func (p *DOCXParser) ConfigureFromSetup(setup map[string]any) {
 //
 // JSON path mirrors python parser.py:_docx() output_format == "json".
 // Markdown path mirrors python naive.py: Docx() → naive_merge_docx().
-func (p *DOCXParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *DOCXParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	doc, err := officeOxide.OpenFromBytes(data, "docx")
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("docx open: %w", err)}
diff --git a/internal/parser/parser/docx_parser_cgo_test.go b/internal/parser/parser/docx_parser_cgo_test.go
index 8eeb0cfca6..65ff846f7f 100644
--- a/internal/parser/parser/docx_parser_cgo_test.go
+++ b/internal/parser/parser/docx_parser_cgo_test.go
@@ -9,10 +9,11 @@ import (
 )
 
 func TestDOCXParser_ParseWithResult_JSON(t *testing.T) {
+	ctx := t.Context()
 	p := NewDOCXParser()
 	p.outputFormat = "json"
 	data := minimalDOCX(t, "Hello from JSON path")
-	res := p.ParseWithResult("sample.docx", data)
+	res := p.ParseWithResult(ctx, "sample.docx", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -33,6 +34,7 @@ func TestDOCXParser_ParseWithResult_JSON(t *testing.T) {
 }
 
 func TestDOCXParser_ConfigureFromSetup_JSON(t *testing.T) {
+	ctx := t.Context()
 	p := NewDOCXParser()
 	p.ConfigureFromSetup(map[string]any{"output_format": "json"})
 	if p.outputFormat != "json" {
@@ -43,7 +45,7 @@ func TestDOCXParser_ConfigureFromSetup_JSON(t *testing.T) {
 	}
 	// Full round-trip: json config → json output
 	data := minimalDOCX(t, "Config test")
-	res := p.ParseWithResult("sample.docx", data)
+	res := p.ParseWithResult(ctx, "sample.docx", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -56,13 +58,14 @@ func TestDOCXParser_ConfigureFromSetup_JSON(t *testing.T) {
 }
 
 func TestDOCXParser_ConfigureFromSetup_Markdown(t *testing.T) {
+	ctx := t.Context()
 	p := NewDOCXParser()
 	p.ConfigureFromSetup(map[string]any{"output_format": "markdown"})
 	if p.outputFormat != "markdown" {
 		t.Fatalf("After ConfigureFromSetup, outputFormat = %q, want %q", p.outputFormat, "markdown")
 	}
 	data := minimalDOCX(t, "Config md test")
-	res := p.ParseWithResult("sample.docx", data)
+	res := p.ParseWithResult(ctx, "sample.docx", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -75,9 +78,10 @@ func TestDOCXParser_ConfigureFromSetup_Markdown(t *testing.T) {
 }
 
 func TestDOCXParser_ParseWithResult_CGOMinimalDocument(t *testing.T) {
+	ctx := t.Context()
 	p := NewDOCXParser()
 	data := minimalDOCX(t, "Hello from DOCX parser")
-	res := p.ParseWithResult("sample.docx", data)
+	res := p.ParseWithResult(ctx, "sample.docx", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/email_parser.go b/internal/parser/parser/email_parser.go
index 2aea280337..6952225710 100644
--- a/internal/parser/parser/email_parser.go
+++ b/internal/parser/parser/email_parser.go
@@ -18,6 +18,7 @@ package parser
 
 import (
 	"bytes"
+	"context"
 	"encoding/base64"
 	"fmt"
 	"io"
@@ -73,7 +74,7 @@ func (p *EmailParser) ConfigureFromSetup(setup map[string]any) {
 	}
 }
 
-func (p *EmailParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *EmailParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	ext := strings.ToLower(filepath.Ext(filename))
 	if ext == ".msg" {
 		return ParseResult{
diff --git a/internal/parser/parser/email_parser_test.go b/internal/parser/parser/email_parser_test.go
index 7ad32196f9..610f9fc3ba 100644
--- a/internal/parser/parser/email_parser_test.go
+++ b/internal/parser/parser/email_parser_test.go
@@ -23,6 +23,7 @@ import (
 )
 
 func TestEmailParser_EmlJSON(t *testing.T) {
+	ctx := t.Context()
 	raw := strings.Join([]string{
 		"From: sender@example.com",
 		"To: recipient@example.com",
@@ -41,7 +42,7 @@ func TestEmailParser_EmlJSON(t *testing.T) {
 		"fields":        []string{"from", "to", "cc", "date", "subject", "body", "metadata"},
 	})
 
-	result := p.ParseWithResult("test.eml", []byte(raw))
+	result := p.ParseWithResult(ctx, "test.eml", []byte(raw))
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
@@ -78,6 +79,7 @@ func TestEmailParser_EmlJSON(t *testing.T) {
 }
 
 func TestEmailParser_EmlText(t *testing.T) {
+	ctx := t.Context()
 	raw := strings.Join([]string{
 		"From: sender@test.com",
 		"To: recipient@test.com",
@@ -93,7 +95,7 @@ func TestEmailParser_EmlText(t *testing.T) {
 		"fields":        []string{"from", "to", "subject", "body"},
 	})
 
-	result := p.ParseWithResult("test.eml", []byte(raw))
+	result := p.ParseWithResult(ctx, "test.eml", []byte(raw))
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
@@ -109,8 +111,9 @@ func TestEmailParser_EmlText(t *testing.T) {
 }
 
 func TestEmailParser_MsgNotSupported(t *testing.T) {
+	ctx := t.Context()
 	p := NewEmailParser()
-	result := p.ParseWithResult("test.msg", []byte{})
+	result := p.ParseWithResult(ctx, "test.msg", []byte{})
 	if result.Err == nil {
 		t.Fatal("expected error for .msg file")
 	}
@@ -120,6 +123,7 @@ func TestEmailParser_MsgNotSupported(t *testing.T) {
 }
 
 func TestEmailParser_Base64Attachment(t *testing.T) {
+	ctx := t.Context()
 	attachmentContent := "Hello! This is the decoded content of the attachment."
 	encoded := base64.StdEncoding.EncodeToString([]byte(attachmentContent))
 	// Simulate MIME line-wrapping (typically 76 chars per line).
@@ -153,7 +157,7 @@ func TestEmailParser_Base64Attachment(t *testing.T) {
 		"fields":        []string{"from", "body", "attachments"},
 	})
 
-	result := p.ParseWithResult("test.eml", []byte(raw))
+	result := p.ParseWithResult(ctx, "test.eml", []byte(raw))
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
@@ -179,6 +183,7 @@ func TestEmailParser_Base64Attachment(t *testing.T) {
 }
 
 func TestEmailParser_Base64AttachmentInMixedMultipart(t *testing.T) {
+	ctx := t.Context()
 	// Simulates the original test email structure:
 	// multipart/mixed → multipart/alternative (text/plain + text/html) + base64 attachment
 	innerBoundary := "inneralt"
@@ -223,7 +228,7 @@ func TestEmailParser_Base64AttachmentInMixedMultipart(t *testing.T) {
 		"fields":        []string{"from", "body", "attachments"},
 	})
 
-	result := p.ParseWithResult("test.eml", []byte(raw))
+	result := p.ParseWithResult(ctx, "test.eml", []byte(raw))
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
@@ -249,6 +254,7 @@ func TestEmailParser_Base64AttachmentInMixedMultipart(t *testing.T) {
 }
 
 func TestEmailParser_Multipart(t *testing.T) {
+	ctx := t.Context()
 	boundary := "boundary123"
 	raw := strings.Join([]string{
 		"From: multipart@test.com",
@@ -273,7 +279,7 @@ func TestEmailParser_Multipart(t *testing.T) {
 		"fields":        []string{"from", "body"},
 	})
 
-	result := p.ParseWithResult("test.eml", []byte(raw))
+	result := p.ParseWithResult(ctx, "test.eml", []byte(raw))
 	if result.Err != nil {
 		t.Fatalf("unexpected error: %v", result.Err)
 	}
diff --git a/internal/parser/parser/epub_parser.go b/internal/parser/parser/epub_parser.go
index 081c1a81c4..29d2c5eebe 100644
--- a/internal/parser/parser/epub_parser.go
+++ b/internal/parser/parser/epub_parser.go
@@ -31,6 +31,7 @@ package parser
 import (
 	"archive/zip"
 	"bytes"
+	"context"
 	"encoding/xml"
 	"fmt"
 	"io"
@@ -53,7 +54,7 @@ func (p *EPUBParser) String() string {
 // ParseWithResult implements ParseResultProducer. It extracts XHTML
 // content from the EPUB spine and emits one JSON item per spine entry
 // with {text, doc_type_kwd:"text"}.
-func (p *EPUBParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *EPUBParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	if len(data) == 0 {
 		return ParseResult{
 			OutputFormat: "json",
diff --git a/internal/parser/parser/html_parser.go b/internal/parser/parser/html_parser.go
index b73020ed0f..315195838e 100644
--- a/internal/parser/parser/html_parser.go
+++ b/internal/parser/parser/html_parser.go
@@ -18,6 +18,7 @@ package parser
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"strings"
 
@@ -47,7 +48,7 @@ func (p *HTMLParser) String() string {
 // (bold / links / images) is intentionally NOT surfaced as a
 // separate ck_type — the python HtmlParser collapses inline
 // formatting into the parent block's text.
-func (p *HTMLParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *HTMLParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	doc, err := html.Parse(bytes.NewReader(data))
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("html parse: %w", err)}
diff --git a/internal/parser/parser/json_parser.go b/internal/parser/parser/json_parser.go
index fe03f9ef7b..9f0481aa9c 100644
--- a/internal/parser/parser/json_parser.go
+++ b/internal/parser/parser/json_parser.go
@@ -32,6 +32,7 @@ package parser
 import (
 	"bufio"
 	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"strings"
@@ -51,7 +52,7 @@ func (p *JSONParser) String() string {
 // ParseWithResult implements ParseResultProducer. It detects the JSON
 // shape (array, single object, or line-delimited) and emits one JSON
 // item per logical record, with {text, doc_type_kwd:"text"}.
-func (p *JSONParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *JSONParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	text := string(bytes.TrimSpace(data))
 	if text == "" {
 		return ParseResult{
diff --git a/internal/parser/parser/markdown_parser.go b/internal/parser/parser/markdown_parser.go
index 482c136b8f..55ce6b5b51 100644
--- a/internal/parser/parser/markdown_parser.go
+++ b/internal/parser/parser/markdown_parser.go
@@ -89,7 +89,7 @@ func (p *MarkdownParser) ConfigureFromSetup(setup map[string]any) {
 // data is resolved and the item carries `doc_type_kwd: "image"` with
 // the base64-encoded image payload. The legacy debug-print path has
 // been removed; callers consume ParseResult directly.
-func (p *MarkdownParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *MarkdownParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	doc := markdownNew().Parse(data)
 	rawText := string(data)
 
diff --git a/internal/parser/parser/markdown_parser_test.go b/internal/parser/parser/markdown_parser_test.go
index d0db085a97..b0c18c71ec 100644
--- a/internal/parser/parser/markdown_parser_test.go
+++ b/internal/parser/parser/markdown_parser_test.go
@@ -8,12 +8,13 @@ import (
 )
 
 func TestMarkdownParser_ParseWithResult_Basic(t *testing.T) {
+	ctx := t.Context()
 	p, err := NewMarkdownParser(GoMarkdown)
 	if err != nil {
 		t.Fatalf("NewMarkdownParser: %v", err)
 	}
 	md := "# Hello\n\nThis is a paragraph.\n\n* List item 1\n* List item 2\n\n```go\nfunc main() {}\n```\n"
-	res := p.ParseWithResult("test.md", []byte(md))
+	res := p.ParseWithResult(ctx, "test.md", []byte(md))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -33,8 +34,9 @@ func TestMarkdownParser_ParseWithResult_Basic(t *testing.T) {
 }
 
 func TestMarkdownParser_ParseWithResult_EmptyInput(t *testing.T) {
+	ctx := t.Context()
 	p, _ := NewMarkdownParser(GoMarkdown)
-	res := p.ParseWithResult("empty.md", []byte(""))
+	res := p.ParseWithResult(ctx, "empty.md", []byte(""))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -44,13 +46,14 @@ func TestMarkdownParser_ParseWithResult_EmptyInput(t *testing.T) {
 }
 
 func TestMarkdownParser_ParseWithResult_ImageDataURI(t *testing.T) {
+	ctx := t.Context()
 	p, _ := NewMarkdownParser(GoMarkdown)
 	// 1×1 pixel transparent PNG encoded as data URI
 	pixelPNG := make([]byte, 68) // minimal 1x1 PNG header
 	pixelB64 := base64.StdEncoding.EncodeToString([]byte("fake-png-data"))
 	md := "Some text with an image\n![test](data:image/png;base64," + pixelB64 + ")\n"
 	_ = pixelPNG
-	res := p.ParseWithResult("test.md", []byte(md))
+	res := p.ParseWithResult(ctx, "test.md", []byte(md))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -71,9 +74,10 @@ func TestMarkdownParser_ParseWithResult_ImageDataURI(t *testing.T) {
 }
 
 func TestMarkdownParser_ParseWithResult_NoImage(t *testing.T) {
+	ctx := t.Context()
 	p, _ := NewMarkdownParser(GoMarkdown)
 	md := "# Title\n\nJust some text, no images here.\n\nMore text."
-	res := p.ParseWithResult("test.md", []byte(md))
+	res := p.ParseWithResult(ctx, "test.md", []byte(md))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/parse_result.go b/internal/parser/parser/parse_result.go
index 21a3c019ea..6dc642cdf8 100644
--- a/internal/parser/parser/parse_result.go
+++ b/internal/parser/parser/parse_result.go
@@ -31,6 +31,8 @@
 
 package parser
 
+import "context"
+
 // ParseResult is the structured return value of a successful parse.
 // Exactly one of the payload fields (JSON / Markdown / Text / HTML)
 // is populated on success, matching the Python contract — see
@@ -84,5 +86,5 @@ type ParseResult struct {
 // ParseResultProducer is the parser package's single structured-output
 // contract. Every parser returned by GetParser must implement it.
 type ParseResultProducer interface {
-	ParseWithResult(filename string, data []byte) ParseResult
+	ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult
 }
diff --git a/internal/parser/parser/parse_result_test.go b/internal/parser/parser/parse_result_test.go
index 36d39adf34..ffb3707755 100644
--- a/internal/parser/parser/parse_result_test.go
+++ b/internal/parser/parser/parse_result_test.go
@@ -130,9 +130,10 @@ func (s sentinelErr) Error() string { return string(s) }
 // tiny: 1 heading, 1 paragraph, 1 unordered list item, no nested
 // formatting.
 func TestMarkdownParser_ParseWithResult(t *testing.T) {
+	ctx := t.Context()
 	p, _ := NewMarkdownParser(GoMarkdown)
 	src := []byte("# Title\n\nFirst paragraph.\n\n- Item one\n")
-	res := p.ParseWithResult("doc.md", src)
+	res := p.ParseWithResult(ctx, "doc.md", src)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/parse_with_result_test.go b/internal/parser/parser/parse_with_result_test.go
index d38670f83c..2c523a6361 100644
--- a/internal/parser/parser/parse_with_result_test.go
+++ b/internal/parser/parser/parse_with_result_test.go
@@ -45,7 +45,8 @@ import (
 func TestTextParser_ParseWithResult_ParaSplit(t *testing.T) {
 	p := NewTextParser()
 	src := []byte("First paragraph.\n\nSecond paragraph.\n\nThird.")
-	res := p.ParseWithResult("doc.txt", src)
+	ctx := t.Context()
+	res := p.ParseWithResult(ctx, "doc.txt", src)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -71,8 +72,9 @@ func TestTextParser_ParseWithResult_ParaSplit(t *testing.T) {
 // sees a non-nil JSON slice. Mirrors the MarkdownParser convention
 // at markdown_parser.go:71-76.
 func TestTextParser_ParseWithResult_Empty(t *testing.T) {
+	ctx := t.Context()
 	p := NewTextParser()
-	res := p.ParseWithResult("empty.txt", []byte{})
+	res := p.ParseWithResult(ctx, "empty.txt", []byte{})
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -85,9 +87,10 @@ func TestTextParser_ParseWithResult_Empty(t *testing.T) {
 // maxItemBytes boundary behaviour. A single paragraph longer
 // than 8192 bytes is sliced at the nearest line boundary.
 func TestTextParser_ParseWithResult_LongParagraphSlicing(t *testing.T) {
+	ctx := t.Context()
 	p := NewTextParser()
 	long := strings.Repeat("a", 9000)
-	res := p.ParseWithResult("long.txt", []byte(long))
+	res := p.ParseWithResult(ctx, "long.txt", []byte(long))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -105,9 +108,10 @@ func TestTextParser_ParseWithResult_LongParagraphSlicing(t *testing.T) {
 // validation rule. Invalid bytes produce an error in the result
 // (matching the python TxtParser's behaviour).
 func TestTextParser_ParseWithResult_InvalidUTF8(t *testing.T) {
+	ctx := t.Context()
 	p := NewTextParser()
 	bad := []byte{0xff, 0xfe, 0xfd}
-	res := p.ParseWithResult("bad.txt", bad)
+	res := p.ParseWithResult(ctx, "bad.txt", bad)
 	if res.Err == nil {
 		t.Fatal("want error for invalid UTF-8, got nil")
 	}
@@ -117,13 +121,14 @@ func TestTextParser_ParseWithResult_InvalidUTF8(t *testing.T) {
 // Three block elements (heading, paragraph, list) yield three
 // items with the python-compatible ck_type vocabulary.
 func TestHTMLParser_ParseWithResult_BlockSplit(t *testing.T) {
+	ctx := t.Context()
 	p := NewHTMLParser()
 	src := []byte(`<!DOCTYPE html><html><body>
 <h1>Title</h1>
 <p>First paragraph.</p>
 <ul><li>Item one</li></ul>
 </body></html>`)
-	res := p.ParseWithResult("doc.html", src)
+	res := p.ParseWithResult(ctx, "doc.html", src)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -157,6 +162,7 @@ func TestHTMLParser_ParseWithResult_BlockSplit(t *testing.T) {
 // rule that <script> / <style> subtrees are skipped entirely so
 // they don't pollute the downstream chunker input.
 func TestHTMLParser_ParseWithResult_SkipsScriptAndStyle(t *testing.T) {
+	ctx := t.Context()
 	p := NewHTMLParser()
 	src := []byte(`<html><body>
 <p>Visible.</p>
@@ -164,7 +170,7 @@ func TestHTMLParser_ParseWithResult_SkipsScriptAndStyle(t *testing.T) {
 <style>body { color: red; }</style>
 <p>Also visible.</p>
 </body></html>`)
-	res := p.ParseWithResult("doc.html", src)
+	res := p.ParseWithResult(ctx, "doc.html", src)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/pdf_parser_cgo.go b/internal/parser/parser/pdf_parser_cgo.go
index 3140340262..17854e430d 100644
--- a/internal/parser/parser/pdf_parser_cgo.go
+++ b/internal/parser/parser/pdf_parser_cgo.go
@@ -12,7 +12,7 @@ import (
 	deepdoctype "ragflow/internal/deepdoc/parser/type"
 )
 
-func (p *PDFParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *PDFParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	if err := p.validateParseMethod(); err != nil {
 		return ParseResult{Err: err}
 	}
@@ -21,13 +21,13 @@ func (p *PDFParser) ParseWithResult(filename string, data []byte) ParseResult {
 	case "plain_text":
 		return parsePDFWithPlainText(filename, data, p)
 	case "mineru":
-		return parsePDFWithMinerU(filename, data, p)
+		return parsePDFWithMinerU(ctx, filename, data, p)
 	case "paddleocr":
-		return parsePDFWithPaddleOCR(filename, data, p)
+		return parsePDFWithPaddleOCR(ctx, filename, data, p)
 	case "docling":
-		return parsePDFWithDocling(filename, data, p)
+		return parsePDFWithDocling(ctx, filename, data, p)
 	case "opendataloader":
-		return parsePDFWithOpenDataLoader(filename, data, p)
+		return parsePDFWithOpenDataLoader(ctx, filename, data, p)
 	case "somark":
 		return parsePDFWithSoMark(filename, data, p)
 	case "tcadp":
@@ -36,7 +36,7 @@ func (p *PDFParser) ParseWithResult(filename string, data []byte) ParseResult {
 	cfg := deepdoctype.DefaultParserConfig()
 	cfg.SkipOCR = false
 	parser := deepdocpdf.NewParser(cfg)
-	res := parsePDFWithDeepDocOptions(context.Background(), filename, data, pdfPostProcessOptions{
+	res := parsePDFWithDeepDocOptions(ctx, filename, data, pdfPostProcessOptions{
 		outputFormat:       p.OutputFormat,
 		zoom:               cfg.Zoom,
 		enableMultiColumn:  p.EnableMultiColumn,
diff --git a/internal/parser/parser/pdf_parser_cgo_test.go b/internal/parser/parser/pdf_parser_cgo_test.go
index 1e76b133ad..4912ad3798 100644
--- a/internal/parser/parser/pdf_parser_cgo_test.go
+++ b/internal/parser/parser/pdf_parser_cgo_test.go
@@ -8,19 +8,21 @@ import (
 )
 
 func TestPDFParser_ParseWithResult_CGOInvalidPDF(t *testing.T) {
+	ctx := t.Context()
 	t.Setenv("DEEPDOC_URL", "")
 	t.Setenv("OSSDEEPDOC_URL", "")
 
 	pdf := NewPDFParser()
-	res := pdf.ParseWithResult("bad.pdf", []byte("not a valid pdf"))
+	res := pdf.ParseWithResult(ctx, "bad.pdf", []byte("not a valid pdf"))
 	if res.Err == nil {
 		t.Fatal("want parse error for invalid PDF bytes, got nil")
 	}
 }
 
 func TestPDFParser_ParseWithResult_CGOEmpty(t *testing.T) {
+	ctx := t.Context()
 	pdf := NewPDFParser()
-	res := pdf.ParseWithResult("empty.pdf", nil)
+	res := pdf.ParseWithResult(ctx, "empty.pdf", nil)
 	if res.Err != nil {
 		t.Fatalf("empty input: want nil err, got %v", res.Err)
 	}
diff --git a/internal/parser/parser/pdf_parser_docling.go b/internal/parser/parser/pdf_parser_docling.go
index 5665a8e071..9d75aee3c2 100644
--- a/internal/parser/parser/pdf_parser_docling.go
+++ b/internal/parser/parser/pdf_parser_docling.go
@@ -36,7 +36,7 @@ type doclingResponse struct {
 	Results   []doclingResult   `json:"results"`
 }
 
