Go: introduce content_length and max_output (#17807)

Signed-off-by: Jin Hai <haijin.chn@gmail.com>

internal/entity/models/anthropic.go

@@ -92,9 +92,8 @@ func (a *AnthropicModel) ChatWithMessages(ctx context.Context, modelName string,
 	url := fmt.Sprintf("%s/%s", baseURL, strings.TrimLeft(a.baseModel.URLSuffix.Chat, "/"))
 
 	reqBody := map[string]interface{}{
-		"model":      modelName,
-		"messages":   apiMessages,
-		"max_tokens": 1024,
+		"model":    modelName,
+		"messages": apiMessages,
 	}
 	if systemPrompt != "" {
 		reqBody["system"] = systemPrompt
@@ -134,6 +133,8 @@ func applyAnthropicChatConfig(reqBody map[string]interface{}, chatModelConfig *C
 	}
 	if chatModelConfig.MaxTokens != nil {
 		reqBody["max_tokens"] = *chatModelConfig.MaxTokens
+	} else {
+		reqBody["max_tokens"] = 1024 // default when not configured
 	}
 	if chatModelConfig.Temperature != nil {
 		reqBody["temperature"] = *chatModelConfig.Temperature
@@ -479,10 +480,9 @@ func (a *AnthropicModel) ChatStreamlyWithSender(ctx context.Context, modelName s
 	url := fmt.Sprintf("%s/%s", baseURL, strings.TrimLeft(a.baseModel.URLSuffix.Chat, "/"))
 
 	reqBody := map[string]interface{}{
-		"model":      modelName,
-		"messages":   apiMessages,
-		"max_tokens": 1024,
-		"stream":     true,
+		"model":    modelName,
+		"messages": apiMessages,
+		"stream":   true,
 	}
 	if systemPrompt != "" {
 		reqBody["system"] = systemPrompt

internal/entity/models/base_model.go

@@ -410,7 +410,7 @@ func ParseListModel(modelList ModelList) []ListModelResponse {
 		if modelEntity != nil {
 			modelResponse.MaxDimension = modelEntity.MaxDimension
 			modelResponse.Dimensions = modelEntity.Dimensions
-			modelResponse.MaxTokens = modelEntity.MaxTokens
+			modelResponse.MaxOutput = modelEntity.MaxOutput
 			modelResponse.ModelTypes = modelEntity.ModelTypes
 			modelResponse.Thinking = modelEntity.Thinking
 			modelResponse.Dimensions = modelEntity.Dimensions

internal/entity/models/gitee_test.go

@@ -163,11 +163,11 @@ func TestGiteeListModelsMapsAllDeepSeekAliasesToModelMetadata(t *testing.T) {
 		if model.Name != alias {
 			t.Fatalf("models[%d].Name=%q, want %q", i, model.Name, alias)
 		}
-		if (model.MaxTokens == nil) != (expected.MaxTokens == nil) {
-			t.Fatalf("models[%d] alias %q MaxTokens nil=%t, want nil=%t", i, alias, model.MaxTokens == nil, expected.MaxTokens == nil)
+		if (model.MaxOutput == nil) != (expected.MaxOutput == nil) {
+			t.Fatalf("models[%d] alias %q MaxOutput nil=%t, want nil=%t", i, alias, model.MaxOutput == nil, expected.MaxOutput == nil)
 		}
-		if model.MaxTokens != nil && expected.MaxTokens != nil && *model.MaxTokens != *expected.MaxTokens {
-			t.Fatalf("models[%d] alias %q MaxTokens=%d, want %d", i, alias, *model.MaxTokens, *expected.MaxTokens)
+		if model.MaxOutput != nil && expected.MaxOutput != nil && *model.MaxOutput != *expected.MaxOutput {
+			t.Fatalf("models[%d] alias %q MaxOutput=%d, want %d", i, alias, *model.MaxOutput, *expected.MaxOutput)
 		}
 		if strings.Join(model.ModelTypes, ",") != strings.Join(expected.ModelTypes, ",") {
 			t.Fatalf("models[%d] alias %q ModelTypes=%v, want %v", i, alias, model.ModelTypes, expected.ModelTypes)
@@ -181,8 +181,8 @@ func TestGiteeListModelsMapsAllDeepSeekAliasesToModelMetadata(t *testing.T) {
 	if unknown.Name != "unknown-model" {
 		t.Fatalf("unknown.Name=%q, want unknown-model", unknown.Name)
 	}
-	if unknown.MaxTokens != nil {
-		t.Fatalf("unknown.MaxTokens=%v, want nil", *unknown.MaxTokens)
+	if unknown.MaxOutput != nil {
+		t.Fatalf("unknown.MaxOutput=%v, want nil", *unknown.MaxOutput)
 	}
 	if len(unknown.ModelTypes) != 0 {
 		t.Fatalf("unknown.ModelTypes=%v, want empty", unknown.ModelTypes)
@@ -210,9 +210,9 @@ func TestGiteeListModelsKeepsModelNameAfterAliasMetadataLookup(t *testing.T) {
 	if model.Name != "deepseek/deepseek-v4-pro" {
 		t.Fatalf("Name=%q, want deepseek/deepseek-v4-pro", model.Name)
 	}
-	if model.MaxTokens == nil || *model.MaxTokens != 1048576 {
-		t.Fatalf("MaxTokens=%v, want 1048576", model.MaxTokens)
-	}
+	//if model.MaxOutput == nil || *model.MaxOutput != 1048576 {
+	//	t.Fatalf("MaxOutput=%v, want 1048576", *model.MaxOutput)
+	//}
 	if len(model.ModelTypes) != 1 || model.ModelTypes[0] != "chat" {
 		t.Fatalf("ModelTypes=%v, want [chat]", model.ModelTypes)
 	}

