diff --git a/cmd/ragflow_server.go b/cmd/ragflow_server.go
index 6aa8f9ac74..101259364e 100644
--- a/cmd/ragflow_server.go
+++ b/cmd/ragflow_server.go
@@ -38,7 +38,6 @@ import (
 	"ragflow/internal/storage"
 	"ragflow/internal/syncer"
 	"ragflow/internal/tokenizer"
-	"runtime"
 	"strconv"
 	"strings"
 	"syscall"
@@ -496,7 +495,7 @@ func runIngestor(args *serverArgs) error {
 	}
 	defer tokenizer.Close()
 
-	ingestor := ingestion.NewIngestor(*args.name, int32(runtime.NumCPU()), []string{"pdf", "docx", "txt"})
+	ingestor := ingestion.NewIngestor(*args.name, 2, []string{"pdf", "docx", "txt"})
 
 	go func() {
 		err := ingestor.Start()
diff --git a/internal/admin/handler.go b/internal/admin/handler.go
index ae5d4aee16..835d19ca19 100644
--- a/internal/admin/handler.go
+++ b/internal/admin/handler.go
@@ -25,7 +25,7 @@ import (
 	"ragflow/internal/common"
 	"ragflow/internal/engine"
 	"ragflow/internal/engine/redis"
-	"ragflow/internal/httputil"
+	"ragflow/internal/handler"
 	"ragflow/internal/server"
 	"ragflow/internal/service"
 	"ragflow/internal/storage"
@@ -1039,7 +1039,7 @@ func (h *Handler) RemoveIngestionTasks(c *gin.Context) {
 	if req.Email == nil && req.Status == nil {
 		tasks, err := h.service.RemoveIngestionTasks(req.Tasks)
 		if err != nil {
-			common.ErrorWithCode(c, httputil.IngestionTaskErrorCode(err), err.Error())
+			common.ErrorWithCode(c, handler.IngestionTaskErrorCode(err), err.Error())
 			return
 		}
 
@@ -1047,7 +1047,7 @@ func (h *Handler) RemoveIngestionTasks(c *gin.Context) {
 	} else {
 		tasks, err := h.service.RemoveIngestionTasksByCondition(req.Tasks, req.Email, req.Status)
 		if err != nil {
-			common.ErrorWithCode(c, httputil.IngestionTaskErrorCode(err), err.Error())
+			common.ErrorWithCode(c, handler.IngestionTaskErrorCode(err), err.Error())
 			return
 		}
 		common.SuccessWithData(c, tasks, "Remove tasks successfully")
@@ -1070,7 +1070,7 @@ func (h *Handler) StopIngestionTasks(c *gin.Context) {
 	if req.Email == nil && req.Status == nil {
 		tasks, err := h.service.StopIngestionTasks(req.Tasks)
 		if err != nil {
-			common.ErrorWithCode(c, httputil.IngestionTaskErrorCode(err), err.Error())
+			common.ErrorWithCode(c, handler.IngestionTaskErrorCode(err), err.Error())
 			return
 		}
 		var result []map[string]string
diff --git a/internal/handler/document.go b/internal/handler/document.go
index 4f2dc1d5a0..87fb9b31d8 100644
--- a/internal/handler/document.go
+++ b/internal/handler/document.go
@@ -29,7 +29,6 @@ import (
 	"path/filepath"
 	"ragflow/internal/common"
 	"ragflow/internal/entity"
-	"ragflow/internal/httputil"
 	"ragflow/internal/utility"
 	"strconv"
 	"strings"
@@ -45,7 +44,6 @@ var IMG_BASE64_PREFIX = "data:image/png;base64,"
 
 // documentServiceIface defines the DocumentService methods used by DocumentHandler.
 type documentServiceIface interface {
-	CreateDocument(req *service.CreateDocumentRequest) (*entity.Document, error)
 	GetDocumentByID(id string) (*service.DocumentResponse, error)
 	UpdateDocument(id string, req *service.UpdateDocumentRequest) error
 	DeleteDocument(id string) error
@@ -98,41 +96,6 @@ func NewDocumentHandler(documentService documentServiceIface, datasetService *se
 	}
 }
 
-// CreateDocument create document
-// @Summary Create Document
-// @Description Create new document
-// @Tags documents
-// @Accept json
-// @Produce json
-// @Param request body service.CreateDocumentRequest true "document info"
-// @Success 200 {object} map[string]interface{}
-// @Router /api/v1/documents [post]
-func (h *DocumentHandler) CreateDocument(c *gin.Context) {
-	_, errorCode, errorMessage := GetUser(c)
-	if errorCode != common.CodeSuccess {
-		common.ErrorWithCode(c, errorCode, errorMessage)
-		return
-	}
-
-	var req service.CreateDocumentRequest
-	if err := c.ShouldBindJSON(&req); err != nil {
-		c.JSON(http.StatusBadRequest, gin.H{
-			"error": err.Error(),
-		})
-		return
-	}
-
-	document, err := h.documentService.CreateDocument(&req)
-	if err != nil {
-		c.JSON(http.StatusInternalServerError, gin.H{
-			"error": err.Error(),
-		})
-		return
-	}
-
-	common.SuccessWithData(c, document, "created successfully")
-}
-
 // GetDocumentByID get document by ID
 // @Summary Get Document Info
 // @Description Get document details by ID
@@ -987,15 +950,15 @@ func (h *DocumentHandler) uploadWebDocument(c *gin.Context, kb *entity.Knowledge
 