-func parsePDFWithDocling(filename string, data []byte, parser *PDFParser) ParseResult {
+func parsePDFWithDocling(ctx context.Context, filename string, data []byte, parser *PDFParser) ParseResult {
 	if len(data) == 0 {
 		return emptyPDFResult(filename)
 	}
@@ -116,7 +116,7 @@ func parsePDFWithDocling(filename string, data []byte, parser *PDFParser) ParseR
 			lastErr = fmt.Errorf("%s: chunked response contained no usable text", candidate.endpoint)
 			continue
 		}
-		if res, ok := parseDoclingStandardResult(filename, body, parser.OutputFormat); ok {
+		if res, ok := parseDoclingStandardResult(ctx, filename, body, parser.OutputFormat); ok {
 			return res
 		}
 		lastErr = fmt.Errorf("%s: standard response contained no parsed document", candidate.endpoint)
@@ -189,7 +189,7 @@ func parseDoclingChunkedResult(filename string, body []byte, outputFormat string
 	return doclingTextsToResult(filename, texts, outputFormat, pageCount), true
 }
 
-func parseDoclingStandardResult(filename string, body []byte, outputFormat string) (ParseResult, bool) {
+func parseDoclingStandardResult(ctx context.Context, filename string, body []byte, outputFormat string) (ParseResult, bool) {
 	var payload doclingResponse
 	if err := json.Unmarshal(body, &payload); err != nil {
 		return ParseResult{}, false
@@ -209,13 +209,13 @@ func parseDoclingStandardResult(filename string, body []byte, outputFormat strin
 	}
 	for _, doc := range docs {
 		if md := strings.TrimSpace(doc.MDContent); md != "" {
-			return parseMinerUMarkdownResult(filename, md, outputFormat, len(docs)), true
+			return parseMinerUMarkdownResult(ctx, filename, md, outputFormat, len(docs)), true
 		}
 		if txt := strings.TrimSpace(doc.TextContent); txt != "" {
 			return doclingTextsToResult(filename, []string{txt}, outputFormat, len(docs)), true
 		}
 		if md, _ := doc.JSONContent["md_content"].(string); strings.TrimSpace(md) != "" {
-			return parseMinerUMarkdownResult(filename, md, outputFormat, len(docs)), true
+			return parseMinerUMarkdownResult(ctx, filename, md, outputFormat, len(docs)), true
 		}
 	}
 	return ParseResult{}, false
diff --git a/internal/parser/parser/pdf_parser_docling_test.go b/internal/parser/parser/pdf_parser_docling_test.go
index 29e400a7a1..43ed17ef53 100644
--- a/internal/parser/parser/pdf_parser_docling_test.go
+++ b/internal/parser/parser/pdf_parser_docling_test.go
@@ -70,7 +70,8 @@ func TestPDFParser_ParseWithResult_DoclingChunkedMarkdownIntegration(t *testing.
 		"docling_api_key":    "doc-secret",
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -132,7 +133,8 @@ func TestPDFParser_ParseWithResult_DoclingFallbackToStandardJSONIntegration(t *t
 		"docling_server_url": server.URL,
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -151,7 +153,8 @@ func TestPDFParser_ParseWithResult_DoclingRequiresServerURL(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "Docling"})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil {
 		t.Fatal("ParseWithResult: want error when docling_server_url is missing, got nil")
 	}
diff --git a/internal/parser/parser/pdf_parser_fixture_cgo_test.go b/internal/parser/parser/pdf_parser_fixture_cgo_test.go
index 3df1cd1e50..c3ea98c23b 100644
--- a/internal/parser/parser/pdf_parser_fixture_cgo_test.go
+++ b/internal/parser/parser/pdf_parser_fixture_cgo_test.go
@@ -19,7 +19,8 @@ func TestPDFParser_ParseWithResult_CGOFixture(t *testing.T) {
 		t.Fatalf("ReadFile(%s): %v", path, err)
 	}
 	pdf := NewPDFParser()
-	res := pdf.ParseWithResult("Doc1.pdf", data)
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "Doc1.pdf", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -49,7 +50,8 @@ func TestPDFParser_ParseWithResult_CGOFixtureMarkdown(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"output_format": "markdown"})
 
-	res := pdf.ParseWithResult("Doc1.pdf", data)
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "Doc1.pdf", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -72,7 +74,8 @@ func TestPDFParser_ParseWithResult_CGOFixturePlainText(t *testing.T) {
 	}
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "plain_text"})
-	res := pdf.ParseWithResult("Doc1.pdf", data)
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "Doc1.pdf", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/pdf_parser_mineru.go b/internal/parser/parser/pdf_parser_mineru.go
index dcf2ff76f6..697946a72a 100644
--- a/internal/parser/parser/pdf_parser_mineru.go
+++ b/internal/parser/parser/pdf_parser_mineru.go
@@ -1,6 +1,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 	"ragflow/internal/common"
 	"strings"
@@ -12,7 +13,7 @@ import (
 const minerUPollTimeout = 30 * time.Second
 const minerUPollInterval = 200 * time.Millisecond
 
-func parsePDFWithMinerU(filename string, data []byte, parser *PDFParser) ParseResult {
+func parsePDFWithMinerU(ctx context.Context, filename string, data []byte, parser *PDFParser) ParseResult {
 	if len(data) == 0 {
 		return emptyPDFResult(filename)
 	}
@@ -50,11 +51,11 @@ func parsePDFWithMinerU(filename string, data []byte, parser *PDFParser) ParseRe
 		apiConfig.ApiKey = &apiKey
 	}
 
-	task, err := driver.ParseFile(&backend, data, nil, apiConfig, &models.ParseFileConfig{}, nil)
+	task, err := driver.ParseFile(ctx, &backend, data, nil, apiConfig, &models.ParseFileConfig{}, nil)
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("parser: MinerU submit: %w", err)}
 	}
-	content, err := pollMinerUTask(driver, task.TaskID, apiConfig, timeout)
+	content, err := pollMinerUTask(ctx, driver, task.TaskID, apiConfig, timeout)
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("parser: MinerU result: %w", err)}
 	}
@@ -62,17 +63,17 @@ func parsePDFWithMinerU(filename string, data []byte, parser *PDFParser) ParseRe
 	if strings.TrimSpace(content) != "" {
 		pageCount = 1
 	}
-	return parseMinerUMarkdownResult(filename, content, parser.OutputFormat, pageCount)
+	return parseMinerUMarkdownResult(ctx, filename, content, parser.OutputFormat, pageCount)
 }
 
-func pollMinerUTask(driver *models.MinerULocalModel, taskID string, apiConfig *models.APIConfig, timeout time.Duration) (string, error) {
+func pollMinerUTask(ctx context.Context, driver *models.MinerULocalModel, taskID string, apiConfig *models.APIConfig, timeout time.Duration) (string, error) {
 	if timeout <= 0 {
 		timeout = minerUPollTimeout
 	}
 	deadline := time.Now().Add(timeout)
 	var lastErr error
 	for {
-		task, err := driver.ShowTask(taskID, apiConfig)
+		task, err := driver.ShowTask(ctx, taskID, apiConfig)
 		if err == nil {
 			for _, segment := range task.Segments {
 				if strings.TrimSpace(segment.Content) != "" {
@@ -93,12 +94,12 @@ func pollMinerUTask(driver *models.MinerULocalModel, taskID string, apiConfig *m
 	}
 }
 
-func parseMinerUMarkdownResult(filename, markdown, outputFormat string, pageCount int) ParseResult {
+func parseMinerUMarkdownResult(ctx context.Context, filename, markdown, outputFormat string, pageCount int) ParseResult {
 	fileMeta := pdfFileMeta(filename, pageCount)
 	switch strings.ToLower(strings.TrimSpace(outputFormat)) {
 	case "", "json":
 		mp, _ := NewMarkdownParser(GoMarkdown)
-		res := mp.ParseWithResult(filename, []byte(markdown))
+		res := mp.ParseWithResult(ctx, filename, []byte(markdown))
 		if res.Err != nil {
 			return res
 		}
diff --git a/internal/parser/parser/pdf_parser_mineru_test.go b/internal/parser/parser/pdf_parser_mineru_test.go
index 88e549efe7..85eab7fe33 100644
--- a/internal/parser/parser/pdf_parser_mineru_test.go
+++ b/internal/parser/parser/pdf_parser_mineru_test.go
@@ -80,7 +80,8 @@ func TestPDFParser_ParseWithResult_MinerUMarkdownIntegration(t *testing.T) {
 		"mineru_backend":   "pipeline",
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -120,7 +121,8 @@ func TestPDFParser_ParseWithResult_MinerUJSONIntegration(t *testing.T) {
 		"mineru_apiserver": server.URL,
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -139,7 +141,8 @@ func TestPDFParser_ParseWithResult_MinerURequiresAPIServer(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "MinerU"})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil {
 		t.Fatal("ParseWithResult: want error when mineru_apiserver is missing, got nil")
 	}
diff --git a/internal/parser/parser/pdf_parser_opendataloader.go b/internal/parser/parser/pdf_parser_opendataloader.go
index 9c4474fcff..089327428a 100644
--- a/internal/parser/parser/pdf_parser_opendataloader.go
+++ b/internal/parser/parser/pdf_parser_opendataloader.go
@@ -13,7 +13,7 @@ import (
 	models "ragflow/internal/entity/models"
 )
 
-func parsePDFWithOpenDataLoader(filename string, data []byte, parser *PDFParser) ParseResult {
+func parsePDFWithOpenDataLoader(ctx context.Context, filename string, data []byte, parser *PDFParser) ParseResult {
 	if len(data) == 0 {
 		return emptyPDFResult(filename)
 	}
@@ -67,7 +67,7 @@ func parsePDFWithOpenDataLoader(filename string, data []byte, parser *PDFParser)
 		}
 	}
 	if strings.TrimSpace(payload.MDText) != "" {
-		return parseMinerUMarkdownResult(filename, payload.MDText, parser.OutputFormat, 1)
+		return parseMinerUMarkdownResult(ctx, filename, payload.MDText, parser.OutputFormat, 1)
 	}
 	return ParseResult{Err: fmt.Errorf("parser: OpenDataLoader returned no parsed content")}
 }
diff --git a/internal/parser/parser/pdf_parser_opendataloader_test.go b/internal/parser/parser/pdf_parser_opendataloader_test.go
index cc1ca4facb..89e2e93d03 100644
--- a/internal/parser/parser/pdf_parser_opendataloader_test.go
+++ b/internal/parser/parser/pdf_parser_opendataloader_test.go
@@ -70,7 +70,8 @@ func TestPDFParser_ParseWithResult_OpenDataLoaderJSONIntegration(t *testing.T) {
 		"sanitize":                 sanitize,
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -99,7 +100,8 @@ func TestPDFParser_ParseWithResult_OpenDataLoaderMarkdownFallback(t *testing.T)
 		"output_format":            "markdown",
 		"opendataloader_apiserver": server.URL,
 	})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -111,7 +113,8 @@ func TestPDFParser_ParseWithResult_OpenDataLoaderMarkdownFallback(t *testing.T)
 func TestPDFParser_ParseWithResult_OpenDataLoaderRequiresAPIServer(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "OpenDataLoader"})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil || !strings.Contains(res.Err.Error(), "opendataloader_apiserver") {
 		t.Fatalf("error = %v, want opendataloader_apiserver context", res.Err)
 	}
diff --git a/internal/parser/parser/pdf_parser_paddleocr.go b/internal/parser/parser/pdf_parser_paddleocr.go
index c48a27ff9b..e537933b0f 100644
--- a/internal/parser/parser/pdf_parser_paddleocr.go
+++ b/internal/parser/parser/pdf_parser_paddleocr.go
@@ -1,6 +1,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 	"ragflow/internal/common"
 	"strings"
@@ -8,7 +9,7 @@ import (
 	models "ragflow/internal/entity/models"
 )
 
-func parsePDFWithPaddleOCR(filename string, data []byte, parser *PDFParser) ParseResult {
+func parsePDFWithPaddleOCR(ctx context.Context, filename string, data []byte, parser *PDFParser) ParseResult {
 	if len(data) == 0 {
 		return emptyPDFResult(filename)
 	}
@@ -45,7 +46,7 @@ func parsePDFWithPaddleOCR(filename string, data []byte, parser *PDFParser) Pars
 		apiConfig.ApiKey = &apiKey
 	}
 
-	resp, err := driver.OCRFile(&algorithm, data, &filename, apiConfig, &models.OCRConfig{
+	resp, err := driver.OCRFile(ctx, &algorithm, data, &filename, apiConfig, &models.OCRConfig{
 		Algorithm: algorithm,
 	}, nil)
 	if err != nil {
@@ -58,5 +59,5 @@ func parsePDFWithPaddleOCR(filename string, data []byte, parser *PDFParser) Pars
 	if resp.Text != nil && strings.TrimSpace(*resp.Text) == "" {
 		pageCount = 0
 	}
-	return parseMinerUMarkdownResult(filename, *resp.Text, parser.OutputFormat, pageCount)
+	return parseMinerUMarkdownResult(ctx, filename, *resp.Text, parser.OutputFormat, pageCount)
 }
diff --git a/internal/parser/parser/pdf_parser_paddleocr_test.go b/internal/parser/parser/pdf_parser_paddleocr_test.go
index 934b84e02a..3b7f8fae0e 100644
--- a/internal/parser/parser/pdf_parser_paddleocr_test.go
+++ b/internal/parser/parser/pdf_parser_paddleocr_test.go
@@ -62,7 +62,8 @@ func TestPDFParser_ParseWithResult_PaddleOCRMarkdownIntegration(t *testing.T) {
 		"paddleocr_api_key":  "paddle-secret",
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -98,7 +99,8 @@ func TestPDFParser_ParseWithResult_PaddleOCRJSONIntegration(t *testing.T) {
 		"paddleocr_base_url": server.URL,
 	})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -117,7 +119,8 @@ func TestPDFParser_ParseWithResult_PaddleOCRRequiresBaseURL(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "PaddleOCR"})
 
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil {
 		t.Fatal("ParseWithResult: want error when paddleocr_base_url is missing, got nil")
 	}
diff --git a/internal/parser/parser/pdf_parser_somark_test.go b/internal/parser/parser/pdf_parser_somark_test.go
index 8733e10854..a86641c499 100644
--- a/internal/parser/parser/pdf_parser_somark_test.go
+++ b/internal/parser/parser/pdf_parser_somark_test.go
@@ -64,7 +64,8 @@ func TestPDFParser_ParseWithResult_SoMarkJSONIntegration(t *testing.T) {
 		"somark_base_url": server.URL,
 		"somark_api_key":  "somark-secret",
 	})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -94,7 +95,8 @@ func TestPDFParser_ParseWithResult_SoMarkMarkdownIntegration(t *testing.T) {
 		"output_format":   "markdown",
 		"somark_base_url": server.URL,
 	})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -107,7 +109,8 @@ func TestPDFParser_ParseWithResult_SoMarkRequiresBaseURL(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "SoMark"})
 	pdf.SoMarkBaseURL = ""
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil || !strings.Contains(res.Err.Error(), "somark_base_url") {
 		t.Fatalf("error = %v, want somark_base_url context", res.Err)
 	}
diff --git a/internal/parser/parser/pdf_parser_tcadp_test.go b/internal/parser/parser/pdf_parser_tcadp_test.go
index 84eca46302..838539f43d 100644
--- a/internal/parser/parser/pdf_parser_tcadp_test.go
+++ b/internal/parser/parser/pdf_parser_tcadp_test.go
@@ -36,7 +36,8 @@ func TestPDFParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 		"tcadp_apiserver": server.URL,
 		"tcadp_api_key":   "tcadp-secret",
 	})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -65,7 +66,8 @@ func TestPDFParser_ParseWithResult_TCADPMarkdownIntegration(t *testing.T) {
 		"output_format":   "markdown",
 		"tcadp_apiserver": server.URL,
 	})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -77,7 +79,8 @@ func TestPDFParser_ParseWithResult_TCADPMarkdownIntegration(t *testing.T) {
 func TestPDFParser_ParseWithResult_TCADPRequiresAPIServer(t *testing.T) {
 	pdf := NewPDFParser()
 	pdf.ConfigureFromSetup(map[string]any{"parse_method": "TCADP parser"})
-	res := pdf.ParseWithResult("sample.pdf", []byte("%PDF-1.4\nmock"))
+	ctx := t.Context()
+	res := pdf.ParseWithResult(ctx, "sample.pdf", []byte("%PDF-1.4\nmock"))
 	if res.Err == nil || !strings.Contains(res.Err.Error(), "tcadp_apiserver") {
 		t.Fatalf("error = %v, want tcadp_apiserver context", res.Err)
 	}
diff --git a/internal/parser/parser/picture_parser.go b/internal/parser/parser/picture_parser.go
index b2a9e005f7..b2964ae1ab 100644
--- a/internal/parser/parser/picture_parser.go
+++ b/internal/parser/parser/picture_parser.go
@@ -30,6 +30,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 	"path/filepath"
 	"strings"
@@ -98,7 +99,7 @@ func (p *PictureParser) ConfigureFromSetup(setup map[string]any) {
 // file extension against the image extension whitelist. The actual
 // OCR and VLM description happens via maybeDispatchImage at the
 // component layer (mirrors Python's picture.py:chunk()).
-func (p *PictureParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *PictureParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	ext := strings.ToLower(filepath.Ext(filename))
 	if len(ext) > 1 && ext[0] == '.' {
 		ext = ext[1:]
diff --git a/internal/parser/parser/picture_parser_test.go b/internal/parser/parser/picture_parser_test.go
index e5327b4f51..797fdd58c5 100644
--- a/internal/parser/parser/picture_parser_test.go
+++ b/internal/parser/parser/picture_parser_test.go
@@ -66,9 +66,10 @@ func TestPictureParser_ConfigureFromSetup(t *testing.T) {
 }
 
 func TestPictureParser_ParseWithResult_NilSetup(t *testing.T) {
+	ctx := t.Context()
 	p := NewPictureParser()
 	p.ConfigureFromSetup(nil)
-	res := p.ParseWithResult("photo.png", []byte("\x89PNG"))
+	res := p.ParseWithResult(ctx, "photo.png", []byte("\x89PNG"))
 	if res.Err != nil {
 		t.Errorf("unexpected error for nil setup: %v", res.Err)
 	}
@@ -82,11 +83,12 @@ func TestPictureParser_ParseWithResult_NilSetup(t *testing.T) {
 }
 
 func TestPictureParser_ParseWithResult_ValidExtensions(t *testing.T) {
+	ctx := t.Context()
 	exts := []string{"png", "jpg", "jpeg", "gif", "bmp", "tiff", "tif", "webp", "svg", "ico", "avif", "heic", "apng"}
 	p := NewPictureParser()
 	for _, ext := range exts {
 		fn := "img." + ext
-		res := p.ParseWithResult(fn, []byte{1, 2, 3})
+		res := p.ParseWithResult(ctx, fn, []byte{1, 2, 3})
 		if res.Err != nil {
 			t.Errorf("unexpected error for .%s: %v", ext, res.Err)
 		}
@@ -97,8 +99,9 @@ func TestPictureParser_ParseWithResult_ValidExtensions(t *testing.T) {
 }
 
 func TestPictureParser_ParseWithResult_InvalidExtension(t *testing.T) {
+	ctx := t.Context()
 	p := NewPictureParser()
-	res := p.ParseWithResult("sound.mp3", []byte{1})
+	res := p.ParseWithResult(ctx, "sound.mp3", []byte{1})
 	if res.Err == nil {
 		t.Error("expected error for .mp3, got nil")
 	}
@@ -108,8 +111,9 @@ func TestPictureParser_ParseWithResult_InvalidExtension(t *testing.T) {
 }
 