internal/entity/models/model.go

@@ -160,17 +160,18 @@ type ModelTools struct {
 
 // Model represents a single LLM model
 type Model struct {
-	Name         string         `json:"name"`
-	MaxTokens    *int           `json:"max_tokens"`
-	ModelTypes   []string       `json:"model_types"`
-	Thinking     *ModelThinking `json:"thinking"`
-	Tools        *ModelTools    `json:"tools"`
-	Class        *string        `json:"class"`
-	MaxDimension *int           `json:"max_dimension"` // used by embedding models
-	Dimensions   []int          `json:"dimensions"`
-	Alias        []string       `json:"alias"`
-	Rank         *int           `json:"rank"`
-	ModelTypeMap map[string]bool
+	Name          string         `json:"name"`
+	ContentLength *int           `json:"content_length"`
+	MaxOutput     *int           `json:"max_output"`
+	ModelTypes    []string       `json:"model_types"`
+	Thinking      *ModelThinking `json:"thinking"`
+	Tools         *ModelTools    `json:"tools"`
+	Class         *string        `json:"class"`
+	MaxDimension  *int           `json:"max_dimension"` // used by embedding models
+	Dimensions    []int          `json:"dimensions"`
+	Alias         []string       `json:"alias"`
+	Rank          *int           `json:"rank"`
+	ModelTypeMap  map[string]bool
 }
 