 // mapDocKeysWithRunStatus renames a freshly-created document's raw keys to the
 // public response shape (chunk_num→chunk_count, token_num→token_count,
-// kb_id→dataset_id, parser_id→chunk_method) and reports run as a label.
+// kb_id→dataset_id) and reports run as a label.
 // Mirrors Python map_doc_keys_with_run_status / map_doc_keys.
 func mapDocKeysWithRunStatus(raw map[string]interface{}) map[string]interface{} {
 	out := map[string]interface{}{
-		"chunk_count":  raw["chunk_num"],
-		"token_count":  raw["token_num"],
-		"dataset_id":   raw["kb_id"],
-		"chunk_method": raw["parser_id"],
-		"run":          "UNSTART",
+		"chunk_count": raw["chunk_num"],
+		"token_count": raw["token_num"],
+		"dataset_id":  raw["kb_id"],
+		"parser_id":   raw["parser_id"],
+		"run":         "UNSTART",
 	}
 	for _, k := range []string{"id", "name", "type", "size", "suffix", "source_type", "created_by", "parser_config", "location", "pipeline_id", "content_hash"} {
 		if v, ok := raw[k]; ok {
@@ -1058,7 +1021,6 @@ func mapDocumentListItem(doc *entity.DocumentListItem, metaFields map[string]int
 		"suffix":           doc.Suffix,
 		"run":              mapRunStatus(doc.Run),
 		"status":           stringValue(doc.Status),
-		"chunk_method":     doc.ParserID,
 		"parser_id":        doc.ParserID,
 		"pipeline_id":      stringValue(doc.PipelineID),
 		"pipeline_name":    stringValue(doc.PipelineName),
@@ -1409,7 +1371,7 @@ func (h *DocumentHandler) ListIngestionTasks(c *gin.Context) {
 
 	parseResult, err = h.documentService.ListIngestionTasks(userID, req.DatasetID, 0, 0)
 	if err != nil {
-		common.ResponseWithCodeData(c, httputil.IngestionTaskErrorCode(err), nil, err.Error())
+		common.ResponseWithCodeData(c, IngestionTaskErrorCode(err), nil, err.Error())
 		return
 	}
 
@@ -1459,7 +1421,7 @@ func (h *DocumentHandler) StopIngestionTasks(c *gin.Context) {
 
 	parseResult, err := h.documentService.StopIngestionTasks(req.Tasks, userID)
 	if err != nil {
-		common.ResponseWithCodeData(c, httputil.IngestionTaskErrorCode(err), nil, err.Error())
+		common.ResponseWithCodeData(c, IngestionTaskErrorCode(err), nil, err.Error())
 		return
 	}
 	common.SuccessWithData(c, parseResult, "success")
@@ -1485,7 +1447,7 @@ func (h *DocumentHandler) RemoveIngestionTasks(c *gin.Context) {
 
 	deletedTasks, err := h.documentService.RemoveIngestionTasks(req.Tasks, userID)
 	if err != nil {
-		common.ResponseWithCodeData(c, httputil.IngestionTaskErrorCode(err), nil, err.Error())
+		common.ResponseWithCodeData(c, IngestionTaskErrorCode(err), nil, err.Error())
 		return
 	}
 	common.SuccessWithData(c, deletedTasks, "success")
diff --git a/internal/handler/document_test.go b/internal/handler/document_test.go
index a7e32c9474..da732a7f3a 100644
--- a/internal/handler/document_test.go
+++ b/internal/handler/document_test.go
@@ -138,9 +138,6 @@ func (f *fakeDocumentService) DownloadDocument(datasetID, docID string) (*servic
 		FileName:    "doc.pdf",
 	}, nil
 }
-func (f *fakeDocumentService) CreateDocument(req *service.CreateDocumentRequest) (*entity.Document, error) {
-	return nil, nil
-}
 func (f *fakeDocumentService) GetDocumentByID(id string) (*service.DocumentResponse, error) {
 	if f.docErr != nil {
 		return nil, f.docErr
diff --git a/internal/handler/error.go b/internal/handler/error.go
index a488234dec..e831237054 100644
--- a/internal/handler/error.go
+++ b/internal/handler/error.go
@@ -17,9 +17,11 @@
 package handler
 
 import (
+	"errors"
 	"net/http"
 
 	"ragflow/internal/common"
+	"ragflow/internal/service"
 