 func TestPictureParser_ParseWithResult_VideoExtension(t *testing.T) {
+	ctx := t.Context()
 	p := NewPictureParser()
-	res := p.ParseWithResult("video.mp4", []byte{1})
+	res := p.ParseWithResult(ctx, "video.mp4", []byte{1})
 	if res.Err == nil {
 		t.Error("expected error for video .mp4, got nil")
 	}
@@ -119,11 +123,12 @@ func TestPictureParser_ParseWithResult_VideoExtension(t *testing.T) {
 }
 
 func TestPictureParser_ParseWithResult_OutputFormat(t *testing.T) {
+	ctx := t.Context()
 	p := NewPictureParser()
 	p.ConfigureFromSetup(map[string]any{
 		"output_format": "json",
 	})
-	res := p.ParseWithResult("photo.jpg", []byte{1, 2, 3})
+	res := p.ParseWithResult(ctx, "photo.jpg", []byte{1, 2, 3})
 	if res.Err != nil {
 		t.Fatalf("unexpected error: %v", res.Err)
 	}
diff --git a/internal/parser/parser/ppt_parser.go b/internal/parser/parser/ppt_parser.go
index 595a27c260..b57322225e 100644
--- a/internal/parser/parser/ppt_parser.go
+++ b/internal/parser/parser/ppt_parser.go
@@ -18,6 +18,8 @@
 
 package parser
 
+import "context"
+
 type PPTParser struct{}
 
 func NewPPTParser() *PPTParser {
@@ -33,6 +35,6 @@ func (p *PPTParser) String() string {
 // hint (OLE binary). The two file families differ only in the
 // binary container; the python parser.py:slides branch treats
 // them uniformly.
-func (p *PPTParser) ParseWithResult(filename string, data []byte) ParseResult {
-	return (&PPTXParser{format: "ppt"}).ParseWithResult(filename, data)
+func (p *PPTParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
+	return (&PPTXParser{format: "ppt"}).ParseWithResult(ctx, filename, data)
 }
diff --git a/internal/parser/parser/ppt_parser_cgo_test.go b/internal/parser/parser/ppt_parser_cgo_test.go
index b615bd6284..2cb7928604 100644
--- a/internal/parser/parser/ppt_parser_cgo_test.go
+++ b/internal/parser/parser/ppt_parser_cgo_test.go
@@ -26,9 +26,10 @@ func TestPPTXParser_FormatField(t *testing.T) {
 // JSON items. Uses office_oxide's own PptxWriter to produce the
 // test data so no external file is needed.
 func TestPPTXParser_ParseWithResult_CGO(t *testing.T) {
+	ctx := t.Context()
 	p := NewPPTXParser()
 	data := buildPPTX(t, "Hello World")
-	res := p.ParseWithResult("test.pptx", data)
+	res := p.ParseWithResult(ctx, "test.pptx", data)
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -47,12 +48,13 @@ func TestPPTXParser_ParseWithResult_CGO(t *testing.T) {
 // delegates correctly to PPTXParser{format:"ppt"} and produces
 // output with File["format"] = "ppt".
 func TestPPTParser_ParseWithResult_CGO(t *testing.T) {
+	ctx := t.Context()
 	p := NewPPTParser()
 	// Use PPTX content — office_oxide may reject it with format="ppt"
 	// hint (expects OLE binary). When it does, skip gracefully; when
 	// it succeeds, verify the metadata contract.
 	data := buildPPTX(t, "Hello")
-	res := p.ParseWithResult("test.ppt", data)
+	res := p.ParseWithResult(ctx, "test.ppt", data)
 	if res.Err != nil {
 		t.Skip("PPTParser with PPTX data (expected maybe to fail):", res.Err)
 	}
diff --git a/internal/parser/parser/pptx_parser.go b/internal/parser/parser/pptx_parser.go
index f5fc806f97..3386d48f5d 100644
--- a/internal/parser/parser/pptx_parser.go
+++ b/internal/parser/parser/pptx_parser.go
@@ -19,6 +19,7 @@
 package parser
 
 import (
+	"context"
 	"fmt"
 	"strings"
 
@@ -45,7 +46,7 @@ func (p *PPTXParser) String() string {
 // ParseWithResult emits one JSON item per slide with the slide's
 // plain text. Mirrors the python parser.py:slides branch which
 // forces output_format="json" for the slide family.
-func (p *PPTXParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *PPTXParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	doc, err := officeOxide.OpenFromBytes(data, p.format)
 	if err != nil {
 		return ParseResult{Err: fmt.Errorf("presentation open: %w", err)}
diff --git a/internal/parser/parser/text_parser.go b/internal/parser/parser/text_parser.go
index de041e0202..6f2f03769f 100644
--- a/internal/parser/parser/text_parser.go
+++ b/internal/parser/parser/text_parser.go
@@ -34,6 +34,7 @@ package parser
 
 import (
 	"bytes"
+	"context"
 	"strings"
 )
 
@@ -60,7 +61,7 @@ func NewTextParser() *TextParser {
 // The items slice is always non-nil so downstream chunkers see a
 // non-empty JSON payload even for an empty input (mirrors the
 // MarkdownParser convention at markdown_parser.go:71-76).
-func (p *TextParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *TextParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	if !utf8Valid(data) {
 		return ParseResult{Err: errInvalidUTF8}
 	}
diff --git a/internal/parser/parser/xls_parser.go b/internal/parser/parser/xls_parser.go
index 5101108eca..216d85c038 100644
--- a/internal/parser/parser/xls_parser.go
+++ b/internal/parser/parser/xls_parser.go
@@ -18,6 +18,7 @@ package parser
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"strings"
 
@@ -73,7 +74,7 @@ func (p *XLSParser) ConfigureFromSetup(setup map[string]any) {
 	}
 }
 
-func (p *XLSParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *XLSParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	method := normalizeXLSXParseMethod(p.ParseMethod)
 	switch method {
 	case "tcadp":
diff --git a/internal/parser/parser/xls_tcadp_test.go b/internal/parser/parser/xls_tcadp_test.go
index 5f1b3c916d..111ca4bb00 100644
--- a/internal/parser/parser/xls_tcadp_test.go
+++ b/internal/parser/parser/xls_tcadp_test.go
@@ -7,6 +7,7 @@ import (
 )
 
 func TestXLSXParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
+	ctx := t.Context()
 	zipPayload := tcadpZipFixture(t)
 	var server *httptest.Server
 	server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -34,7 +35,7 @@ func TestXLSXParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 		"table_result_type":            "1",
 		"markdown_image_response_type": "1",
 	})
-	res := p.ParseWithResult("sample.xlsx", []byte("mock xlsx content"))
+	res := p.ParseWithResult(ctx, "sample.xlsx", []byte("mock xlsx content"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -44,6 +45,7 @@ func TestXLSXParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 }
 
 func TestXLSParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
+	ctx := t.Context()
 	zipPayload := tcadpZipFixture(t)
 	var server *httptest.Server
 	server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -68,7 +70,7 @@ func TestXLSParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 		"tcadp_apiserver": server.URL,
 		"tcadp_api_key":   "tcadp-secret",
 	})
-	res := p.ParseWithResult("sample.xls", []byte("mock xls content"))
+	res := p.ParseWithResult(ctx, "sample.xls", []byte("mock xls content"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -78,6 +80,7 @@ func TestXLSParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 }
 
 func TestCSVParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
+	ctx := t.Context()
 	zipPayload := tcadpZipFixture(t)
 	var server *httptest.Server
 	server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -99,7 +102,7 @@ func TestCSVParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 		"tcadp_apiserver": server.URL,
 		"tcadp_api_key":   "tcadp-secret",
 	})
-	res := p.ParseWithResult("sample.csv", []byte("a,b,c\n1,2,3"))
+	res := p.ParseWithResult(ctx, "sample.csv", []byte("a,b,c\n1,2,3"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -109,6 +112,7 @@ func TestCSVParser_ParseWithResult_TCADPJSONIntegration(t *testing.T) {
 }
 
 func TestXLSXParser_ParseWithResult_TCADPMarkdownIntegration(t *testing.T) {
+	ctx := t.Context()
 	zipPayload := tcadpZipFixture(t)
 	var server *httptest.Server
 	server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -132,7 +136,7 @@ func TestXLSXParser_ParseWithResult_TCADPMarkdownIntegration(t *testing.T) {
 		"output_format":   "markdown",
 		"tcadp_apiserver": server.URL,
 	})
-	res := p.ParseWithResult("sample.xlsx", []byte("mock xlsx content"))
+	res := p.ParseWithResult(ctx, "sample.xlsx", []byte("mock xlsx content"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
@@ -142,63 +146,69 @@ func TestXLSXParser_ParseWithResult_TCADPMarkdownIntegration(t *testing.T) {
 }
 
 func TestXLSXParser_ParseWithResult_TCADPRequiresAPIServer(t *testing.T) {
+	ctx := t.Context()
 	p, err := NewXLSXParser("")
 	if err != nil {
 		t.Fatalf("NewXLSXParser: %v", err)
 	}
 	p.ConfigureFromSetup(map[string]any{"parse_method": "TCADP parser"})
-	res := p.ParseWithResult("sample.xlsx", []byte("mock xlsx content"))
+	res := p.ParseWithResult(ctx, "sample.xlsx", []byte("mock xlsx content"))
 	if res.Err == nil {
 		t.Fatal("expected error about tcadp_apiserver, got nil")
 	}
 }
 
 func TestXLSParser_ParseWithResult_TCADPRequiresAPIServer(t *testing.T) {
+	ctx := t.Context()
 	p, err := NewXLSParser("")
 	if err != nil {
 		t.Fatalf("NewXLSParser: %v", err)
 	}
 	p.ConfigureFromSetup(map[string]any{"parse_method": "TCADP parser"})
-	res := p.ParseWithResult("sample.xls", []byte("mock xls content"))
+	res := p.ParseWithResult(ctx, "sample.xls", []byte("mock xls content"))
 	if res.Err == nil {
 		t.Fatal("expected error about tcadp_apiserver, got nil")
 	}
 }
 
 func TestCSVParser_ParseWithResult_TCADPRequiresAPIServer(t *testing.T) {
+	ctx := t.Context()
 	p := NewCSVParser()
 	p.ConfigureFromSetup(map[string]any{"parse_method": "TCADP parser"})
-	res := p.ParseWithResult("sample.csv", []byte("mock csv content"))
+	res := p.ParseWithResult(ctx, "sample.csv", []byte("mock csv content"))
 	if res.Err == nil {
 		t.Fatal("expected error about tcadp_apiserver, got nil")
 	}
 }
 
 func TestXLSXParser_ParseWithResult_InvalidXLSXHandled(t *testing.T) {
+	ctx := t.Context()
 	p, err := NewXLSXParser("")
 	if err != nil {
 		t.Fatalf("NewXLSXParser: %v", err)
 	}
-	res := p.ParseWithResult("sample.xlsx", []byte("not a valid xlsx"))
+	res := p.ParseWithResult(ctx, "sample.xlsx", []byte("not a valid xlsx"))
 	if res.Err == nil {
 		t.Fatal("expected error for invalid xlsx, got nil")
 	}
 }
 
 func TestXLSParser_ParseWithResult_InvalidXLSHandled(t *testing.T) {
+	ctx := t.Context()
 	p, err := NewXLSParser("")
 	if err != nil {
 		t.Fatalf("NewXLSParser: %v", err)
 	}
-	res := p.ParseWithResult("sample.xls", []byte("not a valid xls"))
+	res := p.ParseWithResult(ctx, "sample.xls", []byte("not a valid xls"))
 	if res.Err == nil {
 		t.Fatal("expected error for invalid xls, got nil")
 	}
 }
 
 func TestCSVParser_ParseWithResult_DefaultCSVBehavior(t *testing.T) {
+	ctx := t.Context()
 	p := NewCSVParser()
-	res := p.ParseWithResult("sample.csv", []byte("a,b\n1,2"))
+	res := p.ParseWithResult(ctx, "sample.csv", []byte("a,b\n1,2"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult: %v", res.Err)
 	}
diff --git a/internal/parser/parser/xlsx_parse_method_test.go b/internal/parser/parser/xlsx_parse_method_test.go
index f50f3086b2..0f965f8480 100644
--- a/internal/parser/parser/xlsx_parse_method_test.go
+++ b/internal/parser/parser/xlsx_parse_method_test.go
@@ -69,7 +69,8 @@ func TestXLSXParser_DeepDocParseMethod(t *testing.T) {
 			}
 			p.ConfigureFromSetup(map[string]any{"parse_method": tc.method})
 
-			res := p.ParseWithResult("test.xlsx", buf.Bytes())
+			ctx := t.Context()
+			res := p.ParseWithResult(ctx, "test.xlsx", buf.Bytes())
 			if res.Err != nil {
 				t.Fatalf("ParseWithResult(%s): %v", tc.method, res.Err)
 			}
@@ -89,7 +90,8 @@ func TestCSVParser_DeepDocParseMethod(t *testing.T) {
 	p := NewCSVParser()
 	p.ConfigureFromSetup(map[string]any{"parse_method": "deepdoc"})
 
-	res := p.ParseWithResult("test.csv", []byte("a,b\n1,2"))
+	ctx := t.Context()
+	res := p.ParseWithResult(ctx, "test.csv", []byte("a,b\n1,2"))
 	if res.Err != nil {
 		t.Fatalf("ParseWithResult(deepdoc): %v", res.Err)
 	}
@@ -114,7 +116,8 @@ func TestXLSParser_DeepDocParseMethod_NoUnsupportedError(t *testing.T) {
 	}
 	p.ConfigureFromSetup(map[string]any{"parse_method": "deepdoc"})
 
-	res := p.ParseWithResult("test.xls", []byte("not a real xls"))
+	ctx := t.Context()
+	res := p.ParseWithResult(ctx, "test.xls", []byte("not a real xls"))
 	if res.Err != nil && strings.Contains(res.Err.Error(), "unsupported XLS parse method") {
 		t.Fatalf("deepdoc must not be rejected as unsupported: %v", res.Err)
 	}
diff --git a/internal/parser/parser/xlsx_parser.go b/internal/parser/parser/xlsx_parser.go
index 0ffd4f3c7a..5d2478d600 100644
--- a/internal/parser/parser/xlsx_parser.go
+++ b/internal/parser/parser/xlsx_parser.go
@@ -18,6 +18,7 @@ package parser
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"strings"
 
@@ -90,7 +91,7 @@ func normalizeXLSXParseMethod(raw string) string {
 	return method
 }
 
-func (p *XLSXParser) ParseWithResult(filename string, data []byte) ParseResult {
+func (p *XLSXParser) ParseWithResult(ctx context.Context, filename string, data []byte) ParseResult {
 	method := normalizeXLSXParseMethod(p.ParseMethod)
 	switch method {
 	case "tcadp":
diff --git a/internal/router/router.go b/internal/router/router.go
index 7cdd126d40..375f19c347 100644
--- a/internal/router/router.go
+++ b/internal/router/router.go
@@ -173,7 +173,7 @@ func (r *Router) Setup(engine *gin.Engine) {
 		apiNoAuth.GET("/pipelines/:id", r.pipelineHandler.GetPipeline)
 
 		// searchbots
-		apiNoAuth.GET("/searchbots/detail", r.searchBotHandler.SearchbotDetail)
+		apiNoAuth.GET("/searchbots/detail", r.searchBotHandler.SearchBotDetail)
 
 		// User login channels endpoint
 		apiNoAuth.GET("/auth/login/channels", r.userHandler.GetLoginChannels)
diff --git a/internal/service/agent.go b/internal/service/agent.go
index e5f797a9a8..ca5db0f568 100644
--- a/internal/service/agent.go
+++ b/internal/service/agent.go
@@ -1281,7 +1281,7 @@ func (s *AgentService) buildRunFunc(canvasID string, versionRow *entity.UserCanv
 			// Only used for ParseAgentUploads (read-only); nil DocRemover means
 			// this FileService MUST NOT be used for DeleteFiles.
 			fileSvc := file.NewFileService(CheckFileTeamPermission, nil)
-			files, ferr := fileSvc.ParseAgentUploads(userID, rawFiles, beginLayoutRecognize(c))
+			files, ferr := fileSvc.ParseAgentUploads(ctx, userID, rawFiles, beginLayoutRecognize(c))
 			if ferr != nil {
 				s.markRunFailed(ctx2, runID, "parse files: "+ferr.Error())
 				return nil, fmt.Errorf("parse agent files: %w", ferr)
diff --git a/internal/service/ask_service.go b/internal/service/ask_service.go
index 40ef84a453..9a50b17a7c 100644
--- a/internal/service/ask_service.go
+++ b/internal/service/ask_service.go
@@ -224,7 +224,7 @@ func (s *AskService) run(ctx context.Context, llm StreamingLLM, userID, question
 	// Attempt citation insertion if embedder is available.
 	chunkVectors := ExtractChunkVectors(result.Chunks)
 	if len(chunkVectors) > 0 && s.embedder != nil {
-		if decorated, cited := InsertCitations(visible, chunks, s.embedder, chunkVectors); len(cited) > 0 {
+		if decorated, cited := InsertCitations(ctx, visible, chunks, s.embedder, chunkVectors); len(cited) > 0 {
 			visible = decorated
 		}
 	}
diff --git a/internal/service/chat_pipeline.go b/internal/service/chat_pipeline.go
index 9b4e04e0a2..fc81b22df9 100644
--- a/internal/service/chat_pipeline.go
+++ b/internal/service/chat_pipeline.go
@@ -333,9 +333,9 @@ func (s *ChatPipelineService) AsyncChat(
 				}
 			}
 			if modelType == "chat" {
-				textAttachmentsList, imageAttachments = splitFileAttachments(userID, files, false)
+				textAttachmentsList, imageAttachments = splitFileAttachments(ctx, userID, files, false)
 			} else {
-				textAttachmentsList, imageFiles = splitFileAttachments(userID, files, true)
+				textAttachmentsList, imageFiles = splitFileAttachments(ctx, userID, files, true)
 			}
 			attachments = strings.Join(textAttachmentsList, "\n\n")
 			common.Debug("Resolved attachments",
@@ -863,7 +863,7 @@ func (s *ChatPipelineService) AsyncChat(
 				out <- AsyncChatResult{
 					Answer:      emptyResp,
 					Reference:   kbinfos,
-					AudioBinary: s.synthesizeTTS(ttsModel, emptyResp),
+					AudioBinary: s.synthesizeTTS(ctx, ttsModel, emptyResp),
 					Prompt:      fmt.Sprintf("\n\n### Query:\n%s", strings.Join(questions, " ")),
 					Final:       true,
 				}
@@ -1112,7 +1112,7 @@ func (s *ChatPipelineService) AsyncChat(
 							out <- AsyncChatResult{
 								Answer:      text,
 								Reference:   map[string]interface{}{},
-								AudioBinary: s.synthesizeTTS(ttsModel, text),
+								AudioBinary: s.synthesizeTTS(ctx, ttsModel, text),
 								CreatedAt:   float64(time.Now().Unix()),
 								Final:       false,
 							}
@@ -1121,7 +1121,7 @@ func (s *ChatPipelineService) AsyncChat(
 					})
 			} else {
 				driverErr = chatDriver.ModelDriver.ChatStreamlyWithSender(
-					*chatDriver.ModelName, chatMessages, chatDriver.APIConfig, chatCfg, nil,
+					ctx, *chatDriver.ModelName, chatMessages, chatDriver.APIConfig, chatCfg, nil,
 					func(answer *string, reason *string) error {
 						if reason != nil && *reason != "" {
 							if thinkState.EnterReasoning() {
@@ -1141,7 +1141,7 @@ func (s *ChatPipelineService) AsyncChat(
 									out <- AsyncChatResult{
 										Answer:      d.Value,
 										Reference:   map[string]interface{}{},
-										AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+										AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 										CreatedAt:   float64(time.Now().Unix()),
 										Final:       false,
 									}
@@ -1156,7 +1156,7 @@ func (s *ChatPipelineService) AsyncChat(
 										out <- AsyncChatResult{
 											Answer:      d.Value,
 											Reference:   map[string]interface{}{},
-											AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+											AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 											CreatedAt:   float64(time.Now().Unix()),
 											Final:       false,
 										}
@@ -1178,7 +1178,7 @@ func (s *ChatPipelineService) AsyncChat(
 									out <- AsyncChatResult{
 										Answer:      d.Value,
 										Reference:   map[string]interface{}{},
-										AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+										AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 										CreatedAt:   float64(time.Now().Unix()),
 										Final:       false,
 									}
@@ -1216,7 +1216,7 @@ func (s *ChatPipelineService) AsyncChat(
 					out <- AsyncChatResult{
 						Answer:      d.Value,
 						Reference:   map[string]interface{}{},
-						AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+						AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 						CreatedAt:   float64(time.Now().Unix()),
 						Final:       false,
 					}
@@ -1258,7 +1258,7 @@ func (s *ChatPipelineService) AsyncChat(
 				answer, _, err = chatDriver.ChatWithTools(ctx, prompt+prompt4citation, chatMessages, chatCfg)
 			} else {
 				resp, respErr := chatDriver.ModelDriver.ChatWithMessages(
-					*chatDriver.ModelName, chatMessages, chatDriver.APIConfig, chatCfg, nil,
+					ctx, *chatDriver.ModelName, chatMessages, chatDriver.APIConfig, chatCfg, nil,
 				)
 				if respErr != nil {
 					err = respErr
@@ -1344,11 +1344,11 @@ func (s *ChatPipelineService) AsyncChatSolo(
 		isImage2Text := modelType == "image2text"
 		if len(messages) > 0 {
 			if files, hasFiles := messages[len(messages)-1]["files"]; hasFiles {
-				attachmentsStr = s.processFileAttachments(userID, files)
+				attachmentsStr = s.processFileAttachments(ctx, userID, files)
 				if isImage2Text {
 					imageFiles = s.extractRawImageURLs(files)
 				} else {
-					imageFiles = s.extractImageFiles(userID, files)
+					imageFiles = s.extractImageFiles(ctx, userID, files)
 				}
 			}
 		}
@@ -1439,7 +1439,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 			timer.Enter(common.PhaseGenerateAnswer)
 