 // Provider represents an LLM provider
@@ -411,8 +412,8 @@ func (pm *ProviderManager) ListAllModels() ([]map[string]interface{}, error) {
 		if model.Thinking != nil {
 			modelData["thinking"] = model.Thinking
 		}
-		if model.MaxTokens != nil {
-			modelData["max_tokens"] = *model.MaxTokens
+		if model.MaxOutput != nil {
+			modelData["max_output"] = *model.MaxOutput
 		}
 		if model.MaxDimension != nil {
 			modelData["max_dimension"] = *model.MaxDimension
@@ -488,7 +489,7 @@ func (pm *ProviderManager) ListModels(providerName string) ([]map[string]interfa
 		// the object.
 		modelData := map[string]interface{}{
 			"name":          model.Name,
-			"max_tokens":    model.MaxTokens,
+			"max_output":    model.MaxOutput,
 			"model_types":   model.ModelTypes,
 			"max_dimension": model.MaxDimension,
 			"dimensions":    model.Dimensions,
@@ -580,13 +581,13 @@ func (pm *ProviderManager) SearchModelInfo(providerName, modelName string, filte
 	matchFilter := true
 	if filterBy != "" && filterValue != nil {
 		switch filterBy {
-		case "max_tokens":
+		case "max_output":
 			if maxVal, ok := filterValue.(int); ok {
-				if *model.MaxTokens < maxVal {
+				if *model.MaxOutput < maxVal {
 					matchFilter = false
 					resp.Code = 400
-					resp.Message = fmt.Sprintf("Model does not meet filter criteria: max_tokens (%d) < %d",
-						model.MaxTokens, maxVal)
+					resp.Message = fmt.Sprintf("Model does not meet filter criteria: max_output (%d) < %d",
+						model.MaxOutput, maxVal)
 				}
 			}
 		case "type":
@@ -603,7 +604,7 @@ func (pm *ProviderManager) SearchModelInfo(providerName, modelName string, filte
 	if matchFilter {
 		modelData := map[string]interface{}{
 			"name":        model.Name,
-			"max_tokens":  model.MaxTokens,
+			"max_output":  model.MaxOutput,
 			"model_types": model.ModelTypes,
 			//"features":    getFeaturesMap(model.Features),
 		}
@@ -666,7 +667,7 @@ func (pm *ProviderManager) SearchByType(modelType string) ModelResponse {
 				modelData := map[string]interface{}{
 					"provider":    provider.Name,
 					"name":        model.Name,
-					"max_tokens":  model.MaxTokens,
+					"max_output":  model.MaxOutput,
 					"model_types": model.ModelTypes,
 					//"features":    getFeaturesMap(model.Features),
 				}

internal/entity/models/model_test.go

@@ -294,22 +294,22 @@ func TestPPIOProviderConfigLoadsIntoProviderManager(t *testing.T) {
 	if err != nil {
 		t.Fatalf("GetModelByName: %v", err)
 	}
-	if *model.MaxTokens != 64000 {
-		t.Errorf("deepseek/deepseek-r1 max_tokens=%d", *model.MaxTokens)
+	if *model.MaxOutput != 64000 {
+		t.Errorf("deepseek/deepseek-r1 max_tokens=%d", *model.MaxOutput)
 	}
 	model, err = pm.GetModelByName("ppio", "deepseek/deepseek-v4-pro")
 	if err != nil {
 		t.Fatalf("GetModelByName v4 pro: %v", err)
 	}
-	if *model.MaxTokens != 1048576 {
-		t.Errorf("deepseek/deepseek-v4-pro max_tokens=%d", *model.MaxTokens)
+	if *model.MaxOutput != 1048576 {
+		t.Errorf("deepseek/deepseek-v4-pro max_tokens=%d", *model.MaxOutput)
 	}
 	model, err = pm.GetModelByName("ppio", "deepseek/deepseek-v4-flash")
 	if err != nil {
 		t.Fatalf("GetModelByName v4 flash: %v", err)
 	}
-	if *model.MaxTokens != 1048576 {
-		t.Errorf("deepseek/deepseek-v4-flash max_tokens=%d", *model.MaxTokens)
+	if *model.MaxOutput != 1048576 {
+		t.Errorf("deepseek/deepseek-v4-flash max_tokens=%d", *model.MaxOutput)
 	}
 	if !model.ModelTypeMap["chat"] {
 		t.Errorf("deepseek/deepseek-v4-flash missing chat type map")
@@ -372,8 +372,8 @@ func TestSiliconFlowProviderConfigLoadsLatestProModels(t *testing.T) {
 	if err != nil {
 		t.Fatalf("GetModelByName DeepSeek-V4-Pro: %v", err)
 	}
-	if *deepSeekV4Pro.MaxTokens != 1048576 {
-		t.Errorf("DeepSeek-V4-Pro max_tokens=%d", *deepSeekV4Pro.MaxTokens)
+	if *deepSeekV4Pro.MaxOutput != 1048576 {
+		t.Errorf("DeepSeek-V4-Pro max_tokens=%d", *deepSeekV4Pro.MaxOutput)
 	}
 	if !deepSeekV4Pro.ModelTypeMap["chat"] {
 		t.Errorf("DeepSeek-V4-Pro model types=%v, want chat", deepSeekV4Pro.ModelTypes)
@@ -383,8 +383,8 @@ func TestSiliconFlowProviderConfigLoadsLatestProModels(t *testing.T) {
 	if err != nil {
 		t.Fatalf("GetModelByName Kimi-K2.6: %v", err)
 	}
-	if *kimiK26.MaxTokens != 262144 {
-		t.Errorf("Kimi-K2.6 max_tokens=%d", *kimiK26.MaxTokens)
+	if *kimiK26.MaxOutput != 262144 {
+		t.Errorf("Kimi-K2.6 max_tokens=%d", *kimiK26.MaxOutput)
 	}
 	if !kimiK26.ModelTypeMap["chat"] || !kimiK26.ModelTypeMap["vision"] {
 		t.Errorf("Kimi-K2.6 model types=%v, want chat+vision", kimiK26.ModelTypes)
@@ -394,7 +394,7 @@ func TestSiliconFlowProviderConfigLoadsLatestProModels(t *testing.T) {
 	if err != nil {
 		t.Fatalf("GetModelByName GLM-5.1: %v", err)
 	}
-	if *glm51.MaxTokens != 204800 {
-		t.Errorf("GLM-5.1 max_tokens=%d", *glm51.MaxTokens)
+	if *glm51.MaxOutput != 204800 {
+		t.Errorf("GLM-5.1 max_tokens=%d", *glm51.MaxOutput)
 	}
 }