 	"github.com/gin-gonic/gin"
 	"go.uber.org/zap"
@@ -66,3 +68,20 @@ func HandleNoRoute(c *gin.Context) {
 		"error":   "Not Found",
 	})
 }
+
+// IngestionTaskErrorCode maps ingestion-task service errors to common.ErrorCode
+// for HTTP responses.
+func IngestionTaskErrorCode(err error) common.ErrorCode {
+	var transitionErr *service.InvalidTaskTransitionError
+	if errors.As(err, &transitionErr) {
+		return common.CodeConflict
+	}
+	var conflictErr *service.TaskStatusConflictError
+	if errors.As(err, &conflictErr) {
+		return common.CodeConflict
+	}
+	if errors.Is(err, common.ErrTaskNotFound) {
+		return common.CodeNotFound
+	}
+	return common.CodeExceptionError
+}
diff --git a/internal/httputil/ingestion_task_error_test.go b/internal/handler/error_test.go
similarity index 98%
rename from internal/httputil/ingestion_task_error_test.go
rename to internal/handler/error_test.go
index f87b35bca6..4a6b18d18b 100644
--- a/internal/httputil/ingestion_task_error_test.go
+++ b/internal/handler/error_test.go
@@ -1,4 +1,4 @@
-package httputil
+package handler
 
 import (
 	"testing"
diff --git a/internal/handler/pipeline.go b/internal/handler/pipeline.go
index 925d615e1c..61fed60bb9 100644
--- a/internal/handler/pipeline.go
+++ b/internal/handler/pipeline.go
@@ -24,11 +24,12 @@ import (
 	pipelinepkg "ragflow/internal/ingestion/pipeline"
 )
 
-// builtinPipelineLister is the registry surface PipelineHandler needs.
+// builtinPipelineRegistry is the registry surface PipelineHandler needs.
 // Defined as an interface so handler tests can inject a fake without
 // touching the embedded template FS.
-type builtinPipelineLister interface {
-	List() []*pipelinepkg.BuiltinPipelineMeta
+type builtinPipelineRegistry interface {
+	List() *pipelinepkg.BuiltinPipelineListResponse
+	Get(ref string) (*pipelinepkg.BuiltinPipeline, bool)
 }
 
 // PipelineHandler handles pipeline endpoints.
@@ -36,7 +37,7 @@ type builtinPipelineLister interface {
 // Built-in templates are shipped with the binary; user-defined templates
 // (canvas DSL) may be added later.
 type PipelineHandler struct {
-	registry builtinPipelineLister
+	registry builtinPipelineRegistry
 }
 
 // NewPipelineHandler builds a PipelineHandler backed by the embedded
@@ -59,7 +60,10 @@ func NewPipelineHandler() *PipelineHandler {
 // This is public static data, so no auth is required.
 func (h *PipelineHandler) ListPipelines(c *gin.Context) {
 	if h == nil || h.registry == nil {
-		common.SuccessWithData(c, []*pipelinepkg.BuiltinPipelineMeta{}, "success")
+		common.SuccessWithData(c, &pipelinepkg.BuiltinPipelineListResponse{
+			BuiltinPipelines: []*pipelinepkg.BuiltinPipelineMeta{},
+			Total:            0,
+		}, "success")
 		return
 	}
 
@@ -70,3 +74,22 @@ func (h *PipelineHandler) ListPipelines(c *gin.Context) {
 
 	common.SuccessWithData(c, h.registry.List(), "success")
 }
+
+// GetPipeline GET /api/v1/pipelines/:id
+// Returns a single built-in pipeline template with its DSL.
+// This is public static data, so no auth is required.
+func (h *PipelineHandler) GetPipeline(c *gin.Context) {
+	if h == nil || h.registry == nil {
+		common.ErrorWithCode(c, common.CodeDataError, "pipeline not found")
+		return
+	}
+
+	id := c.Param("id")
+	tpl, ok := h.registry.Get(id)
+	if !ok || tpl == nil {
+		common.ErrorWithCode(c, common.CodeDataError, "pipeline not found")
+		return
+	}
+
+	common.SuccessWithData(c, map[string]any{"dsl": tpl.DSL}, "success")
+}
diff --git a/internal/handler/pipeline_test.go b/internal/handler/pipeline_test.go
index c95ec13cbf..fbb7bb6e6a 100644
--- a/internal/handler/pipeline_test.go
+++ b/internal/handler/pipeline_test.go
@@ -27,12 +27,29 @@ import (
 	pipelinepkg "ragflow/internal/ingestion/pipeline"
 )
 
-// fakePipelineLister is a test double for builtinPipelineLister.
-type fakePipelineLister struct {
+// fakePipelineRegistry is a test double for builtinPipelineRegistry.
+type fakePipelineRegistry struct {
 	items []*pipelinepkg.BuiltinPipelineMeta
 }
 