 			driverErr := chatModel.ModelDriver.ChatStreamlyWithSender(
-				*chatModel.ModelName, chatMessages, chatModel.APIConfig, chatCfg, nil,
+				ctx, *chatModel.ModelName, chatMessages, chatModel.APIConfig, chatCfg, nil,
 				func(answer *string, reason *string) error {
 					if reason != nil && *reason != "" {
 						if thinkState.EnterReasoning() {
@@ -1459,7 +1459,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 								out <- AsyncChatResult{
 									Answer:      d.Value,
 									Reference:   map[string]interface{}{},
-									AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+									AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 									CreatedAt:   float64(time.Now().Unix()),
 									Final:       false,
 								}
@@ -1474,7 +1474,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 									out <- AsyncChatResult{
 										Answer:      d.Value,
 										Reference:   map[string]interface{}{},
-										AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+										AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 										CreatedAt:   float64(time.Now().Unix()),
 										Final:       false,
 									}
@@ -1496,7 +1496,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 								out <- AsyncChatResult{
 									Answer:      d.Value,
 									Reference:   map[string]interface{}{},
-									AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+									AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 									CreatedAt:   float64(time.Now().Unix()),
 									Final:       false,
 								}
@@ -1530,7 +1530,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 					out <- AsyncChatResult{
 						Answer:      d.Value,
 						Reference:   map[string]interface{}{},
-						AudioBinary: s.synthesizeTTS(ttsModel, d.Value),
+						AudioBinary: s.synthesizeTTS(ctx, ttsModel, d.Value),
 						CreatedAt:   float64(time.Now().Unix()),
 						Final:       false,
 					}
@@ -1565,7 +1565,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 			chatCfg := BuildChatConfig(chat, nil)
 			timer.Enter(common.PhaseGenerateAnswer)
 			resp, err := chatModel.ModelDriver.ChatWithMessages(
-				*chatModel.ModelName, chatMessages, chatModel.APIConfig, chatCfg, nil,
+				ctx, *chatModel.ModelName, chatMessages, chatModel.APIConfig, chatCfg, nil,
 			)
 			timer.Exit(common.PhaseGenerateAnswer)
 			if err != nil {
@@ -1592,7 +1592,7 @@ func (s *ChatPipelineService) AsyncChatSolo(
 			out <- AsyncChatResult{
 				Answer:      answer,
 				Reference:   map[string]interface{}{},
-				AudioBinary: s.synthesizeTTS(ttsModel, answer),
+				AudioBinary: s.synthesizeTTS(ctx, ttsModel, answer),
 				CreatedAt:   float64(time.Now().Unix()),
 				Final:       true,
 			}
@@ -1604,14 +1604,14 @@ func (s *ChatPipelineService) AsyncChatSolo(
 
 // extractImageFiles extracts data-URI image attachments from the files list.
 // Mirrors Python split_file_attachments raw mode.
-func (s *ChatPipelineService) extractImageFiles(userID string, files interface{}) []string {
+func (s *ChatPipelineService) extractImageFiles(ctx context.Context, userID string, files interface{}) []string {
 	// ── File-dict mode ──
 	if fileDicts, ok := parseFileDicts(files); ok {
 		// Only used for GetFileContents (read-only); nil DocRemover means
 		// this FileService MUST NOT be used for DeleteFiles.
 		fileSvc := file.NewFileService(CheckFileTeamPermission, nil)
 		// Use raw=false to get base64 data URIs for images.
-		_, images, err := fileSvc.GetFileContents(userID, fileDicts, false)
+		_, images, err := fileSvc.GetFileContents(ctx, userID, fileDicts, false)
 		if err != nil {
 			common.Warn("GetFileContents failed in extractImageFiles",
 				zap.Error(err))
@@ -2066,13 +2066,13 @@ func lastUserQuestion(messages []map[string]interface{}) string {
 //
 // When files are file dicts (Python-compatible format), calls
 // FileService.GetFileContents to fetch actual blobs from storage.
-func (s *ChatPipelineService) processFileAttachments(userID string, files interface{}) string {
+func (s *ChatPipelineService) processFileAttachments(ctx context.Context, userID string, files interface{}) string {
 	// ── File-dict mode ──
 	if fileDicts, ok := parseFileDicts(files); ok {
 		// Only used for GetFileContents (read-only); nil DocRemover means
 		// this FileService MUST NOT be used for DeleteFiles.
 		fileSvc := file.NewFileService(CheckFileTeamPermission, nil)
-		texts, _, err := fileSvc.GetFileContents(userID, fileDicts, false)
+		texts, _, err := fileSvc.GetFileContents(ctx, userID, fileDicts, false)
 		if err != nil {
 			common.Warn("GetFileContents failed in processFileAttachments",
 				zap.Error(err))
@@ -2123,13 +2123,13 @@ func (s *ChatPipelineService) processFileAttachments(userID string, files interf
 //     URIs → image files.
 //     - raw=true: all items go to textAttachments (Python's FileService.get_files
 //     with raw=True pre-separates images, so non-image content arrives here).
-func splitFileAttachments(userID string, files interface{}, raw bool) (textAttachments []string, imageAttachments []string) {
+func splitFileAttachments(ctx context.Context, userID string, files interface{}, raw bool) (textAttachments []string, imageAttachments []string) {
 	// ── Mode 1: file dicts (Python-compatible) ──
 	if fileDicts, ok := parseFileDicts(files); ok {
 		// Only used for GetFileContents (read-only); nil DocRemover means
 		// this FileService MUST NOT be used for DeleteFiles.
 		fileSvc := file.NewFileService(CheckFileTeamPermission, nil)
-		texts, images, err := fileSvc.GetFileContents(userID, fileDicts, raw)
+		texts, images, err := fileSvc.GetFileContents(ctx, userID, fileDicts, raw)
 		if err != nil {
 			common.Warn("GetFileContents failed, falling back to string splitting",
 				zap.Error(err))
@@ -2253,7 +2253,7 @@ func cleanTTSText(text string) string {
 
 // synthesizeTTS calls the TTS model to convert text to audio.
 // Mirrors dialog_service.py:1426-1432.
-func (s *ChatPipelineService) synthesizeTTS(ttsModel *modelModule.ChatModel, text string) interface{} {
+func (s *ChatPipelineService) synthesizeTTS(ctx context.Context, ttsModel *modelModule.ChatModel, text string) interface{} {
 	if ttsModel == nil || text == "" {
 		return nil
 	}
@@ -2262,7 +2262,7 @@ func (s *ChatPipelineService) synthesizeTTS(ttsModel *modelModule.ChatModel, tex
 		return nil
 	}
 	ttsResp, err := ttsModel.ModelDriver.AudioSpeech(
-		ttsModel.ModelName, &text, ttsModel.APIConfig, &modelModule.TTSConfig{Format: "mp3"}, nil,
+		ctx, ttsModel.ModelName, &text, ttsModel.APIConfig, &modelModule.TTSConfig{Format: "mp3"}, nil,
 	)
 	if err != nil {
 		common.Warn("TTS synthesis failed", zap.Error(err))
@@ -2651,9 +2651,9 @@ type embeddingModelEmbedder struct {
 	embModel *modelModule.EmbeddingModel
 }
 
-func (e *embeddingModelEmbedder) Encode(texts []string) ([][]float64, error) {
+func (e *embeddingModelEmbedder) Encode(ctx context.Context, texts []string) ([][]float64, error) {
 	config := &modelModule.EmbeddingConfig{Dimension: 0}
-	embeds, err := e.embModel.ModelDriver.Embed(e.embModel.ModelName, texts, e.embModel.APIConfig, config, nil)
+	embeds, err := e.embModel.ModelDriver.Embed(ctx, e.embModel.ModelName, texts, e.embModel.APIConfig, config, nil)
 	if err != nil {
 		return nil, err
 	}
@@ -2737,7 +2737,7 @@ func (s *ChatPipelineService) decorateAnswer(
 				}
 				if allVec {
 					embedder := &embeddingModelEmbedder{embModel: embModel}
-					if decorated, cited := InsertCitations(ans, NewSourcedChunks(chunksRaw), embedder, chunkVectors); len(cited) > 0 {
+					if decorated, cited := InsertCitations(ctx, ans, NewSourcedChunks(chunksRaw), embedder, chunkVectors); len(cited) > 0 {
 						ans = decorated
 						citationIdx = make(map[int]struct{})
 						for _, ci := range cited {
@@ -2852,7 +2852,7 @@ func (s *ChatPipelineService) decorateAnswer(
 	}
 
 	// TTS synthesis for the final answer.
-	audioBinary := s.synthesizeTTS(ttsModel, think+ans)
+	audioBinary := s.synthesizeTTS(ctx, ttsModel, think+ans)
 
 	// Langfuse generation end observation.
 	if langfuseTraceID != "" {
@@ -3610,7 +3610,7 @@ func chatForSQL(
 		modelModule.Message{Role: "user", Content: userPrompt},
 	}
 	resp, err := chatModel.ModelDriver.ChatWithMessages(
-		modelName, msgs, chatModel.APIConfig, cfg, nil,
+		ctx, modelName, msgs, chatModel.APIConfig, cfg, nil,
 	)
 	if err != nil {
 		return "", err
diff --git a/internal/service/chunk/chunk.go b/internal/service/chunk/chunk.go
index 5164575f92..5f02016a31 100644
--- a/internal/service/chunk/chunk.go
+++ b/internal/service/chunk/chunk.go
@@ -1245,7 +1245,7 @@ func (s *ChunkService) RemoveChunks(req *service.RemoveChunksRequest, userID str
 	return deletedCount, nil
 }
 
-func (s *ChunkService) AddChunk(req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
+func (s *ChunkService) AddChunk(ctx context.Context, req *service.AddChunkRequest, userID string) (*service.AddChunkResponse, error) {
 	if s.docEngine == nil {
 		return nil, addChunkError{code: common.CodeServerError, message: "doc engine not initialized"}
 	}
@@ -1351,7 +1351,7 @@ func (s *ChunkService) AddChunk(req *service.AddChunkRequest, userID string) (*s
 	if len(questionKwd) > 0 {
 		embeddingText = strings.Join(questionKwd, "\n")
 	}
-	embeddings, err := embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{docName, embeddingText}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
+	embeddings, err := embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{docName, embeddingText}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
 	if err != nil {
 		return nil, addChunkError{code: common.CodeServerError, message: fmt.Sprintf("encode chunk embedding: %v", err)}
 	}
diff --git a/internal/service/chunk/chunk_test.go b/internal/service/chunk/chunk_test.go
index c03f625c04..a178fcfdda 100644
--- a/internal/service/chunk/chunk_test.go
+++ b/internal/service/chunk/chunk_test.go
@@ -279,6 +279,7 @@ func TestParseRejectsRunningDocument(t *testing.T) {
 }
 
 func TestAddChunkSuccess(t *testing.T) {
+	ctx := t.Context()
 	db := setupChunkTestDB(t)
 	pushChunkTestDB(t, db)
 	userID, datasetID, documentID := "user-1", "kb-1", "doc-1"
@@ -320,7 +321,7 @@ func TestAddChunkSuccess(t *testing.T) {
 		numTokensFunc:           func(text string) int { return len(text) },
 	}
 
-	resp, err := svc.AddChunk(&service.AddChunkRequest{
+	resp, err := svc.AddChunk(ctx, &service.AddChunkRequest{
 		DatasetID:         datasetID,
 		DocumentID:        documentID,
 		Content:           "chunk body",
@@ -370,6 +371,7 @@ func TestAddChunkSuccess(t *testing.T) {
 }
 
 func TestAddChunkValidationErrors(t *testing.T) {
+	ctx := t.Context()
 	db := setupChunkTestDB(t)
 	pushChunkTestDB(t, db)
 	insertChunkTestKB(t, "kb-1", "user-1")
@@ -410,7 +412,7 @@ func TestAddChunkValidationErrors(t *testing.T) {
 			if tt.setup != nil {
 				tt.setup(svc)
 			}
-			_, err := svc.AddChunk(tt.req, "user-1")
+			_, err := svc.AddChunk(ctx, tt.req, "user-1")
 			if err == nil || !strings.Contains(err.Error(), tt.wantMsg) {
 				t.Fatalf("error = %v, want substring %q", err, tt.wantMsg)
 			}
@@ -419,6 +421,7 @@ func TestAddChunkValidationErrors(t *testing.T) {
 }
 
 func TestAddChunkImageAndTagFeatureValidation(t *testing.T) {
+	ctx := t.Context()
 	db := setupChunkTestDB(t)
 	pushChunkTestDB(t, db)
 	userID, datasetID, documentID := "user-1", "kb-1", "doc-1"
@@ -457,7 +460,7 @@ func TestAddChunkImageAndTagFeatureValidation(t *testing.T) {
 		},
 	}
 
-	_, err := svc.AddChunk(&service.AddChunkRequest{
+	_, err := svc.AddChunk(ctx, &service.AddChunkRequest{
 		DatasetID:   datasetID,
 		DocumentID:  documentID,
 		Content:     "chunk body",
@@ -469,7 +472,7 @@ func TestAddChunkImageAndTagFeatureValidation(t *testing.T) {
 
 	validJPEG := "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO2pRZ0AAAAASUVORK5CYII="
 
-	resp, err := svc.AddChunk(&service.AddChunkRequest{
+	resp, err := svc.AddChunk(ctx, &service.AddChunkRequest{
 		DatasetID:  datasetID,
 		DocumentID: documentID,
 		Content:    "chunk body",
@@ -479,7 +482,7 @@ func TestAddChunkImageAndTagFeatureValidation(t *testing.T) {
 		t.Fatalf("expected tag_feas validation error, got resp=%#v err=%v", resp, err)
 	}
 
-	resp, err = svc.AddChunk(&service.AddChunkRequest{
+	resp, err = svc.AddChunk(ctx, &service.AddChunkRequest{
 		DatasetID:   datasetID,
 		DocumentID:  documentID,
 		Content:     "chunk body",
@@ -498,6 +501,7 @@ func TestAddChunkImageAndTagFeatureValidation(t *testing.T) {
 }
 
 func TestAddChunkIncrementsStatsAfterInsert(t *testing.T) {
+	ctx := t.Context()
 	db := setupChunkTestDB(t)
 	pushChunkTestDB(t, db)
 	userID, datasetID, documentID := "user-1", "kb-1", "doc-1"
@@ -533,7 +537,7 @@ func TestAddChunkIncrementsStatsAfterInsert(t *testing.T) {
 		numTokensFunc:           func(text string) int { return len(text) },
 	}
 
-	_, err := svc.AddChunk(&service.AddChunkRequest{
+	_, err := svc.AddChunk(ctx, &service.AddChunkRequest{
 		DatasetID:  datasetID,
 		DocumentID: documentID,
 		Content:    "chunk body",
@@ -1082,47 +1086,47 @@ type stubEmbeddingDriver struct {
 
 func (d *stubEmbeddingDriver) NewInstance(map[string]string) models.ModelDriver { return d }
 func (d *stubEmbeddingDriver) Name() string                                     { return "stub" }
-func (d *stubEmbeddingDriver) ChatWithMessages(string, []models.Message, *models.APIConfig, *models.ChatConfig, *common.ModelUsage) (*models.ChatResponse, error) {
+func (d *stubEmbeddingDriver) ChatWithMessages(context.Context, string, []models.Message, *models.APIConfig, *models.ChatConfig, *common.ModelUsage) (*models.ChatResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) ChatStreamlyWithSender(string, []models.Message, *models.APIConfig, *models.ChatConfig, *common.ModelUsage, func(*string, *string) error) error {
+func (d *stubEmbeddingDriver) ChatStreamlyWithSender(context.Context, string, []models.Message, *models.APIConfig, *models.ChatConfig, *common.ModelUsage, func(*string, *string) error) error {
 	return nil
 }
-func (d *stubEmbeddingDriver) Embed(*string, []string, *models.APIConfig, *models.EmbeddingConfig, *common.ModelUsage) ([]models.EmbeddingData, error) {
+func (d *stubEmbeddingDriver) Embed(context.Context, *string, []string, *models.APIConfig, *models.EmbeddingConfig, *common.ModelUsage) ([]models.EmbeddingData, error) {
 	return d.embeddings, d.embedErr
 }
-func (d *stubEmbeddingDriver) Rerank(*string, string, []string, *models.APIConfig, *models.RerankConfig, *common.ModelUsage) (*models.RerankResponse, error) {
+func (d *stubEmbeddingDriver) Rerank(context.Context, *string, string, []string, *models.APIConfig, *models.RerankConfig, *common.ModelUsage) (*models.RerankResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) TranscribeAudio(*string, *string, *models.APIConfig, *models.ASRConfig, *common.ModelUsage) (*models.ASRResponse, error) {
+func (d *stubEmbeddingDriver) TranscribeAudio(context.Context, *string, *string, *models.APIConfig, *models.ASRConfig, *common.ModelUsage) (*models.ASRResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) TranscribeAudioWithSender(*string, *string, *models.APIConfig, *models.ASRConfig, *common.ModelUsage, func(*string, *string) error) error {
+func (d *stubEmbeddingDriver) TranscribeAudioWithSender(context.Context, *string, *string, *models.APIConfig, *models.ASRConfig, *common.ModelUsage, func(*string, *string) error) error {
 	return nil
 }
-func (d *stubEmbeddingDriver) AudioSpeech(*string, *string, *models.APIConfig, *models.TTSConfig, *common.ModelUsage) (*models.TTSResponse, error) {
+func (d *stubEmbeddingDriver) AudioSpeech(context.Context, *string, *string, *models.APIConfig, *models.TTSConfig, *common.ModelUsage) (*models.TTSResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) AudioSpeechWithSender(*string, *string, *models.APIConfig, *models.TTSConfig, *common.ModelUsage, func(*string, *string) error) error {
+func (d *stubEmbeddingDriver) AudioSpeechWithSender(context.Context, *string, *string, *models.APIConfig, *models.TTSConfig, *common.ModelUsage, func(*string, *string) error) error {
 	return nil
 }
-func (d *stubEmbeddingDriver) OCRFile(*string, []byte, *string, *models.APIConfig, *models.OCRConfig, *common.ModelUsage) (*models.OCRFileResponse, error) {
+func (d *stubEmbeddingDriver) OCRFile(context.Context, *string, []byte, *string, *models.APIConfig, *models.OCRConfig, *common.ModelUsage) (*models.OCRFileResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) ParseFile(*string, []byte, *string, *models.APIConfig, *models.ParseFileConfig, *common.ModelUsage) (*models.ParseFileResponse, error) {
+func (d *stubEmbeddingDriver) ParseFile(context.Context, *string, []byte, *string, *models.APIConfig, *models.ParseFileConfig, *common.ModelUsage) (*models.ParseFileResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) ListModels(*models.APIConfig) ([]models.ListModelResponse, error) {
+func (d *stubEmbeddingDriver) ListModels(context.Context, *models.APIConfig) ([]models.ListModelResponse, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) Balance(*models.APIConfig) (map[string]interface{}, error) {
+func (d *stubEmbeddingDriver) Balance(context.Context, *models.APIConfig) (map[string]interface{}, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) CheckConnection(*models.APIConfig) error { return nil }
-func (d *stubEmbeddingDriver) ListTasks(*models.APIConfig) ([]models.ListTaskStatus, error) {
+func (d *stubEmbeddingDriver) CheckConnection(context.Context, *models.APIConfig) error { return nil }
+func (d *stubEmbeddingDriver) ListTasks(context.Context, *models.APIConfig) ([]models.ListTaskStatus, error) {
 	return nil, nil
 }
-func (d *stubEmbeddingDriver) ShowTask(string, *models.APIConfig) (*models.TaskResponse, error) {
+func (d *stubEmbeddingDriver) ShowTask(context.Context, string, *models.APIConfig) (*models.TaskResponse, error) {
 	return nil, nil
 }
 
diff --git a/internal/service/citation.go b/internal/service/citation.go
index c7c536cfb4..ea54adae7c 100644
--- a/internal/service/citation.go
+++ b/internal/service/citation.go
@@ -17,6 +17,7 @@
 package service
 
 import (
+	"context"
 	"math"
 	"regexp"
 	"strings"
@@ -30,7 +31,7 @@ const minSentenceLen = 5
 
 // Embedder abstracts embedding-model access so InsertCitations is testable.
 type Embedder interface {
-	Encode(texts []string) ([][]float64, error)
+	Encode(ctx context.Context, texts []string) ([][]float64, error)
 }
 