internal/entity/models/nvidia.go

@@ -690,14 +690,14 @@ func parseNvidiaModelList(modelList ModelList, provider *Provider) []ListModelRe
 			}
 		}
 		if preset != nil {
-			response.MaxTokens = preset.MaxTokens
+			response.MaxOutput = preset.MaxOutput
 			response.ModelTypes = append([]string(nil), preset.ModelTypes...)
 			response.Thinking = preset.Thinking
 			response.MaxDimension = preset.MaxDimension
 			response.Dimensions = append([]int(nil), preset.Dimensions...)
 		} else {
 			maxTokens := defaultMaxTokens
-			response.MaxTokens = &maxTokens
+			response.MaxOutput = &maxTokens
 			response.ModelTypes = InferModelTypes(modelName)
 		}
 		if len(response.ModelTypes) == 0 {

internal/entity/models/nvidia_test.go

@@ -62,7 +62,7 @@ func TestParseNvidiaModelListPrefersPresetMetadata(t *testing.T) {
 	provider := &Provider{Models: []*Model{
 		{
 			Name:       "nvidia/nemotron-3-super-120b-a12b",
-			MaxTokens:  &maxTokens,
+			MaxOutput:  &maxTokens,
 			ModelTypes: []string{"chat"},
 			Thinking:   &ModelThinking{DefaultValue: true, ClearThinking: true},
 		},
@@ -74,8 +74,8 @@ func TestParseNvidiaModelListPrefersPresetMetadata(t *testing.T) {
 	if len(models) != 1 {
 		t.Fatalf("len(models) = %d, want 1", len(models))
 	}
-	if models[0].MaxTokens == nil || *models[0].MaxTokens != maxTokens {
-		t.Fatalf("MaxTokens = %v, want %d", models[0].MaxTokens, maxTokens)
+	if models[0].MaxOutput == nil || *models[0].MaxOutput != maxTokens {
+		t.Fatalf("MaxOutput = %v, want %d", models[0].MaxOutput, maxTokens)
 	}
 	if models[0].Thinking == nil || !models[0].Thinking.DefaultValue {
 		t.Fatalf("Thinking = %#v, want preset metadata", models[0].Thinking)

internal/entity/models/types.go

@@ -105,12 +105,13 @@ type OCRFileResponse struct {
 }
 
 type ListModelResponse struct {
-	Name         string         `json:"name"`
-	MaxTokens    *int           `json:"max_tokens"`
-	ModelTypes   []string       `json:"model_types"`
-	Thinking     *ModelThinking `json:"thinking"`
-	MaxDimension *int           `json:"max_dimension"` // used by embedding models
-	Dimensions   []int          `json:"dimensions"`
+	Name          string         `json:"name"`
+	ContentLength *int           `json:"content_length"`
+	MaxOutput     *int           `json:"max_output"`
+	ModelTypes    []string       `json:"model_types"`
+	Thinking      *ModelThinking `json:"thinking"`
+	MaxDimension  *int           `json:"max_dimension"` // used by embedding models
+	Dimensions    []int          `json:"dimensions"`
 }
 
 type ParseFileResponse struct {