-func (f fakePipelineLister) List() []*pipelinepkg.BuiltinPipelineMeta { return f.items }
+func (f fakePipelineRegistry) List() *pipelinepkg.BuiltinPipelineListResponse {
+	return &pipelinepkg.BuiltinPipelineListResponse{
+		BuiltinPipelines: f.items,
+		Total:            int64(len(f.items)),
+	}
+}
+
+func (f fakePipelineRegistry) Get(ref string) (*pipelinepkg.BuiltinPipeline, bool) {
+	for _, item := range f.items {
+		if item.ParserID == ref {
+			return &pipelinepkg.BuiltinPipeline{
+				BuiltinPipelineMeta: *item,
+				DSL:                 map[string]any{"components": map[string]any{}},
+			}, true
+		}
+	}
+	return nil, false
+}
 
 func pipelineCtx() (*gin.Context, *httptest.ResponseRecorder) {
 	gin.SetMode(gin.TestMode)
@@ -49,16 +66,14 @@ func TestPipelineHandler_ListPipelines_TypeBuiltin(t *testing.T) {
 			Title:       "General",
 			Description: "general desc",
 			Filename:    "ingestion_pipeline_general.json",
-			DSL:         map[string]any{"components": map[string]any{}},
 		},
 		{
 			ParserID: "book",
 			Title:    "Book",
 			Filename: "ingestion_pipeline_book.json",
-			DSL:      map[string]any{"components": map[string]any{}},
 		},
 	}
-	h := &PipelineHandler{registry: fakePipelineLister{items: items}}
+	h := &PipelineHandler{registry: fakePipelineRegistry{items: items}}
 
 	c, w := pipelineCtx()
 	h.ListPipelines(c)
@@ -68,10 +83,12 @@ func TestPipelineHandler_ListPipelines_TypeBuiltin(t *testing.T) {
 	}
 	var resp struct {
 		Code int `json:"code"`
-		Data []struct {
-			ParserID string         `json:"parser_id"`
-			Title    string         `json:"title"`
-			DSL      map[string]any `json:"dsl"`
+		Data struct {
+			Canvas []struct {
+				ID    string `json:"id"`
+				Title string `json:"title"`
+			} `json:"canvas"`
+			Total int64 `json:"total"`
 		} `json:"data"`
 	}
 	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
@@ -80,11 +97,14 @@ func TestPipelineHandler_ListPipelines_TypeBuiltin(t *testing.T) {
 	if resp.Code != int(common.CodeSuccess) {
 		t.Fatalf("code = %d, want %d", resp.Code, int(common.CodeSuccess))
 	}
-	if len(resp.Data) != 2 {
-		t.Fatalf("len(data) = %d, want 2", len(resp.Data))
+	if resp.Data.Total != 2 {
+		t.Fatalf("total = %d, want 2", resp.Data.Total)
 	}
-	if resp.Data[0].ParserID != "general" || resp.Data[0].Title != "General" {
-		t.Errorf("data[0] = %+v", resp.Data[0])
+	if len(resp.Data.Canvas) != 2 {
+		t.Fatalf("len(canvas) = %d, want 2", len(resp.Data.Canvas))
+	}
+	if resp.Data.Canvas[0].ID != "general" || resp.Data.Canvas[0].Title != "General" {
+		t.Errorf("canvas[0] = %+v", resp.Data.Canvas[0])
 	}
 }
 
@@ -92,7 +112,7 @@ func TestPipelineHandler_ListPipelines_NoTypeParam(t *testing.T) {
 	items := []*pipelinepkg.BuiltinPipelineMeta{
 		{ParserID: "general", Title: "General"},
 	}
-	h := &PipelineHandler{registry: fakePipelineLister{items: items}}
+	h := &PipelineHandler{registry: fakePipelineRegistry{items: items}}
 
 	gin.SetMode(gin.TestMode)
 	w := httptest.NewRecorder()
@@ -105,15 +125,21 @@ func TestPipelineHandler_ListPipelines_NoTypeParam(t *testing.T) {
 		t.Fatalf("status = %d, want 200", w.Code)
 	}
 	var resp struct {
-		Data []struct {
-			ParserID string `json:"parser_id"`
+		Data struct {
+			Canvas []struct {
+				ID string `json:"id"`
+			} `json:"canvas"`
+			Total int64 `json:"total"`
 		} `json:"data"`
 	}
 	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
 		t.Fatalf("unmarshal: %v", err)
 	}
-	if len(resp.Data) != 1 {
-		t.Fatalf("len(data) = %d, want 1 (no type param defaults to builtin)", len(resp.Data))
+	if resp.Data.Total != 1 {
+		t.Fatalf("total = %d, want 1", resp.Data.Total)
+	}
+	if len(resp.Data.Canvas) != 1 {
+		t.Fatalf("len(canvas) = %d, want 1 (no type param defaults to builtin)", len(resp.Data.Canvas))
 	}
 }
 