 // CitationMarkerPattern matches "[ID:N]" or bare "[N]" with Arabic digit support,
@@ -57,13 +58,13 @@ var badCitationPatterns = []*regexp.Regexp{
 //  6. Rebuild answer text with [ID:n] markers inserted after cited sentences.
 //
 // Returns the decorated answer and the set of cited chunk indices.
-func InsertCitations(answer string, chunks []SourcedChunk, embedder Embedder, chunkVectors [][]float64) (string, []int) {
+func InsertCitations(ctx context.Context, answer string, chunks []SourcedChunk, embedder Embedder, chunkVectors [][]float64) (string, []int) {
 	sentences, sentenceIdx := splitAnswer(answer)
 	if len(sentences) == 0 || len(chunks) == 0 || len(chunkVectors) == 0 {
 		return answer, nil
 	}
 
-	sentenceVecs, err := embedder.Encode(sentences)
+	sentenceVecs, err := embedder.Encode(ctx, sentences)
 	if err != nil || len(sentenceVecs) == 0 {
 		return answer, nil
 	}
diff --git a/internal/service/citation_test.go b/internal/service/citation_test.go
index 47723f50c5..fb27bbe78d 100644
--- a/internal/service/citation_test.go
+++ b/internal/service/citation_test.go
@@ -17,6 +17,7 @@
 package service
 
 import (
+	"context"
 	"fmt"
 	"math"
 	"strings"
@@ -244,7 +245,7 @@ type fakeEmbedder struct {
 	err  error
 }
 
-func (f *fakeEmbedder) Encode(texts []string) ([][]float64, error) {
+func (f *fakeEmbedder) Encode(ctx context.Context, texts []string) ([][]float64, error) {
 	if f.err != nil {
 		return nil, f.err
 	}
@@ -252,10 +253,11 @@ func (f *fakeEmbedder) Encode(texts []string) ([][]float64, error) {
 }
 
 func TestInsertCitations_Happy(t *testing.T) {
+	ctx := t.Context()
 	chunks := []SourcedChunk{{ID: "abc123"}, {ID: "def456"}}
 	chunkVectors := [][]float64{{1, 0, 0}, {0, 1, 0}}
 	embedder := &fakeEmbedder{vecs: [][]float64{{1, 0, 0}, {0, 1, 0}}}
-	answer, cited := InsertCitations("First sentence. Second sentence here.", chunks, embedder, chunkVectors)
+	answer, cited := InsertCitations(ctx, "First sentence. Second sentence here.", chunks, embedder, chunkVectors)
 	if len(cited) == 0 {
 		t.Fatalf("expected citations, got none. answer=%q", answer)
 	}
@@ -265,28 +267,32 @@ func TestInsertCitations_Happy(t *testing.T) {
 }
 
 func TestInsertCitations_EmptyAnswer(t *testing.T) {
-	c, _ := InsertCitations("", nil, &fakeEmbedder{}, nil)
+	ctx := t.Context()
+	c, _ := InsertCitations(ctx, "", nil, &fakeEmbedder{}, nil)
 	if c != "" {
 		t.Errorf("empty answer: %q", c)
 	}
 }
 
 func TestInsertCitations_NoChunks(t *testing.T) {
-	c, _ := InsertCitations("Hello world.", nil, &fakeEmbedder{}, [][]float64{})
+	ctx := t.Context()
+	c, _ := InsertCitations(ctx, "Hello world.", nil, &fakeEmbedder{}, [][]float64{})
 	if c != "Hello world." {
 		t.Errorf("no chunks should return original: %q", c)
 	}
 }
 
 func TestInsertCitations_EncodeError(t *testing.T) {
-	c, _ := InsertCitations("Hello world.", []SourcedChunk{{ID: "c1"}}, &fakeEmbedder{err: fmt.Errorf("offline")}, [][]float64{{1, 0}})
+	ctx := t.Context()
+	c, _ := InsertCitations(ctx, "Hello world.", []SourcedChunk{{ID: "c1"}}, &fakeEmbedder{err: fmt.Errorf("offline")}, [][]float64{{1, 0}})
 	if c != "Hello world." {
 		t.Errorf("encode error should return original: %q", c)
 	}
 }
 
 func TestInsertCitations_EncodeEmpty(t *testing.T) {
-	c, _ := InsertCitations("Hello world.", []SourcedChunk{{ID: "c1"}}, &fakeEmbedder{vecs: [][]float64{}}, [][]float64{{1, 0}})
+	ctx := t.Context()
+	c, _ := InsertCitations(ctx, "Hello world.", []SourcedChunk{{ID: "c1"}}, &fakeEmbedder{vecs: [][]float64{}}, [][]float64{{1, 0}})
 	if c != "Hello world." {
 		t.Errorf("empty encode result should return original: %q", c)
 	}
diff --git a/internal/service/dataset/index.go b/internal/service/dataset/index.go
index ae92cda43c..0e5c1cf75e 100644
--- a/internal/service/dataset/index.go
+++ b/internal/service/dataset/index.go
@@ -419,7 +419,7 @@ type embeddingCheckSample struct {
 	QuestionKeywords  []string
 }
 
-func (d *DatasetService) CheckEmbedding(userID, datasetID string, req *service.CheckEmbeddingRequest) (*service.EmbeddingCheckResponse, common.ErrorCode, error) {
+func (d *DatasetService) CheckEmbedding(ctx context.Context, userID, datasetID string, req *service.CheckEmbeddingRequest) (*service.EmbeddingCheckResponse, common.ErrorCode, error) {
 	if datasetID == "" {
 		return nil, common.CodeDataError, errors.New(`Lack of "Dataset ID"`)
 	}
@@ -493,14 +493,14 @@ func (d *DatasetService) CheckEmbedding(userID, datasetID string, req *service.C
 
 		var titleVector [][]float64
 		if title != "" {
-			titleVector, err = datasetEncodeEmbedding(embeddingModel, []string{title})
+			titleVector, err = datasetEncodeEmbedding(ctx, embeddingModel, []string{title})
 			if err != nil {
 				return nil, common.CodeServerError, err
 			}
 		}
 		var contentVector [][]float64
 		if content != "" {
-			contentVector, err = datasetEncodeEmbedding(embeddingModel, []string{content})
+			contentVector, err = datasetEncodeEmbedding(ctx, embeddingModel, []string{content})
 			if err != nil {
 				return nil, common.CodeServerError, err
 			}
diff --git a/internal/service/dataset/utils.go b/internal/service/dataset/utils.go
index bf49a00c62..a9afbbda56 100644
--- a/internal/service/dataset/utils.go
+++ b/internal/service/dataset/utils.go
@@ -1,6 +1,7 @@
 package dataset
 
 import (
+	"context"
 	"encoding/json"
 	"math"
 	"regexp"
@@ -228,7 +229,7 @@ func datasetCleanEmbeddingText(s string) string {
 	return re.ReplaceAllString(s, "")
 }
 
-func datasetEncodeEmbedding(embeddingModel *modelModule.EmbeddingModel, texts []string) ([][]float64, error) {
+func datasetEncodeEmbedding(ctx context.Context, embeddingModel *modelModule.EmbeddingModel, texts []string) ([][]float64, error) {
 	if len(texts) == 0 {
 		return nil, nil
 	}
@@ -237,7 +238,7 @@ func datasetEncodeEmbedding(embeddingModel *modelModule.EmbeddingModel, texts []
 		cleaned[i] = datasetCleanEmbeddingText(t)
 	}
 	embeddingConfig := &modelModule.EmbeddingConfig{Dimension: 0}
-	embeddings, err := embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, cleaned, embeddingModel.APIConfig, embeddingConfig, nil)
+	embeddings, err := embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, cleaned, embeddingModel.APIConfig, embeddingConfig, nil)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/service/deep_researcher.go b/internal/service/deep_researcher.go
index bc9a93193c..24db418b0c 100644
--- a/internal/service/deep_researcher.go
+++ b/internal/service/deep_researcher.go
@@ -671,7 +671,7 @@ func (dr *DeepResearcher) chatOnce(
 		modelModule.Message{Role: "user", Content: userPrompt},
 	}
 	resp, err := dr.ChatModel.ModelDriver.ChatWithMessages(
-		modelName, msgs, dr.ChatModel.APIConfig, cfg, nil,
+		ctx, modelName, msgs, dr.ChatModel.APIConfig, cfg, nil,
 	)
 	if err != nil {
 		return "", err
diff --git a/internal/service/file/file_content.go b/internal/service/file/file_content.go
index ccb4504cf1..b18afd4da1 100644
--- a/internal/service/file/file_content.go
+++ b/internal/service/file/file_content.go
@@ -1,6 +1,7 @@
 package file
 
 import (
+	"context"
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
@@ -96,7 +97,7 @@ func (s *FileService) DownloadAgentFile(tenantID, location string) ([]byte, erro
 // for the given file dicts.
 //   - raw=false: images returned as base64 data URIs in images; non-images parsed and returned as text.
 //   - raw=true:  images returned as raw bytes in images; non-images parsed and returned as text.
-func (s *FileService) GetFileContents(uid string, fileDicts []map[string]interface{}, raw bool) (texts []string, images []string, err error) {
+func (s *FileService) GetFileContents(ctx context.Context, uid string, fileDicts []map[string]interface{}, raw bool) (texts []string, images []string, err error) {
 	storageImpl := storage.GetStorageFactory().GetStorage()
 	if storageImpl == nil {
 		return nil, nil, fmt.Errorf("storage not initialized")
@@ -128,7 +129,7 @@ func (s *FileService) GetFileContents(uid string, fileDicts []map[string]interfa
 				images = append(images, "data:"+mime+";base64,"+base64.StdEncoding.EncodeToString(data))
 			}
 		} else {
-			texts = append(texts, parseFileContent(file.Name, data))
+			texts = append(texts, parseFileContent(ctx, file.Name, data))
 		}
 	}
 	return texts, images, nil
@@ -136,7 +137,7 @@ func (s *FileService) GetFileContents(uid string, fileDicts []map[string]interfa
 
 // parseAgentUploads resolves descriptors returned by upload_info from the
 // caller's downloads bucket and converts them to sys.files values.
-func (s *FileService) ParseAgentUploads(userID string, fileDicts []map[string]interface{}, layoutRecognize string) ([]string, error) {
+func (s *FileService) ParseAgentUploads(ctx context.Context, userID string, fileDicts []map[string]interface{}, layoutRecognize string) ([]string, error) {
 	storageImpl := storage.GetStorageFactory().GetStorage()
 	if storageImpl == nil {
 		return nil, fmt.Errorf("storage not initialized")
@@ -169,7 +170,7 @@ func (s *FileService) ParseAgentUploads(userID string, fileDicts []map[string]in
 			continue
 		}
 
-		content, err := parseAgentUploadContent(name, data, layoutRecognize)
+		content, err := parseAgentUploadContent(ctx, name, data, layoutRecognize)
 		if err != nil {
 			return nil, fmt.Errorf("file %q: parse upload: %w", name, err)
 		}
@@ -178,7 +179,7 @@ func (s *FileService) ParseAgentUploads(userID string, fileDicts []map[string]in
 	return contents, nil
 }
 
-func parseAgentUploadContent(filename string, data []byte, layoutRecognize string) (string, error) {
+func parseAgentUploadContent(ctx context.Context, filename string, data []byte, layoutRecognize string) (string, error) {
 	content := string(data)
 	fileType := utility.GetFileType(filename)
 	if fileType != utility.FileTypeOTHER {
@@ -189,7 +190,7 @@ func parseAgentUploadContent(filename string, data []byte, layoutRecognize strin
 		if configurable, ok := fp.(interface{ ConfigureFromSetup(map[string]any) }); ok {
 			configurable.ConfigureFromSetup(map[string]any{"layout_recognize": layoutRecognize})
 		}
-		res := fp.ParseWithResult(filename, data)
+		res := fp.ParseWithResult(ctx, filename, data)
 		if res.Err != nil {
 			return "", res.Err
 		}
@@ -221,7 +222,7 @@ func parseAgentUploadContent(filename string, data []byte, layoutRecognize strin
 
 // parseFileContent tries to parse a file's contents using the appropriate parser.
 // Falls back to returning raw text if no parser is available.
-func parseFileContent(filename string, data []byte) string {
+func parseFileContent(ctx context.Context, filename string, data []byte) string {
 	fileType := utility.GetFileType(filename)
 	if fileType == utility.FileTypeOTHER {
 		return string(data)
@@ -230,7 +231,7 @@ func parseFileContent(filename string, data []byte) string {
 	if err != nil {
 		return string(data)
 	}
-	res := fp.ParseWithResult(filename, data)
+	res := fp.ParseWithResult(ctx, filename, data)
 	if res.Err != nil {
 		return string(data)
 	}
diff --git a/internal/service/file/file_content_test.go b/internal/service/file/file_content_test.go
index e50eef9cc2..9fc4ca7f40 100644
--- a/internal/service/file/file_content_test.go
+++ b/internal/service/file/file_content_test.go
@@ -39,8 +39,9 @@ func TestLooksLikeHTML(t *testing.T) {
 }
 
 func TestParseFileContent_HTMLOutputFormat(t *testing.T) {
+	ctx := t.Context()
 	// CSV parser produces OutputFormat "html".
-	result := parseFileContent("data.csv", []byte("a,b,c\n1,2,3\n"))
+	result := parseFileContent(ctx, "data.csv", []byte("a,b,c\n1,2,3\n"))
 	if result == "" || result == string([]byte("a,b,c\n1,2,3\n")) {
 		t.Skip("CSV parser not available or returned raw text; integration-only test")
 	}
@@ -51,8 +52,9 @@ func TestParseFileContent_HTMLOutputFormat(t *testing.T) {
 }
 
 func TestParseFileContent_MarkdownOutputFormat(t *testing.T) {
+	ctx := t.Context()
 	// .md files fall through to default parser or produce text/markdown.
-	result := parseFileContent("doc.md", []byte("# Title\n\nBody"))
+	result := parseFileContent(ctx, "doc.md", []byte("# Title\n\nBody"))
 	// This is integration-dependent; verify it doesn't crash and returns something.
 	if result == "" {
 		t.Skip("markdown parser returned empty (integration-only)")
diff --git a/internal/service/file/file_test.go b/internal/service/file/file_test.go
index 06e7257f59..4018aaca1a 100644
--- a/internal/service/file/file_test.go
+++ b/internal/service/file/file_test.go
@@ -187,6 +187,7 @@ func setupFileContentPermissionDB(t *testing.T, accessible bool) {
 }
 
 func TestFileService_GetFileContents_NotAccessible(t *testing.T) {
+	ctx := t.Context()
 	setupFileContentPermissionDB(t, false)
 
 	orig := testFilePerm
@@ -200,7 +201,7 @@ func TestFileService_GetFileContents_NotAccessible(t *testing.T) {
 	t.Cleanup(func() { factory.SetStorage(originalStorage) })
 
 	svc := testFileService()
-	texts, images, err := svc.GetFileContents("user-1", []map[string]interface{}{{"id": "file-1"}}, false)
+	texts, images, err := svc.GetFileContents(ctx, "user-1", []map[string]interface{}{{"id": "file-1"}}, false)
 	if err == nil {
 		t.Fatal("expected authorization error")
 	}
@@ -216,6 +217,7 @@ func TestFileService_GetFileContents_NotAccessible(t *testing.T) {
 }
 
 func TestFileService_GetFileContents_Accessible(t *testing.T) {
+	ctx := t.Context()
 	setupFileContentPermissionDB(t, true)
 
 	mockStorage := &fakeStorage{blob: []byte("allowed content")}
@@ -225,7 +227,7 @@ func TestFileService_GetFileContents_Accessible(t *testing.T) {
 	t.Cleanup(func() { factory.SetStorage(originalStorage) })
 
 	svc := testFileService()
-	texts, images, err := svc.GetFileContents("user-1", []map[string]interface{}{{"id": "file-1"}}, false)
+	texts, images, err := svc.GetFileContents(ctx, "user-1", []map[string]interface{}{{"id": "file-1"}}, false)
 	if err != nil {
 		t.Fatalf("GetFileContents failed: %v", err)
 	}
@@ -244,6 +246,7 @@ func TestFileService_GetFileContents_Accessible(t *testing.T) {
 }
 
 func TestFileService_ParseAgentUploads_TextAndImageInRequestOrder(t *testing.T) {
+	ctx := t.Context()
 	memory := storage.NewMemoryStorage()
 	if err := memory.Put("user-1-downloads", "text-id", []byte("uploaded text")); err != nil {
 		t.Fatalf("put text: %v", err)
@@ -256,7 +259,7 @@ func TestFileService_ParseAgentUploads_TextAndImageInRequestOrder(t *testing.T)
 	factory.SetStorage(memory)
 	t.Cleanup(func() { factory.SetStorage(originalStorage) })
 
-	contents, err := testFileService().ParseAgentUploads("user-1", []map[string]interface{}{
+	contents, err := testFileService().ParseAgentUploads(ctx, "user-1", []map[string]interface{}{
 		{"id": "text-id", "name": "notes.txt", "mime_type": "text/plain", "created_by": "user-1"},
 		{"id": "image-id", "name": "photo.bin", "mime_type": "image/png", "created_by": "user-1"},
 	}, "Plain Text")
@@ -275,13 +278,14 @@ func TestFileService_ParseAgentUploads_TextAndImageInRequestOrder(t *testing.T)
 }
 
 func TestFileService_ParseAgentUploads_RejectsForeignOwner(t *testing.T) {
+	ctx := t.Context()
 	memory := storage.NewMemoryStorage()
 	factory := storage.GetStorageFactory()
 	originalStorage := factory.GetStorage()
 	factory.SetStorage(memory)
 	t.Cleanup(func() { factory.SetStorage(originalStorage) })
 
-	_, err := testFileService().ParseAgentUploads("user-1", []map[string]interface{}{
+	_, err := testFileService().ParseAgentUploads(ctx, "user-1", []map[string]interface{}{
 		{"id": "file-id", "name": "secret.txt", "mime_type": "text/plain", "created_by": "user-2"},
 	}, "")
 	if err == nil || !strings.Contains(err.Error(), "created_by does not match") {
@@ -290,13 +294,14 @@ func TestFileService_ParseAgentUploads_RejectsForeignOwner(t *testing.T) {
 }
 
 func TestFileService_ParseAgentUploads_MissingObjectFails(t *testing.T) {
+	ctx := t.Context()
 	memory := storage.NewMemoryStorage()
 	factory := storage.GetStorageFactory()
 	originalStorage := factory.GetStorage()
 	factory.SetStorage(memory)
 	t.Cleanup(func() { factory.SetStorage(originalStorage) })
 
-	_, err := testFileService().ParseAgentUploads("user-1", []map[string]interface{}{
+	_, err := testFileService().ParseAgentUploads(ctx, "user-1", []map[string]interface{}{
 		{"id": "missing", "name": "missing.txt", "mime_type": "text/plain", "created_by": "user-1"},
 	}, "")
 	if err == nil || !strings.Contains(err.Error(), "read upload") {
diff --git a/internal/service/generator.go b/internal/service/generator.go
index 18e7bb70ce..f22a85122d 100644
--- a/internal/service/generator.go
+++ b/internal/service/generator.go
@@ -73,7 +73,7 @@ func KeywordExtraction(ctx context.Context, chatModel *modelModule.ChatModel, co
 	}
 
 	// Call LLM using ChatModel
-	response, err := chatModel.ModelDriver.ChatWithMessages(*chatModel.ModelName, messages, chatModel.APIConfig, modelConfig, nil)
+	response, err := chatModel.ModelDriver.ChatWithMessages(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, modelConfig, nil)
 	if err != nil {
 		return "", fmt.Errorf("failed to extract keywords: %w", err)
 	}
@@ -176,7 +176,7 @@ func CrossLanguages(ctx context.Context, tenantID string, llmID string, query st
 	}
 
 	// Call LLM using ChatModel
-	response, err := chatModel.ModelDriver.ChatWithMessages(*chatModel.ModelName, messages, chatModel.APIConfig, modelConfig, nil)
+	response, err := chatModel.ModelDriver.ChatWithMessages(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, modelConfig, nil)
 	if err != nil {
 		return query, fmt.Errorf("failed to translate question: %w", err)
 	}
@@ -341,7 +341,7 @@ func FullQuestion(
 		{Role: "user", Content: "Output: "},
 	}
 	resp, err := chatModel.ModelDriver.ChatWithMessages(
-		modelName, msgs, chatModel.APIConfig, nil, nil,
+		ctx, modelName, msgs, chatModel.APIConfig, nil, nil,
 	)
 	if err != nil {
 		return fallbackToLatestUser(messages), err
diff --git a/internal/service/graph/pipeline.go b/internal/service/graph/pipeline.go
index 07a1b5506c..4b2f481df7 100644
--- a/internal/service/graph/pipeline.go
+++ b/internal/service/graph/pipeline.go
@@ -23,11 +23,12 @@ import (
 	"strings"
 	"sync"
 
-	"go.uber.org/zap"
 	"ragflow/internal/common"
 	"ragflow/internal/engine"
 	"ragflow/internal/engine/types"
 	modelModule "ragflow/internal/entity/models"
+
+	"go.uber.org/zap"
 )
 
 // Pipeline encapsulates the knowledge graph retrieval pipeline.
@@ -131,7 +132,7 @@ func (p *Pipeline) Retrieval(ctx context.Context) (map[string]interface{}, error
 		data, _ := json.Marshal(typeSamples)
 		ty2entsJSON = string(data)
 	}
-	typeKeywords, entities := queryRewrite(p.chatModel, p.question, ty2entsJSON)
+	typeKeywords, entities := queryRewrite(ctx, p.chatModel, p.question, ty2entsJSON)
 
 	// 2-4. Search entities, types, and relations in parallel (mutually independent)
 	var (
@@ -201,7 +202,7 @@ func (p *Pipeline) searchEntities(ctx context.Context, entities []string) (map[s
 		Filter:       map[string]interface{}{"knowledge_graph_kwd": "entity"},
 	}
 	if len(entities) > 0 {
-		entsReq.MatchExprs = buildSearchExprs(p.embModel, &types.MatchTextExpr{
+		entsReq.MatchExprs = buildSearchExprs(ctx, p.embModel, &types.MatchTextExpr{
 			Fields:       []string{"entity_kwd^10", "content_ltks^2"},
 			MatchingText: strings.Join(entities, " "),
 			TopN:         50,
@@ -263,7 +264,7 @@ func (p *Pipeline) searchRelations(ctx context.Context, entities []string) map[E
 		Filter:       map[string]interface{}{"knowledge_graph_kwd": "relation"},
 	}
 	if len(entities) > 0 {
-		relsReq.MatchExprs = buildSearchExprs(p.embModel, &types.MatchTextExpr{
+		relsReq.MatchExprs = buildSearchExprs(ctx, p.embModel, &types.MatchTextExpr{
 			Fields:       []string{"content_ltks", "from_entity_kwd", "to_entity_kwd"},
 			MatchingText: strings.Join(entities, " "),
 			TopN:         50,
diff --git a/internal/service/graph/retrieval.go b/internal/service/graph/retrieval.go
index 7bbb276d2f..275bf82dee 100644
--- a/internal/service/graph/retrieval.go
+++ b/internal/service/graph/retrieval.go
@@ -194,12 +194,12 @@ func relationFromChunk(chunk map[string]interface{}) (Edge, KGRelation) {
 // When embModel is nil, returns text-only match expression.
 // When embModel is non-nil, embeds the question and returns hybrid
 // (text + dense + fusion) expressions for vector+keyword search.
-func buildSearchExprs(embModel *modelModule.EmbeddingModel, matchText *types.MatchTextExpr, simThreshold float64, denseTopK int) []interface{} {
+func buildSearchExprs(ctx context.Context, embModel *modelModule.EmbeddingModel, matchText *types.MatchTextExpr, simThreshold float64, denseTopK int) []interface{} {
 	if embModel == nil || embModel.ModelDriver == nil {
 		return []interface{}{matchText}
 	}
 	embeddingConfig := &modelModule.EmbeddingConfig{Dimension: 0}
-	embeddings, err := embModel.ModelDriver.Embed(embModel.ModelName, []string{matchText.MatchingText}, embModel.APIConfig, embeddingConfig, nil)
+	embeddings, err := embModel.ModelDriver.Embed(ctx, embModel.ModelName, []string{matchText.MatchingText}, embModel.APIConfig, embeddingConfig, nil)
 	if err != nil || len(embeddings) == 0 {
 		return []interface{}{matchText}
 	}
@@ -233,7 +233,7 @@ func buildFusionExpr(textWeight, vectorWeight float64, topN int) *types.FusionEx
 }
 
 // queryRewrite attempts LLM-based query rewrite, falling back to raw question.
-func queryRewrite(chatModel *modelModule.ChatModel, question string, ty2entsJSON string) (typeKeywords, entities []string) {
+func queryRewrite(ctx context.Context, chatModel *modelModule.ChatModel, question string, ty2entsJSON string) (typeKeywords, entities []string) {
 	if question == "" {
 		return nil, nil
 	}
@@ -243,7 +243,7 @@ func queryRewrite(chatModel *modelModule.ChatModel, question string, ty2entsJSON
 			{Role: "system", Content: prompt},
 			{Role: "user", Content: "Output:"},
 		}
-		response, err := chatModel.ModelDriver.ChatWithMessages(*chatModel.ModelName, messages, chatModel.APIConfig, nil, nil)
+		response, err := chatModel.ModelDriver.ChatWithMessages(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, nil, nil)
 		if err == nil && response != nil && response.Answer != nil {
 			result, parseErr := common.ParseQueryRewriteResponse(*response.Answer)
 			if parseErr == nil && result != nil {
diff --git a/internal/service/graph/retrieval_test.go b/internal/service/graph/retrieval_test.go
index 26ce25d6ad..f105b5fe5e 100644
--- a/internal/service/graph/retrieval_test.go
+++ b/internal/service/graph/retrieval_test.go
@@ -297,7 +297,8 @@ func (e *searchCaptureEngine) Search(ctx context.Context, req *types.SearchReque
 // --- queryRewrite ---
 
 func TestQueryRewrite_Fallback(t *testing.T) {
-	typeKeywords, entities := queryRewrite(nil, "What is SpaceX?", "{}")
+	ctx := t.Context()
+	typeKeywords, entities := queryRewrite(ctx, nil, "What is SpaceX?", "{}")
 	if typeKeywords != nil {
 		t.Errorf("expected nil typeKeywords when no LLM, got %v", typeKeywords)
 	}
@@ -307,7 +308,8 @@ func TestQueryRewrite_Fallback(t *testing.T) {
 }
 
 func TestQueryRewrite_EmptyQuestion(t *testing.T) {
-	typeKeywords, entities := queryRewrite(nil, "", "")
+	ctx := t.Context()
+	typeKeywords, entities := queryRewrite(ctx, nil, "", "")
 	if typeKeywords != nil || entities != nil {
 		t.Errorf("expected nil for empty question, got type=%v entities=%v", typeKeywords, entities)
 	}
@@ -322,7 +324,7 @@ type spyEmbedDriver struct {
 	err           error
 }
 
-func (s *spyEmbedDriver) Embed(_ *string, texts []string, _ *modelModule.APIConfig, _ *modelModule.EmbeddingConfig, _ *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
+func (s *spyEmbedDriver) Embed(ctx context.Context, _ *string, texts []string, _ *modelModule.APIConfig, _ *modelModule.EmbeddingConfig, _ *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
 	s.capturedTexts = texts
 	if s.err != nil {
 		return nil, s.err
@@ -390,12 +392,13 @@ func TestBuildFusionExpr_AsymmetricWeights(t *testing.T) {
 // --- buildSearchExprs ---
 
 func TestBuildSearchExprs_NoEmbModel(t *testing.T) {
+	ctx := t.Context()
 	matchText := &types.MatchTextExpr{
 		Fields:       []string{"entity_kwd^10"},
 		MatchingText: "test",
 		TopN:         10,
 	}
-	exprs := buildSearchExprs(nil, matchText, 0, 0)
+	exprs := buildSearchExprs(ctx, nil, matchText, 0, 0)
 	if len(exprs) != 1 {
 		t.Fatalf("expected 1 expr, got %d", len(exprs))
 	}
@@ -409,6 +412,7 @@ func TestBuildSearchExprs_NoEmbModel(t *testing.T) {
 }
 
 func TestBuildSearchExprs_WithEmbModel(t *testing.T) {
+	ctx := t.Context()
 	driver := &spyEmbedDriver{vector: []float64{0.1, 0.2, 0.3}}
 	embModel := modelModule.NewEmbeddingModel(driver, strPtr("text-embedding"), &modelModule.APIConfig{}, 512)
 	matchText := &types.MatchTextExpr{
@@ -416,7 +420,7 @@ func TestBuildSearchExprs_WithEmbModel(t *testing.T) {
 		MatchingText: "Elon Musk SpaceX",
 		TopN:         50,
 	}
-	exprs := buildSearchExprs(embModel, matchText, defaultSimThreshold, defaultDenseTopK)
+	exprs := buildSearchExprs(ctx, embModel, matchText, defaultSimThreshold, defaultDenseTopK)
 	// Verify Embed was called with matchText.MatchingText, not raw question
 	if len(driver.capturedTexts) != 1 || driver.capturedTexts[0] != "Elon Musk SpaceX" {
 		t.Errorf("expected Embed to receive %q, got %v", "Elon Musk SpaceX", driver.capturedTexts)
@@ -457,6 +461,7 @@ func TestBuildSearchExprs_WithEmbModel(t *testing.T) {
 }
 
 func TestBuildSearchExprs_EmbModelFallback(t *testing.T) {
+	ctx := t.Context()
 	driver := &spyEmbedDriver{err: assertError("embed failed")}
 	embModel := modelModule.NewEmbeddingModel(driver, strPtr("text-embedding"), &modelModule.APIConfig{}, 512)
 	matchText := &types.MatchTextExpr{
@@ -464,7 +469,7 @@ func TestBuildSearchExprs_EmbModelFallback(t *testing.T) {
 		MatchingText: "fallback test",
 		TopN:         10,
 	}
-	exprs := buildSearchExprs(embModel, matchText, defaultSimThreshold, defaultDenseTopK)
+	exprs := buildSearchExprs(ctx, embModel, matchText, defaultSimThreshold, defaultDenseTopK)
 	// Should fall back to text-only when Embed fails
 	if len(exprs) != 1 {
 		t.Fatalf("expected 1 expr (text-only fallback), got %d", len(exprs))
diff --git a/internal/service/graph/search.go b/internal/service/graph/search.go
index 21a1ec64bb..ba27543dd2 100644
--- a/internal/service/graph/search.go
+++ b/internal/service/graph/search.go
@@ -37,7 +37,7 @@ func NhopEntityNames(nHopJSON string) []string {
 
 // SearchEntities searches for KG entities matching a question.
 func SearchEntities(ctx context.Context, docEngine engine.DocEngine, kbIDs []string, question string, embModel *modelModule.EmbeddingModel, topN int) ([]KGEntity, error) {
-	dense, err := buildDenseExpr(embModel, question, topN)
+	dense, err := buildDenseExpr(ctx, embModel, question, topN)
 	if err != nil {
 		return nil, err
 	}
@@ -61,7 +61,7 @@ func SearchEntitiesByTypes(ctx context.Context, docEngine engine.DocEngine, kbID
 
 // SearchRelations searches for KG relations matching a question.
 func SearchRelations(ctx context.Context, docEngine engine.DocEngine, kbIDs []string, question string, embModel *modelModule.EmbeddingModel, topN int) ([]KGRelation, error) {
-	dense, err := buildDenseExpr(embModel, question, topN)
+	dense, err := buildDenseExpr(ctx, embModel, question, topN)
 	if err != nil {
 		return nil, err
 	}
@@ -94,12 +94,12 @@ func SearchTypeSamples(ctx context.Context, docEngine engine.DocEngine, kbIDs []
 }
 
 // buildDenseExpr computes the query vector and returns a MatchDenseExpr.
-func buildDenseExpr(embModel *modelModule.EmbeddingModel, question string, topN int) (*types.MatchDenseExpr, error) {
+func buildDenseExpr(ctx context.Context, embModel *modelModule.EmbeddingModel, question string, topN int) (*types.MatchDenseExpr, error) {
 	if embModel == nil || question == "" {
 		return nil, nil
 	}
 	embCfg := &modelModule.EmbeddingConfig{Dimension: 0}
-	embeddings, err := embModel.ModelDriver.Embed(embModel.ModelName, []string{question}, embModel.APIConfig, embCfg, nil)
+	embeddings, err := embModel.ModelDriver.Embed(ctx, embModel.ModelName, []string{question}, embModel.APIConfig, embCfg, nil)
 	if err != nil {
 		return nil, fmt.Errorf("KG entity embed failed: %w", err)
 	}
diff --git a/internal/service/graph/search_test.go b/internal/service/graph/search_test.go
index 1eb59e3f4c..6b77bd0c41 100644
--- a/internal/service/graph/search_test.go
+++ b/internal/service/graph/search_test.go
@@ -344,11 +344,12 @@ type fakeEmbedDriver struct {
 
 func (f *fakeEmbedDriver) Name() string { return f.name }
 
-func (f *fakeEmbedDriver) Embed(modelName *string, texts []string, apiConfig *modelModule.APIConfig, config *modelModule.EmbeddingConfig, usage *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
+func (f *fakeEmbedDriver) Embed(ctx context.Context, modelName *string, texts []string, apiConfig *modelModule.APIConfig, config *modelModule.EmbeddingConfig, usage *common.ModelUsage) ([]modelModule.EmbeddingData, error) {
 	return []modelModule.EmbeddingData{{Embedding: f.vector}}, nil
 }
 
 func TestBuildKGDenseExpr_WithModel(t *testing.T) {
+	ctx := t.Context()
 	embModel := &modelModule.EmbeddingModel{
 		ModelName: strPtr("test-model"),
 		ModelDriver: &fakeEmbedDriver{
@@ -357,7 +358,7 @@ func TestBuildKGDenseExpr_WithModel(t *testing.T) {
 		},
 		APIConfig: &modelModule.APIConfig{},
 	}
-	dense, err := buildDenseExpr(embModel, "test question", 10)
+	dense, err := buildDenseExpr(ctx, embModel, "test question", 10)
 	if err != nil {
 		t.Fatalf("buildDenseExpr failed: %v", err)
 	}
@@ -373,14 +374,16 @@ func TestBuildKGDenseExpr_WithModel(t *testing.T) {
 }
 
 func TestBuildKGDenseExpr_NilModel(t *testing.T) {
-	dense, err := buildDenseExpr(nil, "test", 10)
+	ctx := t.Context()
+	dense, err := buildDenseExpr(ctx, nil, "test", 10)
 	if dense != nil || err != nil {
 		t.Errorf("expected nil,nil for nil model, got dense=%v err=%v", dense, err)
 	}
 }
 
 func TestBuildKGDenseExpr_EmptyQuestion(t *testing.T) {
-	dense, err := buildDenseExpr(&modelModule.EmbeddingModel{}, "", 10)
+	ctx := t.Context()
+	dense, err := buildDenseExpr(ctx, &modelModule.EmbeddingModel{}, "", 10)
 	if dense != nil || err != nil {
 		t.Errorf("expected nil,nil for empty question, got dense=%v err=%v", dense, err)
 	}
diff --git a/internal/service/memory.go b/internal/service/memory.go
index 12b28d727e..1d8e67090e 100644
--- a/internal/service/memory.go
+++ b/internal/service/memory.go
@@ -1045,7 +1045,7 @@ func (s *MemoryService) queryMessage(ctx context.Context, memories []*entity.Mem
 	matchExprs := make([]interface{}, 0, 3)
 	if question != "" {
 		matchText := memoryMessageTextExpr(question, similarityThreshold)
-		matchDense, err := s.memoryMessageDenseExpr(question, memories[0], topN, similarityThreshold)
+		matchDense, err := s.memoryMessageDenseExpr(ctx, question, memories[0], topN, similarityThreshold)
 		if err != nil {
 			return nil, common.CodeServerError, err
 		}
@@ -1289,13 +1289,13 @@ func memoryMessageTextExpr(question string, similarityThreshold float64) *engine
 	return matchText
 }
 
-func (s *MemoryService) memoryMessageDenseExpr(question string, memory *entity.Memory, topN int, similarityThreshold float64) (*enginetypes.MatchDenseExpr, error) {
+func (s *MemoryService) memoryMessageDenseExpr(ctx context.Context, question string, memory *entity.Memory, topN int, similarityThreshold float64) (*enginetypes.MatchDenseExpr, error) {
 	driver, modelName, apiConfig, maxTokens, err := NewModelProviderService().ResolveModelConfig(memory.TenantID, entity.ModelTypeEmbedding, memory.EmbdID)
 	if err != nil {
 		return nil, err
 	}
 	embeddingModel := models.NewEmbeddingModel(driver, &modelName, apiConfig, maxTokens)
-	embeddings, err := embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{question}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
+	embeddings, err := embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{question}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/service/memory_message_service.go b/internal/service/memory_message_service.go
index 27c84f53ad..78a727a997 100644
--- a/internal/service/memory_message_service.go
+++ b/internal/service/memory_message_service.go
@@ -267,7 +267,7 @@ func (s *MemoryMessageService) embedAndSave(ctx context.Context, mem *CreateMemo
 		return err
 	}
 	embeddingModel := models.NewEmbeddingModel(driver, &modelName, apiConfig, maxTokens)
-	embeddings, err := embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{content}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
+	embeddings, err := embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{content}, embeddingModel.APIConfig, &models.EmbeddingConfig{Dimension: 0}, nil)
 	if err != nil {
 		return err
 	}
diff --git a/internal/service/metadata_filter.go b/internal/service/metadata_filter.go
index decc83a06a..f27e7975f1 100644
--- a/internal/service/metadata_filter.go
+++ b/internal/service/metadata_filter.go
@@ -204,7 +204,7 @@ func GenMetaFilter(ctx context.Context, chatModel *modelModule.ChatModel, metaDa
 	}
 
 	// Call LLM using ChatModel
-	response, err := chatModel.ModelDriver.ChatWithMessages(*chatModel.ModelName, messages, chatModel.APIConfig, nil, nil)
+	response, err := chatModel.ModelDriver.ChatWithMessages(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, nil, nil)
 	if err != nil {
 		common.Warn("ChatWithMessages failed for GenMetaFilter",
 			zap.String("model",
diff --git a/internal/service/model_chat.go b/internal/service/model_chat.go
index d5edb89469..ac1946e01d 100644
--- a/internal/service/model_chat.go
+++ b/internal/service/model_chat.go
@@ -34,12 +34,12 @@ const streamDoneSentinel = "[DONE]"
 // errStreamDone aborts the driver loop once the terminal sentinel arrives.
 var errStreamDone = errors.New("chat stream done")
 
-func (m *ModelProviderService) Chat(tenantID, modelID string, messages []modelModule.Message, config *modelModule.ChatConfig) (*modelModule.ChatResponse, error) {
+func (m *ModelProviderService) Chat(ctx context.Context, tenantID, modelID string, messages []modelModule.Message, config *modelModule.ChatConfig) (*modelModule.ChatResponse, error) {
 	chatModel, err := m.GetChatModel(tenantID, modelID)
 	if err != nil {
 		return nil, err
 	}
-	return chatModel.ModelDriver.ChatWithMessages(*chatModel.ModelName, messages, chatModel.APIConfig, config, nil)
+	return chatModel.ModelDriver.ChatWithMessages(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, config, nil)
 }
 
 func (m *ModelProviderService) ChatStream(ctx context.Context, tenantID, modelID string, messages []modelModule.Message, config *modelModule.ChatConfig) (<-chan string, error) {
@@ -54,7 +54,7 @@ func chatStreamWithContext(ctx context.Context, chatModel *modelModule.ChatModel
 	ch := make(chan string, 256)
 	go func() {
 		defer close(ch)
-		if err := chatModel.ModelDriver.ChatStreamlyWithSender(*chatModel.ModelName, messages, chatModel.APIConfig, config, nil,
+		if err := chatModel.ModelDriver.ChatStreamlyWithSender(ctx, *chatModel.ModelName, messages, chatModel.APIConfig, config, nil,
 			func(delta *string, _ *string) error {
 				if delta == nil {
 					return nil
diff --git a/internal/service/model_chat_test.go b/internal/service/model_chat_test.go
index 215299b86f..890821fc52 100644
--- a/internal/service/model_chat_test.go
+++ b/internal/service/model_chat_test.go
@@ -32,7 +32,7 @@ type doneEmittingDriver struct {
 	deltas []string
 }
 
-func (d *doneEmittingDriver) ChatStreamlyWithSender(modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
+func (d *doneEmittingDriver) ChatStreamlyWithSender(ctx context.Context, modelName string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
 	for _, delta := range d.deltas {
 		if err := sender(&delta, nil); err != nil {
 			return err
diff --git a/internal/service/model_service.go b/internal/service/model_service.go
index 1eca432382..604b0acaf9 100644
--- a/internal/service/model_service.go
+++ b/internal/service/model_service.go
@@ -17,6 +17,7 @@
 package service
 
 import (
+	"context"
 	"encoding/binary"
 	"encoding/json"
 	"errors"
@@ -334,7 +335,7 @@ func (m *ModelProviderService) DeleteModelProvider(userID, providerName string)
 	return common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) ListSupportedModels(providerName, instanceName, userID string) ([]map[string]interface{}, error) {
+func (m *ModelProviderService) ListSupportedModels(ctx context.Context, providerName, instanceName, userID string) ([]map[string]interface{}, error) {
 	providerName = strings.TrimSpace(providerName)
 
 	// Get tenant ID from user
@@ -390,7 +391,7 @@ func (m *ModelProviderService) ListSupportedModels(providerName, instanceName, u
 		}
 	}
 
-	modelList, err := driver.ListModels(apiConfig)
+	modelList, err := driver.ListModels(ctx, apiConfig)
 	if err != nil {
 		return nil, err
 	}
@@ -424,7 +425,7 @@ func (m *ModelProviderService) getProviderByIDOrName(tenantID, providerIDOrName
 	return m.modelProviderDAO.GetByTenantIDAndProviderName(tenantID, strings.TrimSpace(providerIDOrName))
 }
 
-func (m *ModelProviderService) CreateProviderInstance(providerIDOrName, instanceName, apiKey, baseURL, region, userID string, modelInfo []CreateInstanceModelInfo) (common.ErrorCode, error) {
+func (m *ModelProviderService) CreateProviderInstance(ctx context.Context, providerIDOrName, instanceName, apiKey, baseURL, region, userID string, modelInfo []CreateInstanceModelInfo) (common.ErrorCode, error) {
 	providerIDOrName = strings.TrimSpace(providerIDOrName)
 