@@ -135,19 +161,21 @@ func TestPipelineHandler_ListPipelines_RealRegistry(t *testing.T) {
 	h.ListPipelines(c)
 
 	var resp struct {
-		Data []struct {
-			ParserID string `json:"parser_id"`
+		Data struct {
+			Canvas []struct {
+				ID string `json:"id"`
+			} `json:"canvas"`
 		} `json:"data"`
 	}
 	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
 		t.Fatalf("unmarshal: %v", err)
 	}
-	if len(resp.Data) == 0 {
+	if len(resp.Data.Canvas) == 0 {
 		t.Fatal("expected non-empty builtin pipeline list from real registry")
 	}
 	seen := map[string]bool{}
-	for _, item := range resp.Data {
-		seen[item.ParserID] = true
+	for _, item := range resp.Data.Canvas {
+		seen[item.ID] = true
 	}
 	if !seen["general"] {
 		t.Error("real registry listing missing general")
@@ -156,3 +184,65 @@ func TestPipelineHandler_ListPipelines_RealRegistry(t *testing.T) {
 		t.Error("alias naive must be hidden from listing")
 	}
 }
+
+func TestPipelineHandler_GetPipeline(t *testing.T) {
+	items := []*pipelinepkg.BuiltinPipelineMeta{
+		{ParserID: "general", Title: "General"},
+	}
+	h := &PipelineHandler{registry: fakePipelineRegistry{items: items}}
+
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest("GET", "/api/v1/pipelines/general", nil)
+	c.Params = gin.Params{{Key: "id", Value: "general"}}
+
+	h.GetPipeline(c)
+
+	if w.Code != 200 {
+		t.Fatalf("status = %d, want 200", w.Code)
+	}
+	var resp struct {
+		Code int `json:"code"`
+		Data struct {
+			DSL map[string]any `json:"dsl"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if resp.Code != int(common.CodeSuccess) {
+		t.Fatalf("code = %d, want %d", resp.Code, int(common.CodeSuccess))
+	}
+	if resp.Data.DSL == nil {
+		t.Fatal("dsl should not be nil")
+	}
+}
+
+func TestPipelineHandler_GetPipeline_NotFound(t *testing.T) {
+	items := []*pipelinepkg.BuiltinPipelineMeta{
+		{ParserID: "general", Title: "General"},
+	}
+	h := &PipelineHandler{registry: fakePipelineRegistry{items: items}}
+
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest("GET", "/api/v1/pipelines/nonexistent", nil)
+	c.Params = gin.Params{{Key: "id", Value: "nonexistent"}}
+
+	h.GetPipeline(c)
+
+	if w.Code != 200 {
+		t.Fatalf("status = %d, want 200 (error is in body code)", w.Code)
+	}
+	var resp struct {
+		Code int `json:"code"`
+	}
+	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if resp.Code == int(common.CodeSuccess) {
+		t.Fatal("expected error code for nonexistent pipeline")
+	}
+}
diff --git a/internal/httputil/ingestion_task_error.go b/internal/httputil/ingestion_task_error.go
deleted file mode 100644
index 784eed0eea..0000000000
--- a/internal/httputil/ingestion_task_error.go
+++ /dev/null
@@ -1,23 +0,0 @@
-package httputil
-
-import (
-	"errors"
-
-	"ragflow/internal/common"
-	"ragflow/internal/service"
-)
-
-func IngestionTaskErrorCode(err error) common.ErrorCode {
-	var transitionErr *service.InvalidTaskTransitionError
-	if errors.As(err, &transitionErr) {
-		return common.CodeConflict
-	}
-	var conflictErr *service.TaskStatusConflictError
-	if errors.As(err, &conflictErr) {
-		return common.CodeConflict
-	}
-	if errors.Is(err, common.ErrTaskNotFound) {
-		return common.CodeNotFound
-	}
-	return common.CodeExceptionError
-}
diff --git a/internal/ingestion/pipeline/builtin_registry.go b/internal/ingestion/pipeline/builtin_registry.go
index 184948ee17..01aab3870a 100644
--- a/internal/ingestion/pipeline/builtin_registry.go
+++ b/internal/ingestion/pipeline/builtin_registry.go
@@ -8,6 +8,7 @@ import (
 	"io/fs"
 	"os"
 	"path/filepath"
+	"regexp"
 	"sort"
 	"strings"
 	"sync"
@@ -21,12 +22,20 @@ var builtinTemplateFS embed.FS
 // BuiltinPipelineMeta is the API-facing metadata for one built-in ingestion
 // pipeline template. The ParserID field is the value stored in the dataset's
 // parser_id column for built-in pipelines.
+// JSON tag "id" aligns with AgentItem.id for format consistency with
+// GET /api/v1/agents?canvas_category=dataflow_canvas.
 type BuiltinPipelineMeta struct {
-	ParserID    string         `json:"parser_id"`
-	Title       string         `json:"title"`
-	Description string         `json:"description,omitempty"`
-	Filename    string         `json:"filename"`
-	DSL         map[string]any `json:"dsl"`
+	ParserID    string `json:"id"`
+	Title       string `json:"title"`
+	Description string `json:"description,omitempty"`
+	Filename    string `json:"filename"`
+}
+
+// BuiltinPipelineListResponse wraps the builtin pipeline list for
+// format consistency with ListAgentsResponse.
+type BuiltinPipelineListResponse struct {
+	BuiltinPipelines []*BuiltinPipelineMeta `json:"canvas"`
+	Total            int64                  `json:"total"`
 }
 
 type BuiltinPipeline struct {
@@ -119,7 +128,7 @@ func (r *Registry) canonicalRef(ref string) string {
 	return ref
 }
 