 	// Get tenant ID from user
@@ -453,7 +454,7 @@ func (m *ModelProviderService) CreateProviderInstance(providerIDOrName, instance
 
 	// Verify the API key against the provider.
 	// Mirrors Python's verify_api_key (provider_api_service.py:596).
-	modelVerifyResult := m.verifyProviderAPIKey(providerName, apiKey, region, baseURL, modelInfo)
+	modelVerifyResult := m.verifyProviderAPIKey(ctx, providerName, apiKey, region, baseURL, modelInfo)
 
 	instanceID := utility.GenerateToken()
 
@@ -582,7 +583,7 @@ func (m *ModelProviderService) CreateNameOnlyProviderInstance(providerIDOrName,
 // verifyProviderAPIKey verifies the API key against the provider by calling
 // the driver's CheckConnection. It returns a map from model name to verify
 // status (success/fail/unknown).
-func (m *ModelProviderService) verifyProviderAPIKey(providerName, apiKey, region, baseURL string, modelInfo []CreateInstanceModelInfo) map[string]string {
+func (m *ModelProviderService) verifyProviderAPIKey(ctx context.Context, providerName, apiKey, region, baseURL string, modelInfo []CreateInstanceModelInfo) map[string]string {
 	result := make(map[string]string)
 
 	providerInfo := dao.GetModelProviderManager().FindProvider(providerName)
@@ -619,7 +620,7 @@ func (m *ModelProviderService) verifyProviderAPIKey(providerName, apiKey, region
 		BaseURL: &baseURL,
 	}
 
-	verifyErr := driver.CheckConnection(apiConfig)
+	verifyErr := driver.CheckConnection(ctx, apiConfig)
 	verifyStatus := entity.ModelVerifySuccess
 	if verifyErr != nil {
 		verifyStatus = entity.ModelVerifyFail
@@ -815,7 +816,7 @@ func (m *ModelProviderService) ShowProviderInstance(providerName, instanceIDOrNa
 	return result, common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) ShowInstanceBalance(providerName, instanceName, userID string) (map[string]interface{}, common.ErrorCode, error) {
+func (m *ModelProviderService) ShowInstanceBalance(ctx context.Context, providerName, instanceName, userID string) (map[string]interface{}, common.ErrorCode, error) {
 
 	// Get tenant ID from user
 	tenants, err := m.userTenantDAO.GetByUserIDAndRole(userID, "owner")
@@ -863,14 +864,14 @@ func (m *ModelProviderService) ShowInstanceBalance(providerName, instanceName, u
 	apiConfig.BaseURL = &baseURL
 
 	var result map[string]interface{}
-	result, err = providerInfo.ModelDriver.Balance(apiConfig)
+	result, err = providerInfo.ModelDriver.Balance(ctx, apiConfig)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
 	return result, common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) CheckConnection(providerName, apiKey, region, baseURL, instanceID, userID string, modelInfo []string) (common.ErrorCode, error) {
+func (m *ModelProviderService) CheckConnection(ctx context.Context, providerName, apiKey, region, baseURL, instanceID, userID string, modelInfo []string) (common.ErrorCode, error) {
 	providerInfo := dao.GetModelProviderManager().FindProvider(providerName)
 	if providerInfo == nil {
 		return common.CodeServerError, fmt.Errorf("provider %s not found", providerName)
@@ -901,14 +902,14 @@ func (m *ModelProviderService) CheckConnection(providerName, apiKey, region, bas
 		Region: &region,
 	}
 
-	err := driver.CheckConnection(apiConfig)
+	err := driver.CheckConnection(ctx, apiConfig)
 	if err != nil {
 		return common.CodeServerError, err
 	}
 
 	// Mirror Python verify_api_key: verify each model by making a real
 	// lightweight API request.  Returns per-model verify results.
-	modelVerifyResult, verifyErr := verifyProviderModel(driver, providerInfo.Models, apiConfig, modelInfo)
+	modelVerifyResult, verifyErr := verifyProviderModel(ctx, driver, providerInfo.Models, apiConfig, modelInfo)
 
 	// When instanceID is provided (frontend passes it), persist the verify
 	// results to the database — mirrors Python's per-model update_model calls
@@ -979,7 +980,7 @@ func (m *ModelProviderService) updateModelVerifyResults(userID, providerName, in
 // and returns a map of modelName → verify status ("success"/"fail") so the
 // caller can persist the results to the database.  A nil error means at least
 // one model passed verification.
-func verifyProviderModel(driver modelModule.ModelDriver, providerModels []*modelModule.Model, apiConfig *modelModule.APIConfig, modelInfo []string) (map[string]string, error) {
+func verifyProviderModel(ctx context.Context, driver modelModule.ModelDriver, providerModels []*modelModule.Model, apiConfig *modelModule.APIConfig, modelInfo []string) (map[string]string, error) {
 	modelVerifyResult := make(map[string]string)
 
 	// Determine which models to verify: prefer the caller-supplied modelInfo
@@ -1025,18 +1026,18 @@ func verifyProviderModel(driver modelModule.ModelDriver, providerModels []*model
 			switch mtLower {
 			case "chat", "vision":
 				msg := []modelModule.Message{{Role: "user", Content: "Hi"}}
-				_, err = driver.ChatWithMessages(modelName, msg, apiConfig, nil, nil)
+				_, err = driver.ChatWithMessages(ctx, modelName, msg, apiConfig, nil, nil)
 			case "embedding":
-				_, err = driver.Embed(&modelName, []string{"test"}, apiConfig, nil, nil)
+				_, err = driver.Embed(ctx, &modelName, []string{"test"}, apiConfig, nil, nil)
 			case "rerank":
-				_, err = driver.Rerank(&modelName, "test", []string{"test"}, apiConfig, &modelModule.RerankConfig{}, nil)
+				_, err = driver.Rerank(ctx, &modelName, "test", []string{"test"}, apiConfig, &modelModule.RerankConfig{}, nil)
 			case "tts":
 				content := "hello"
-				_, err = driver.AudioSpeech(&modelName, &content, apiConfig, nil, nil)
+				_, err = driver.AudioSpeech(ctx, &modelName, &content, apiConfig, nil, nil)
 			case "asr":
-				err = verifyASRModel(driver, modelName, apiConfig)
+				err = verifyASRModel(ctx, driver, modelName, apiConfig)
 			case "ocr":
-				err = verifyOCRModel(driver, modelName, apiConfig)
+				err = verifyOCRModel(ctx, driver, modelName, apiConfig)
 			default:
 				continue
 			}
@@ -1136,7 +1137,7 @@ func generateTestWAV() []byte {
 // verifyASRModel mirrors Python sequence2txt_model.py's check_available:
 // generates a minimal test WAV, writes it to a temp file, calls
 // TranscribeAudio, and checks the result for errors.
-func verifyASRModel(driver modelModule.ModelDriver, modelName string, apiConfig *modelModule.APIConfig) error {
+func verifyASRModel(ctx context.Context, driver modelModule.ModelDriver, modelName string, apiConfig *modelModule.APIConfig) error {
 	wavData := generateTestWAV()
 
 	tmpFile, err := os.CreateTemp("", "ragflow-asr-verify-*.wav")
@@ -1152,7 +1153,7 @@ func verifyASRModel(driver modelModule.ModelDriver, modelName string, apiConfig
 	}
 	tmpFile.Close()
 
-	resp, err := driver.TranscribeAudio(&modelName, &tmpPath, apiConfig, nil, nil)
+	resp, err := driver.TranscribeAudio(ctx, &modelName, &tmpPath, apiConfig, nil, nil)
 	if err != nil {
 		return err
 	}
@@ -1166,11 +1167,11 @@ func verifyASRModel(driver modelModule.ModelDriver, modelName string, apiConfig
 // minimal PNG (1×1 pixel white) through the OCR pipeline.  Most OCR
 // providers return an error or empty text for such a trivial image;
 // we accept any non-error response as a successful connectivity check.
-func verifyOCRModel(driver modelModule.ModelDriver, modelName string, apiConfig *modelModule.APIConfig) error {
+func verifyOCRModel(ctx context.Context, driver modelModule.ModelDriver, modelName string, apiConfig *modelModule.APIConfig) error {
 	// Send a minimal 1×1 white PNG through the OCR pipeline to verify
 	// connectivity. Most OCRModel drivers only check server reachability
 	// rather than performing full document parsing.
-	_, err := driver.OCRFile(&modelName, minimalPNG(), nil, apiConfig, nil, nil)
+	_, err := driver.OCRFile(ctx, &modelName, minimalPNG(), nil, apiConfig, nil, nil)
 	if err != nil {
 		return err
 	}
@@ -1192,7 +1193,7 @@ func minimalPNG() []byte {
 	}
 }
 
-func (m *ModelProviderService) CheckInstanceConnection(providerName, instanceName, userID string) (common.ErrorCode, error) {
+func (m *ModelProviderService) CheckInstanceConnection(ctx context.Context, providerName, instanceName, userID string) (common.ErrorCode, error) {
 
 	// Get tenant ID from user
 	tenants, err := m.userTenantDAO.GetByUserIDAndRole(userID, "owner")
@@ -1245,14 +1246,14 @@ func (m *ModelProviderService) CheckInstanceConnection(providerName, instanceNam
 		}
 	}
 
-	err = driver.CheckConnection(apiConfig)
+	err = driver.CheckConnection(ctx, apiConfig)
 	if err != nil {
 		return common.CodeServerError, err
 	}
 	return common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) ListTasks(providerName, instanceName, userID string) ([]modelModule.ListTaskStatus, common.ErrorCode, error) {
+func (m *ModelProviderService) ListTasks(ctx context.Context, providerName, instanceName, userID string) ([]modelModule.ListTaskStatus, common.ErrorCode, error) {
 
 	// Get tenant ID from user
 	tenants, err := m.userTenantDAO.GetByUserIDAndRole(userID, "owner")
@@ -1306,14 +1307,14 @@ func (m *ModelProviderService) ListTasks(providerName, instanceName, userID stri
 	}
 
 	var listTaskResponse []modelModule.ListTaskStatus
-	listTaskResponse, err = driver.ListTasks(apiConfig)
+	listTaskResponse, err = driver.ListTasks(ctx, apiConfig)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
 	return listTaskResponse, common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) ShowTask(providerName, instanceName, taskID, userID string) (*modelModule.TaskResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) ShowTask(ctx context.Context, providerName, instanceName, taskID, userID string) (*modelModule.TaskResponse, common.ErrorCode, error) {
 
 	// Get tenant ID from user
 	tenants, err := m.userTenantDAO.GetByUserIDAndRole(userID, "owner")
@@ -1367,7 +1368,7 @@ func (m *ModelProviderService) ShowTask(providerName, instanceName, taskID, user
 	}
 
 	var taskResponse *modelModule.TaskResponse
-	taskResponse, err = driver.ShowTask(taskID, apiConfig)
+	taskResponse, err = driver.ShowTask(ctx, taskID, apiConfig)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -1803,7 +1804,7 @@ func (m *ModelProviderService) ensureOCRProviderFromEnv(tenantID, providerName,
 	return nil
 }
 
-func (m *ModelProviderService) AlterProviderInstance(userID, providerIDOrName, instanceIDOrName, newInstanceName, apiKey, baseURL, region string, modelInfo []CreateInstanceModelInfo, verify bool) (common.ErrorCode, error) {
+func (m *ModelProviderService) AlterProviderInstance(ctx context.Context, userID, providerIDOrName, instanceIDOrName, newInstanceName, apiKey, baseURL, region string, modelInfo []CreateInstanceModelInfo, verify bool) (common.ErrorCode, error) {
 	providerIDOrName = strings.TrimSpace(providerIDOrName)
 
 	tenants, err := m.userTenantDAO.GetByUserIDAndRole(userID, "owner")
@@ -1838,7 +1839,7 @@ func (m *ModelProviderService) AlterProviderInstance(userID, providerIDOrName, i
 	// Verify API key if requested.
 	modelVerifyResult := make(map[string]string)
 	if verify {
-		modelVerifyResult = m.verifyProviderAPIKey(providerName, apiKey, region, baseURL, modelInfo)
+		modelVerifyResult = m.verifyProviderAPIKey(ctx, providerName, apiKey, region, baseURL, modelInfo)
 	}
 
 	// Update instance record.
@@ -2571,7 +2572,7 @@ func (m *ModelProviderService) getModelInstanceAndProviderByID(modelID *string,
 }
 
 // ChatToModelWithMessages sends messages to the model with messages array
-func (m *ModelProviderService) ChatToModelWithMessages(providerName, instanceName, modelName, modelID *string, userID string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage) (*modelModule.ChatResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) ChatToModelWithMessages(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage) (*modelModule.ChatResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2630,7 +2631,7 @@ func (m *ModelProviderService) ChatToModelWithMessages(providerName, instanceNam
 	modelUsage.InstanceID = info.InstanceEntity.ID
 	modelUsage.APIKey = info.InstanceEntity.APIKey
 
-	response, err = modelDriver.ChatWithMessages(resolvedModelName, messages, info.APIConfig, modelConfig, modelUsage)
+	response, err = modelDriver.ChatWithMessages(ctx, resolvedModelName, messages, info.APIConfig, modelConfig, modelUsage)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -2642,7 +2643,7 @@ func (m *ModelProviderService) ChatToModelWithMessages(providerName, instanceNam
 }
 
 // ChatToModelStreamWithSender streams chat response directly via sender function ( the best performance, no channel)
-func (m *ModelProviderService) ChatToModelStreamWithSender(providerName, instanceName, modelName, modelID *string, userID string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) (common.ErrorCode, error) {
+func (m *ModelProviderService) ChatToModelStreamWithSender(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, messages []modelModule.Message, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) (common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2697,7 +2698,7 @@ func (m *ModelProviderService) ChatToModelStreamWithSender(providerName, instanc
 	modelUsage.InstanceID = info.InstanceEntity.ID
 	modelUsage.APIKey = info.InstanceEntity.APIKey
 
-	err = modelDriver.ChatStreamlyWithSender(resolvedModelName, messages, info.APIConfig, modelConfig, modelUsage, sender)
+	err = modelDriver.ChatStreamlyWithSender(ctx, resolvedModelName, messages, info.APIConfig, modelConfig, modelUsage, sender)
 	if err != nil {
 		return common.CodeServerError, err
 	}
@@ -2735,7 +2736,7 @@ func validateEmbeddingDimension(model *modelModule.Model, requested int) error {
 }
 
 // EmbedText sends texts to the embedding model
-func (m *ModelProviderService) EmbedText(providerName, instanceName, modelName, modelID *string, userID string, texts []string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.EmbeddingConfig) ([]modelModule.EmbeddingData, common.ErrorCode, error) {
+func (m *ModelProviderService) EmbedText(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, texts []string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.EmbeddingConfig) ([]modelModule.EmbeddingData, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2789,7 +2790,7 @@ func (m *ModelProviderService) EmbedText(providerName, instanceName, modelName,
 	}
 
 	var response []modelModule.EmbeddingData
-	response, err = modelDriver.Embed(&resolvedModelName, texts, info.APIConfig, modelConfig, nil)
+	response, err = modelDriver.Embed(ctx, &resolvedModelName, texts, info.APIConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -2801,7 +2802,7 @@ func (m *ModelProviderService) EmbedText(providerName, instanceName, modelName,
 }
 
 // RerankDocument sends texts to the embedding model
-func (m *ModelProviderService) RerankDocument(providerName, instanceName, modelName, modelID *string, userID, query string, documents []string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.RerankConfig) (*modelModule.RerankResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) RerankDocument(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID, query string, documents []string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.RerankConfig) (*modelModule.RerankResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2851,7 +2852,7 @@ func (m *ModelProviderService) RerankDocument(providerName, instanceName, modelN
 	}
 
 	var response *modelModule.RerankResponse
-	response, err = modelDriver.Rerank(&resolvedModelName, query, documents, info.APIConfig, modelConfig, nil)
+	response, err = modelDriver.Rerank(ctx, &resolvedModelName, query, documents, info.APIConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -2860,7 +2861,7 @@ func (m *ModelProviderService) RerankDocument(providerName, instanceName, modelN
 }
 
 // TranscribeAudio transcribe audio file to text
-func (m *ModelProviderService) TranscribeAudio(providerName, instanceName, modelName, modelID *string, userID string, audioFile *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ASRConfig) (*modelModule.ASRResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) TranscribeAudio(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, audioFile *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ASRConfig) (*modelModule.ASRResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2905,7 +2906,7 @@ func (m *ModelProviderService) TranscribeAudio(providerName, instanceName, model
 	}
 
 	var response *modelModule.ASRResponse
-	response, err = modelDriver.TranscribeAudio(modelName, audioFile, apiConfig, modelConfig, nil)
+	response, err = modelDriver.TranscribeAudio(ctx, modelName, audioFile, apiConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -2917,7 +2918,7 @@ func (m *ModelProviderService) TranscribeAudio(providerName, instanceName, model
 }
 
 // TranscribeAudioStream transcribe audio file to text stream directly via sender function ( the best performance, no channel)
-func (m *ModelProviderService) TranscribeAudioStream(providerName, instanceName, modelName, modelID *string, userID string, audioFile *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ASRConfig, sender func(*string, *string) error) (common.ErrorCode, error) {
+func (m *ModelProviderService) TranscribeAudioStream(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, audioFile *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ASRConfig, sender func(*string, *string) error) (common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -2961,7 +2962,7 @@ func (m *ModelProviderService) TranscribeAudioStream(providerName, instanceName,
 		}
 	}
 
-	err = modelDriver.TranscribeAudioWithSender(modelName, audioFile, apiConfig, modelConfig, nil, sender)
+	err = modelDriver.TranscribeAudioWithSender(ctx, modelName, audioFile, apiConfig, modelConfig, nil, sender)
 	if err != nil {
 		return common.CodeServerError, err
 	}
@@ -2970,7 +2971,7 @@ func (m *ModelProviderService) TranscribeAudioStream(providerName, instanceName,
 }
 
 // AudioSpeech convert audio to speech
-func (m *ModelProviderService) AudioSpeech(providerName, instanceName, modelName, modelID *string, userID string, audioContent *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.TTSConfig) (*modelModule.TTSResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) AudioSpeech(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, audioContent *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.TTSConfig) (*modelModule.TTSResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -3015,7 +3016,7 @@ func (m *ModelProviderService) AudioSpeech(providerName, instanceName, modelName
 	}
 
 	var response *modelModule.TTSResponse
-	response, err = modelDriver.AudioSpeech(modelName, audioContent, apiConfig, modelConfig, nil)
+	response, err = modelDriver.AudioSpeech(ctx, modelName, audioContent, apiConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -3026,7 +3027,7 @@ func (m *ModelProviderService) AudioSpeech(providerName, instanceName, modelName
 	return response, common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) AudioSpeechStream(providerName, instanceName, modelName, modelID *string, userID string, audioContent *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.TTSConfig, sender func(*string, *string) error) (common.ErrorCode, error) {
+func (m *ModelProviderService) AudioSpeechStream(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, audioContent *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.TTSConfig, sender func(*string, *string) error) (common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -3070,7 +3071,7 @@ func (m *ModelProviderService) AudioSpeechStream(providerName, instanceName, mod
 		}
 	}
 
-	err = modelDriver.AudioSpeechWithSender(modelName, audioContent, apiConfig, modelConfig, nil, sender)
+	err = modelDriver.AudioSpeechWithSender(ctx, modelName, audioContent, apiConfig, modelConfig, nil, sender)
 	if err != nil {
 		return common.CodeServerError, err
 	}
@@ -3078,7 +3079,7 @@ func (m *ModelProviderService) AudioSpeechStream(providerName, instanceName, mod
 	return common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) OCRFile(providerName, instanceName, modelName, modelID *string, userID string, content []byte, url *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.OCRConfig) (*modelModule.OCRFileResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) OCRFile(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, content []byte, url *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.OCRConfig) (*modelModule.OCRFileResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -3123,7 +3124,7 @@ func (m *ModelProviderService) OCRFile(providerName, instanceName, modelName, mo
 	}
 
 	var response *modelModule.OCRFileResponse
-	response, err = modelDriver.OCRFile(modelName, content, url, apiConfig, modelConfig, nil)
+	response, err = modelDriver.OCRFile(ctx, modelName, content, url, apiConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -3134,7 +3135,7 @@ func (m *ModelProviderService) OCRFile(providerName, instanceName, modelName, mo
 	return response, common.CodeSuccess, nil
 }
 