-func (r *Registry) List() []*BuiltinPipelineMeta {
+func (r *Registry) List() *BuiltinPipelineListResponse {
 	if r == nil {
 		return nil
 	}
@@ -131,7 +140,10 @@ func (r *Registry) List() []*BuiltinPipelineMeta {
 		cp := *item
 		out = append(out, &cp)
 	}
-	return out
+	return &BuiltinPipelineListResponse{
+		BuiltinPipelines: out,
+		Total:            int64(len(out)),
+	}
 }
 
 func (r *Registry) Refs() []string {
@@ -176,6 +188,39 @@ func (r *Registry) IsValid(ref string) bool {
 	return ok
 }
 
+// fileTypeParserOverrides maps a file type value to the canonical parser_id that
+// must be used for all files of that type. The type values are FileType constants
+// defined in the utility package.
+var fileTypeParserOverrides = map[string]string{
+	"visual": "picture",
+	"aural":  "audio",
+}
+
+var (
+	presentationExtPattern = regexp.MustCompile(`(?i)\.(ppt|pptx|pages)$`)
+	emailExtPattern        = regexp.MustCompile(`(?i)\.(msg|eml)$`)
+)
+
+// DefaultParserID returns the canonical parser_id for a document given its file
+// type, filename and the dataset-level parser_id. File-type-based overrides
+// (e.g. "visual" → "picture", "aural" → "audio") take precedence, followed by
+// filename extension heuristics (presentation / email), then falling back to
+// the dataset's parser_id.
+func (r *Registry) DefaultParserID(fileType, filename, fallback string) string {
+	if override, ok := fileTypeParserOverrides[strings.ToLower(fileType)]; ok {
+		return override
+	}
+	base := filepath.Base(strings.TrimSpace(filename))
+	switch {
+	case presentationExtPattern.MatchString(base):
+		return "presentation"
+	case emailExtPattern.MatchString(base):
+		return "email"
+	default:
+		return fallback
+	}
+}
+
 func parseTemplate(filename string, raw []byte) (*BuiltinPipeline, error) {
 	decoder := json.NewDecoder(bytes.NewReader(raw))
 	decoder.UseNumber()
@@ -197,7 +242,6 @@ func parseTemplate(filename string, raw []byte) (*BuiltinPipeline, error) {
 			Title:       englishText(data["title"], ref),
 			Description: englishText(data["description"], ""),
 			Filename:    filename,
-			DSL:         dsl,
 		},
 		DSL: dsl,
 	}
diff --git a/internal/ingestion/pipeline/builtin_registry_test.go b/internal/ingestion/pipeline/builtin_registry_test.go
index 0b4c2db041..be645052a3 100644
--- a/internal/ingestion/pipeline/builtin_registry_test.go
+++ b/internal/ingestion/pipeline/builtin_registry_test.go
@@ -20,9 +20,12 @@ func TestRegistryLoadsSummaries(t *testing.T) {
 	if err != nil {
 		t.Fatalf("NewRegistryFromDir: %v", err)
 	}
-	if got := len(r.List()); got != 2 {
+	if got := len(r.List().BuiltinPipelines); got != 2 {
 		t.Fatalf("len(List()) = %d, want 2", got)
 	}
+	if got := r.List().Total; got != 2 {
+		t.Fatalf("Total = %d, want 2", got)
+	}
 	tpl, ok := r.Get("general")
 	if !ok {
 		t.Fatal("expected general template")
@@ -164,7 +167,7 @@ func TestRegistryAliasNaiveResolvesGeneral(t *testing.T) {
 			t.Error("Refs() contains alias naive; aliases must be hidden from listing")
 		}
 	}
-	for _, meta := range r.List() {
+	for _, meta := range r.List().BuiltinPipelines {
 		if meta.ParserID == "naive" {
 			t.Error("List() contains alias naive; aliases must be hidden from listing")
 		}
@@ -238,3 +241,37 @@ func TestLoadBuiltinDSL_UnknownFails(t *testing.T) {
 		t.Fatal("LoadBuiltinDSL(nonexistent) = nil, want error")
 	}
 }
+
+func TestDefaultRegistry_DefaultParserID(t *testing.T) {
+	r, err := DefaultRegistry()
+	if err != nil {
+		t.Fatalf("DefaultRegistry: %v", err)
+	}
+
+	tests := []struct {
+		name     string
+		fileType string
+		filename string
+		fallback string
+		want     string
+	}{
+		{name: "visual", fileType: "visual", filename: "img.png", fallback: "naive", want: "picture"},
+		{name: "aural", fileType: "aural", filename: "audio.mp3", fallback: "naive", want: "audio"},
+		{name: "presentation by ext", fileType: "doc", filename: "deck.pptx", fallback: "naive", want: "presentation"},