-func (m *ModelProviderService) ParseFile(providerName, instanceName, modelName, modelID *string, userID string, content []byte, url *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ParseFileConfig) (*modelModule.ParseFileResponse, common.ErrorCode, error) {
+func (m *ModelProviderService) ParseFile(ctx context.Context, providerName, instanceName, modelName, modelID *string, userID string, content []byte, url *string, apiConfig *modelModule.APIConfig, modelConfig *modelModule.ParseFileConfig) (*modelModule.ParseFileResponse, common.ErrorCode, error) {
 
 	var err error
 	var info *ModelInstanceAndProviderInfo
@@ -3178,7 +3179,7 @@ func (m *ModelProviderService) ParseFile(providerName, instanceName, modelName,
 	}
 
 	var response *modelModule.ParseFileResponse
-	response, err = modelDriver.ParseFile(modelName, content, url, apiConfig, modelConfig, nil)
+	response, err = modelDriver.ParseFile(ctx, modelName, content, url, apiConfig, modelConfig, nil)
 	if err != nil {
 		return nil, common.CodeServerError, err
 	}
@@ -3269,7 +3270,7 @@ func (m *ModelProviderService) GetModelConfigByID(userID string, modelType entit
 	if modelEntity.Status != "active" {
 		return nil, "", nil, 0, fmt.Errorf("tenant model id=%s is disabled", modelID)
 	}
-	if modelType != 0 && !entity.ModelType(modelEntity.ModelType).Has(modelType) {
+	if !entity.ModelType(modelEntity.ModelType).Has(modelType) {
 		return nil, "", nil, 0, fmt.Errorf("tenant model id=%s cannot be used as %s model", modelID, modelType.String())
 	}
 
diff --git a/internal/service/nlp/reranker.go b/internal/service/nlp/reranker.go
index 1e3241ce5b..234e54fdee 100644
--- a/internal/service/nlp/reranker.go
+++ b/internal/service/nlp/reranker.go
@@ -15,6 +15,7 @@
 package nlp
 
 import (
+	"context"
 	"encoding/json"
 	"math"
 	"regexp"
@@ -53,6 +54,7 @@ type SearchResult struct {
 //   - tsim: token similarity scores
 //   - vsim: vector similarity scores
 func Rerank(
+	ctx context.Context,
 	rerankModel *models.RerankModel,
 	chunks []map[string]interface{},
 	total int,
@@ -67,7 +69,7 @@ func Rerank(
 ) (sim []float64, tsim []float64, vsim []float64) {
 	// If reranker model is provided and there are results, use model reranking
 	if rerankModel != nil && total > 0 {
-		return RerankByModel(rerankModel, chunks, nil, nil, query, tkWeight, vtWeight, cfield, qb, rankFeature)
+		return RerankByModel(ctx, rerankModel, chunks, nil, nil, query, tkWeight, vtWeight, cfield, qb, rankFeature)
 	}
 
 	// Otherwise, use fallback logic based on engine type
@@ -87,6 +89,7 @@ func Rerank(
 
 // RerankByModel performs reranking using a reranker model
 func RerankByModel(
+	ctx context.Context,
 	rerankModel *models.RerankModel,
 	chunks []map[string]interface{},
 	ids []string,
@@ -148,7 +151,7 @@ func RerankByModel(
 	tsim = TokenSimilarity(keywords, insTw, qb)
 
 	// Get similarity scores from reranker model
-	rerankResponse, err := rerankModel.ModelDriver.Rerank(rerankModel.ModelName, query, docs, rerankModel.APIConfig, &models.RerankConfig{}, nil)
+	rerankResponse, err := rerankModel.ModelDriver.Rerank(ctx, rerankModel.ModelName, query, docs, rerankModel.APIConfig, &models.RerankConfig{}, nil)
 	if err != nil {
 		common.Error("RerankByModel: rerankModel.Rerank failed; falling back to token-only similarity", err)
 		// If model fails, fall back to token similarity only
diff --git a/internal/service/nlp/retrieval.go b/internal/service/nlp/retrieval.go
index afb8b4f86c..6fdb9d1139 100644
--- a/internal/service/nlp/retrieval.go
+++ b/internal/service/nlp/retrieval.go
@@ -179,6 +179,7 @@ func (s *RetrievalService) Retrieval(ctx context.Context, req *RetrievalRequest)
 	if req.RerankModel != nil && searchResult.Total > 0 {
 		// External rerank model path - use RerankByModel
 		sim, term_similarity, vector_similarity = RerankByModel(
+			ctx,
 			req.RerankModel,
 			searchResult.Chunks,
 			searchResult.IDs,
@@ -652,7 +653,7 @@ func (s *RetrievalService) Search(ctx context.Context, req *RetrievalSearchReque
 			if similarityForGetVector <= 0 {
 				similarityForGetVector = 0.1
 			}
-			matchDense, err := s.GetVector(req.Question, req.EmbeddingModel, topk, similarityForGetVector)
+			matchDense, err := s.GetVector(ctx, req.Question, req.EmbeddingModel, topk, similarityForGetVector)
 			if err != nil {
 				return nil, fmt.Errorf("GetVector failed: %w", err)
 			}
@@ -781,11 +782,11 @@ func (s *RetrievalService) Search(ctx context.Context, req *RetrievalSearchReque
 }
 
 // GetVector computes query vector and returns MatchDenseExpr for hybrid search
-func (s *RetrievalService) GetVector(txt string, embModel *models.EmbeddingModel, topk int, similarity float64) (*types.MatchDenseExpr, error) {
+func (s *RetrievalService) GetVector(ctx context.Context, txt string, embModel *models.EmbeddingModel, topk int, similarity float64) (*types.MatchDenseExpr, error) {
 	embeddingConfig := &models.EmbeddingConfig{
 		Dimension: 0,
 	}
-	embeddings, err := embModel.ModelDriver.Embed(embModel.ModelName, []string{txt}, embModel.APIConfig, embeddingConfig, nil)
+	embeddings, err := embModel.ModelDriver.Embed(ctx, embModel.ModelName, []string{txt}, embModel.APIConfig, embeddingConfig, nil)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/service/related_question.go b/internal/service/related_question.go
index ddb99749cd..a582eb810e 100644
--- a/internal/service/related_question.go
+++ b/internal/service/related_question.go
@@ -17,6 +17,7 @@
 package service
 
 import (
+	"context"
 	"encoding/json"
 	"fmt"
 	"regexp"
@@ -27,7 +28,7 @@ import (
 )
 
 // GenerateRelatedQuestions generates related search questions for chat/searchbot endpoints.
-func GenerateRelatedQuestions(tenantID, question, searchID string, searchSvc *SearchService, tenantSvc *TenantService, modelProviderSvc *ModelProviderService) ([]string, error) {
+func GenerateRelatedQuestions(ctx context.Context, tenantID, question, searchID string, searchSvc *SearchService, tenantSvc *TenantService, modelProviderSvc *ModelProviderService) ([]string, error) {
 	if modelProviderSvc == nil {
 		return nil, fmt.Errorf("model provider service not configured")
 	}
@@ -41,7 +42,7 @@ func GenerateRelatedQuestions(tenantID, question, searchID string, searchSvc *Se
 		{Role: "system", Content: prompt},
 		{Role: "user", Content: "\nKeywords: " + question + "\nRelated search terms:\n    "},
 	}
-	response, err := modelProviderSvc.Chat(tenantID, modelID, messages, relatedQuestionsConfig(searchConfig))
+	response, err := modelProviderSvc.Chat(ctx, tenantID, modelID, messages, relatedQuestionsConfig(searchConfig))
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/service/skill_indexer.go b/internal/service/skill_indexer.go
index fa63daee0a..cf51c17687 100644
--- a/internal/service/skill_indexer.go
+++ b/internal/service/skill_indexer.go
@@ -935,7 +935,7 @@ func (s *SkillIndexerService) generateEmbedding(ctx context.Context, text, embdI
 	truncatedText := truncate(text, maxLen-10)
 
 	var response []models.EmbeddingData
-	response, err = embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{truncatedText}, embeddingModel.APIConfig, nil, nil)
+	response, err = embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{truncatedText}, embeddingModel.APIConfig, nil, nil)
 	if err != nil {
 		return nil, fmt.Errorf("failed to encode text: %w", err)
 	}
@@ -979,7 +979,7 @@ func (s *SkillIndexerService) generateEmbeddings(ctx context.Context, texts []st
 	common.Info(fmt.Sprintf("Encoding %d texts", len(truncatedTexts)))
 	// Use batch encode API (consistent with Python's encode(texts: list))
 	var response []models.EmbeddingData
-	response, err = embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, truncatedTexts, embeddingModel.APIConfig, nil, nil)
+	response, err = embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, truncatedTexts, embeddingModel.APIConfig, nil, nil)
 	if err != nil {
 		common.Error(fmt.Sprintf("Failed to encode texts: %v", err), err)
 		return nil, fmt.Errorf("failed to encode texts: %w", err)
@@ -1026,7 +1026,7 @@ func (s *SkillIndexerService) getEmbeddingDimension(ctx context.Context, tenantI
 	// Use simple test text like Python does: embedding_model.encode(["ok"])
 	testText := "ok"
 	var response []models.EmbeddingData
-	response, err = embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{testText}, embeddingModel.APIConfig, nil, nil)
+	response, err = embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{testText}, embeddingModel.APIConfig, nil, nil)
 	if err != nil {
 		return 0, fmt.Errorf("failed to encode test text: %w", err)
 	}
diff --git a/internal/service/skill_search.go b/internal/service/skill_search.go
index 33d24019a4..5a946219aa 100644
--- a/internal/service/skill_search.go
+++ b/internal/service/skill_search.go
@@ -680,7 +680,7 @@ func (s *SkillSearchService) getEmbedding(ctx context.Context, text, embdID, ten
 	truncatedText := truncate(text, maxLen-10)
 
 	var response []models.EmbeddingData
-	response, err = embeddingModel.ModelDriver.Embed(embeddingModel.ModelName, []string{truncatedText}, embeddingModel.APIConfig, nil, nil)
+	response, err = embeddingModel.ModelDriver.Embed(ctx, embeddingModel.ModelName, []string{truncatedText}, embeddingModel.APIConfig, nil, nil)
 	if err != nil {
 		return nil, fmt.Errorf("failed to encode query: %w", err)
 	}
diff --git a/internal/service/toc_enhancer.go b/internal/service/toc_enhancer.go
index 1533cd7f0a..3f41496f3b 100644
--- a/internal/service/toc_enhancer.go
+++ b/internal/service/toc_enhancer.go
@@ -479,7 +479,7 @@ func (e *TOCEnhancer) scoreEntries(ctx context.Context, entries []tocEntry, limi
 			modelName = *e.chatModel.ModelName
 		}
 		resp, err := e.chatModel.ModelDriver.ChatWithMessages(
-			modelName, msgs, e.chatModel.APIConfig, cfg, nil,
+			ctx, modelName, msgs, e.chatModel.APIConfig, cfg, nil,
 		)
 		if err != nil {
 			return nil, err
diff --git a/internal/service/user.go b/internal/service/user.go
index 7c29c20dec..649d567af0 100644
--- a/internal/service/user.go
+++ b/internal/service/user.go
@@ -389,7 +389,7 @@ func (s *UserService) Login(req *LoginRequest) (*entity.User, common.ErrorCode,
 // - CodeAuthenticationError (109): Email not registered or password mismatch
 // - CodeServerError (500): Password decryption failure
 // - CodeForbidden (403): Account disabled
-func (s *UserService) LoginByEmail(req *EmailLoginRequest) (*entity.User, common.ErrorCode, error) {
+func (s *UserService) LoginByEmail(ctx context.Context, req *EmailLoginRequest) (*entity.User, common.ErrorCode, error) {
 	user, err := s.userDAO.GetByEmail(req.Email)
 	if err != nil {
 		return nil, common.CodeAuthenticationError, fmt.Errorf("email: %s is not registered!", req.Email)
@@ -414,7 +414,7 @@ func (s *UserService) LoginByEmail(req *EmailLoginRequest) (*entity.User, common
 	now := time.Now().Truncate(time.Second)
 	user.LastLoginTime = &now
 
-	if err := s.userDAO.Update(user); err != nil {
+	if err = s.userDAO.Update(user); err != nil {
 		return nil, common.CodeServerError, fmt.Errorf("failed to update user: %w", err)
 	}
 
@@ -422,8 +422,8 @@ func (s *UserService) LoginByEmail(req *EmailLoginRequest) (*entity.User, common
 }
 
 // GetUserByID get user by ID
-func (s *UserService) GetUserByID(id uint) (*UserResponse, common.ErrorCode, error) {
-	user, err := s.userDAO.GetByID(id)
+func (s *UserService) GetUserByID(ctx context.Context, id uint) (*UserResponse, common.ErrorCode, error) {
+	user, err := s.userDAO.GetByID(ctx, id)
 	if err != nil {
 		return nil, common.CodeNotFound, err
 	}
@@ -586,7 +586,7 @@ func defaultUserLanguage() string {
 // GetUserByToken gets user by authorization header
 // The token parameter is the authorization header value, which needs to be decrypted
 // using itsdangerous URLSafeTimedSerializer to get the actual access_token
-func (s *UserService) GetUserByToken(authorization string) (*entity.User, common.ErrorCode, error) {
+func (s *UserService) GetUserByToken(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error) {
 	// Get secret key from config
 	secretKey, err := server.GetSecretKey(redis.Get())
 	if err != nil {
@@ -632,7 +632,7 @@ func (s *UserService) Logout(user *entity.User) (common.ErrorCode, error) {
 }
 
 // GetUserProfile returns user profile information
-func (s *UserService) GetUserProfile(user *entity.User) map[string]interface{} {
+func (s *UserService) GetUserProfile(ctx context.Context, user *entity.User) map[string]interface{} {
 	// Format create time and date (from database fields)
 	createTime := user.CreateTime
 	createDate := ""
@@ -737,7 +737,7 @@ func (s *UserService) GetUserProfile(user *entity.User) map[string]interface{} {
 }
 
 // UpdateUserSettings updates user settings
-func (s *UserService) UpdateUserSettings(user *entity.User, req *UpdateSettingsRequest) (common.ErrorCode, error) {
+func (s *UserService) UpdateUserSettings(ctx context.Context, user *entity.User, req *UpdateSettingsRequest) (common.ErrorCode, error) {
 	// Update fields if provided
 	if req.Password != nil {
 		ciphertext, err := base64.StdEncoding.DecodeString(*req.Password)
@@ -754,20 +754,22 @@ func (s *UserService) UpdateUserSettings(user *entity.User, req *UpdateSettingsR
 			oldPassword = string(oldPasswordBytes)
 		}
 		if user.Password == nil || !s.VerifyPassword(*user.Password, oldPassword) {
-			return common.CodeAuthenticationError, fmt.Errorf("Password error!")
+			return common.CodeAuthenticationError, fmt.Errorf("password error")
 		}
 
 		if req.NewPassword != nil {
-			ciphertext, err := base64.StdEncoding.DecodeString(*req.NewPassword)
+			ciphertext, err = base64.StdEncoding.DecodeString(*req.NewPassword)
 			if err != nil {
 				return common.CodeExceptionError, fmt.Errorf("Error('Incorrect padding')")
 			}
-			newPasswordBytes, err := rsa.DecryptPKCS1v15(nil, privateKey, ciphertext)
+			var newPasswordBytes []byte
+			newPasswordBytes, err = rsa.DecryptPKCS1v15(nil, privateKey, ciphertext)
 			if err != nil {
 				return common.CodeExceptionError, err
 			}
 
-			hashedPassword, err := common.GenerateWerkzeugPasswordHash(string(newPasswordBytes))
+			var hashedPassword string
+			hashedPassword, err = common.GenerateWerkzeugPasswordHash(string(newPasswordBytes))
 			if err != nil {
 				return common.CodeExceptionError, err
 			}
@@ -803,7 +805,7 @@ func (s *UserService) UpdateUserSettings(user *entity.User, req *UpdateSettingsR
 }
 
 // ChangePassword changes user password
-func (s *UserService) ChangePassword(user *entity.User, req *ChangePasswordRequest) (common.ErrorCode, error) {
+func (s *UserService) ChangePassword(ctx context.Context, user *entity.User, req *ChangePasswordRequest) (common.ErrorCode, error) {
 	// If password is provided, verify current password
 	if req.Password != nil {
 		if user.Password == nil || !s.VerifyPassword(*user.Password, *req.Password) {
@@ -873,7 +875,7 @@ type SetTenantInfoRequest struct {
 }
 
 // SetTenantInfo updates tenant model configuration
-func (s *UserService) SetTenantInfo(userID string, req *SetTenantInfoRequest) (common.ErrorCode, error) {
+func (s *UserService) SetTenantInfo(ctx context.Context, userID string, req *SetTenantInfoRequest) (common.ErrorCode, error) {
 	_ = userID
 	tenantDAO := dao.NewTenantDAO()
 	updates := make(map[string]interface{})
@@ -999,7 +1001,7 @@ func convertToUserTenantRelation(userTenant *entity.UserTenant) *UserTenantRelat
 // GetUserByAPIToken gets user by access key from Authorization header
 // This is used for API token authentication
 // The authorization parameter should be in format: "Bearer <token>" or just "<token>"
-func (s *UserService) GetUserByAPIToken(authorization string) (*entity.User, common.ErrorCode, error) {
+func (s *UserService) GetUserByAPIToken(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error) {
 	if authorization == "" {
 		return nil, common.CodeUnauthorized, fmt.Errorf("authorization header is empty")
 	}
@@ -1043,7 +1045,7 @@ func (s *UserService) GetUserByAPIToken(authorization string) (*entity.User, com
 // to expose DialogID (the real agent_id) to downstream handlers
 // without re-parsing the Authorization header. Mirrors
 // `APIToken.query(beta=token)` from python bot_api.py:agent_bot_logs.
-func (s *UserService) GetAPITokenByBeta(authorization string) (*entity.APIToken, error) {
+func (s *UserService) GetAPITokenByBeta(ctx context.Context, authorization string) (*entity.APIToken, error) {
 	authorization = strings.TrimSpace(authorization)
 	if authorization == "" {
 		return nil, fmt.Errorf("authorization header is empty")
@@ -1076,7 +1078,7 @@ func (s *UserService) GetAPITokenByBeta(authorization string) (*entity.APIToken,
 
 // GetUserByBetaAPIToken gets user by beta access key from Authorization
 // header. This mirrors Python's AUTH_BETA flow used by public bot endpoints.
-func (s *UserService) GetUserByBetaAPIToken(authorization string) (*entity.User, common.ErrorCode, error) {
+func (s *UserService) GetUserByBetaAPIToken(ctx context.Context, authorization string) (*entity.User, common.ErrorCode, error) {
 	authorization = strings.TrimSpace(authorization)
 	if authorization == "" {
 		return nil, common.CodeUnauthorized, fmt.Errorf("authorization header is empty")
diff --git a/internal/service/user_beta_test.go b/internal/service/user_beta_test.go
index 470d2c783c..adc6dd2cf9 100644
--- a/internal/service/user_beta_test.go
+++ b/internal/service/user_beta_test.go
@@ -34,6 +34,7 @@ func setupUserBetaServiceDB(t *testing.T) {
 }
 
 func TestUserServiceGetUserByBetaAPIToken(t *testing.T) {
+	ctx := t.Context()
 	setupUserBetaServiceDB(t)
 
 	accessToken := "access-token"
@@ -62,7 +63,7 @@ func TestUserServiceGetUserByBetaAPIToken(t *testing.T) {
 
 	svc := NewUserService()
 	for _, auth := range []string{"Bearer " + beta, beta, "Bearer    " + beta} {
-		user, code, err := svc.GetUserByBetaAPIToken(auth)
+		user, code, err := svc.GetUserByBetaAPIToken(ctx, auth)
 		if err != nil {
 			t.Fatalf("GetUserByBetaAPIToken(%q) failed: %v", auth, err)
 		}
@@ -76,9 +77,10 @@ func TestUserServiceGetUserByBetaAPIToken(t *testing.T) {
 }
 
 func TestUserServiceGetUserByBetaAPITokenRejectsInvalidToken(t *testing.T) {
+	ctx := t.Context()
 	setupUserBetaServiceDB(t)
 
-	_, code, err := NewUserService().GetUserByBetaAPIToken("Bearer missing")
+	_, code, err := NewUserService().GetUserByBetaAPIToken(ctx, "Bearer missing")
 	if err == nil {
 		t.Fatal("expected error for invalid beta token")
 	}
@@ -88,10 +90,11 @@ func TestUserServiceGetUserByBetaAPITokenRejectsInvalidToken(t *testing.T) {
 }
 
 func TestUserServiceGetUserByBetaAPITokenRejectsEmptyOrWhitespaceToken(t *testing.T) {
+	ctx := t.Context()
 	setupUserBetaServiceDB(t)
 
 	for _, auth := range []string{"Bearer ", "   ", "\t"} {
-		_, code, err := NewUserService().GetUserByBetaAPIToken(auth)
+		_, code, err := NewUserService().GetUserByBetaAPIToken(ctx, auth)
 		if err == nil {
 			t.Fatalf("expected error for auth %q", auth)
 		}
@@ -102,6 +105,7 @@ func TestUserServiceGetUserByBetaAPITokenRejectsEmptyOrWhitespaceToken(t *testin
 }
 
 func TestUserServiceGetUserByBetaAPITokenRejectsUserWithEmptyAccessToken(t *testing.T) {
+	ctx := t.Context()
 	setupUserBetaServiceDB(t)
 
 	status := "1"
@@ -128,7 +132,7 @@ func TestUserServiceGetUserByBetaAPITokenRejectsUserWithEmptyAccessToken(t *test
 		t.Fatalf("failed to create api token: %v", err)
 	}
 
-	_, code, err := NewUserService().GetUserByBetaAPIToken("Bearer " + beta)
+	_, code, err := NewUserService().GetUserByBetaAPIToken(ctx, "Bearer "+beta)
 	if err == nil {
 		t.Fatal("expected error for empty access token")
 	}