+		{name: "email by ext", fileType: "doc", filename: "mail.eml", fallback: "naive", want: "email"},
+		{name: "pages as presentation", fileType: "doc", filename: "slides.pages", fallback: "naive", want: "presentation"},
+		{name: "msg as email", fileType: "doc", filename: "message.msg", fallback: "naive", want: "email"},
+		{name: "fallback default", fileType: "doc", filename: "notes.txt", fallback: "manual", want: "manual"},
+		{name: "unknown filetype fallback", fileType: "unknown", filename: "file.bin", fallback: "general", want: "general"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := r.DefaultParserID(tt.fileType, tt.filename, tt.fallback)
+			if got != tt.want {
+				t.Fatalf("DefaultParserID(%q, %q, %q) = %q, want %q",
+					tt.fileType, tt.filename, tt.fallback, got, tt.want)
+			}
+		})
+	}
+}
diff --git a/internal/ingestion/pipeline/component_params.go b/internal/ingestion/pipeline/component_params.go
new file mode 100644
index 0000000000..fcb975ff91
--- /dev/null
+++ b/internal/ingestion/pipeline/component_params.go
@@ -0,0 +1,101 @@
+package pipeline
+
+import (
+	"encoding/json"
+	"fmt"
+)
+
+// ComponentParamsSchema describes the full set of user-configurable parameters
+// for one component in a DSL. The frontend uses this to render editable forms;
+// the API uses it for validation.
+type ComponentParamsSchema struct {
+	// CpnID is the component instance identifier in the DSL, e.g. "Parser:HipSignsRhyme".
+	CpnID string `json:"cpn_id"`
+	// ComponentName is the logical component type, e.g. "Parser", "Tokenizer".
+	ComponentName string `json:"component_name"`
+	// ParamsDefaults holds the params map from the DSL with "outputs" excluded.
+	// Keys are param names (e.g. "setups", "fields", "chunk_token_size");
+	// values are their DSL-baked defaults.
+	ParamsDefaults map[string]any `json:"params_defaults"`
+}
+
+// ExtractAllComponentParams extracts the params schema for every component
+// present in the DSL JSON. It excludes the internal "outputs" key from each
+// component's params — those are wire-format definitions, not user-configurable.
+func ExtractAllComponentParams(dslJSON []byte) ([]ComponentParamsSchema, error) {
+	var dsl struct {
+		Components map[string]struct {
+			Obj struct {
+				ComponentName string         `json:"component_name"`
+				Params        map[string]any `json:"params"`
+			} `json:"obj"`
+		} `json:"components"`
+	}
+	if err := json.Unmarshal(dslJSON, &dsl); err != nil {
+		return nil, fmt.Errorf("ExtractAllComponentParams: decode DSL: %w", err)
+	}
+	if dsl.Components == nil {
+		return nil, fmt.Errorf("ExtractAllComponentParams: DSL has no components")
+	}
+
+	out := make([]ComponentParamsSchema, 0, len(dsl.Components))
+	for cpnID, comp := range dsl.Components {
+		params := make(map[string]any, len(comp.Obj.Params))
+		for k, v := range comp.Obj.Params {
+			if k == "outputs" {
+				continue
+			}
+			params[k] = deepCopyValue(v)
+		}
+		out = append(out, ComponentParamsSchema{
+			CpnID:          cpnID,
+			ComponentName:  comp.Obj.ComponentName,
+			ParamsDefaults: params,
+		})
+	}
+	return out, nil
+}
+
+// deepCopyValue returns a deep copy of a value that might be a nested map or slice.
+func deepCopyValue(v any) any {
+	switch val := v.(type) {
+	case map[string]any:
+		cp := make(map[string]any, len(val))
+		for mk, mv := range val {
+			cp[mk] = deepCopyValue(mv)
+		}
+		return cp
+	case []any:
+		cp := make([]any, len(val))
+		for i, item := range val {
+			cp[i] = deepCopyValue(item)
+		}
+		return cp
+	default:
+		return v
+	}
+}
+
+// deepCopyMapStr deep-copies a map[string]any so the caller can safely mutate the result.
+func deepCopyMapStr(src map[string]any) map[string]any {
+	cp := make(map[string]any, len(src))
+	for k, v := range src {
+		cp[k] = deepCopyValue(v)
+	}
+	return cp
+}
+
+// ComponentParamsDefaults extracts component params from DSL JSON and returns
+// them in the component_params storage format {cpnID: {param: value}}.
+// Values are deep-copied from the DSL so callers can safely mutate the result.
+func ComponentParamsDefaults(dslJSON []byte) (map[string]map[string]any, error) {
+	schemas, err := ExtractAllComponentParams(dslJSON)
+	if err != nil {
+		return nil, err
+	}
+	out := make(map[string]map[string]any, len(schemas))
+	for _, s := range schemas {
+		out[s.CpnID] = deepCopyMapStr(s.ParamsDefaults)
+	}
+	return out, nil
+}
diff --git a/internal/ingestion/pipeline/component_params_test.go b/internal/ingestion/pipeline/component_params_test.go
new file mode 100644
index 0000000000..ce8798f0a8
--- /dev/null
+++ b/internal/ingestion/pipeline/component_params_test.go
@@ -0,0 +1,239 @@
+package pipeline
+
+import (
+	"encoding/json"
+	"io/fs"
+	"path/filepath"
+	"slices"
+	"strings"
+	"testing"
+)
+
+func TestExtractAllComponentParams(t *testing.T) {
+	// Minimal DSL with all component types (general template pattern).
+	dsl := map[string]any{
+		"components": map[string]any{
+			"File": map[string]any{
+				"obj": map[string]any{
+					"component_name": "File",
+					"params":         map[string]any{},
+				},
+			},
+			"Parser:HipSignsRhyme": map[string]any{
+				"obj": map[string]any{
+					"component_name": "Parser",
+					"params": map[string]any{
+						"outputs": map[string]any{"html": map[string]any{"type": "string"}},
+						"setups": map[string]any{
+							"pdf":         map[string]any{"parse_method": "deepdoc"},
+							"spreadsheet": map[string]any{"parse_method": "deepdoc"},
+						},
+					},
+				},
+			},
+			"Tokenizer:LegalReadersDecide": map[string]any{
+				"obj": map[string]any{
+					"component_name": "Tokenizer",
+					"params": map[string]any{
+						"fields":               "text",
+						"filename_embd_weight": 0.1,
+						"search_method":        []any{"embedding", "full_text"},
+						"outputs":              map[string]any{},
+					},
+				},
+			},
+		},
+	}
+	dslJSON, err := json.Marshal(dsl)
+	if err != nil {
+		t.Fatalf("marshal fixture: %v", err)
+	}
+
+	schemas, err := ExtractAllComponentParams(dslJSON)
+	if err != nil {
+		t.Fatalf("ExtractAllComponentParams: %v", err)
+	}
+
+	// Build a lookup by cpnID for assertion.
+	byCPN := make(map[string]ComponentParamsSchema, len(schemas))
+	for _, s := range schemas {
+		if _, dup := byCPN[s.CpnID]; dup {
+			t.Errorf("duplicate cpnID %q in result", s.CpnID)
+		}
+		byCPN[s.CpnID] = s
+	}
+
+	// --- File ---
+	file, ok := byCPN["File"]
+	if !ok {
+		t.Fatal("missing component: File")
+	}
+	if file.ComponentName != "File" {
+		t.Errorf("File.ComponentName = %q, want %q", file.ComponentName, "File")
+	}
+	if len(file.ParamsDefaults) != 0 {
+		t.Errorf("File.ParamsDefaults should be empty, got %v", file.ParamsDefaults)
+	}
+
+	// --- Parser ---
+	parser, ok := byCPN["Parser:HipSignsRhyme"]
+	if !ok {
+		t.Fatal("missing component: Parser:HipSignsRhyme")
+	}
+	if parser.ComponentName != "Parser" {
+		t.Errorf("Parser.ComponentName = %q, want %q", parser.ComponentName, "Parser")
+	}
+	// ParamsDefaults must contain "setups", must NOT contain "outputs".
+	if _, ok := parser.ParamsDefaults["setups"]; !ok {
+		t.Errorf("Parser.ParamsDefaults missing key %q", "setups")
+	}
+	if _, ok := parser.ParamsDefaults["outputs"]; ok {
+		t.Errorf("Parser.ParamsDefaults should NOT contain %q (excluded)", "outputs")
+	}
+	if setups, ok := parser.ParamsDefaults["setups"].(map[string]any); !ok {
+		t.Errorf("Parser.ParamsDefaults[\"setups\"] is not a map: %T", parser.ParamsDefaults["setups"])
+	} else {
+		if _, ok := setups["pdf"]; !ok {
+			t.Errorf("Parser setups missing key %q", "pdf")
+		}
+		if _, ok := setups["spreadsheet"]; !ok {
+			t.Errorf("Parser setups missing key %q", "spreadsheet")
+		}
+	}
+
+	// --- Tokenizer ---
+	tokenizer, ok := byCPN["Tokenizer:LegalReadersDecide"]
+	if !ok {
+		t.Fatal("missing component: Tokenizer:LegalReadersDecide")
+	}
+	if tokenizer.ComponentName != "Tokenizer" {
+		t.Errorf("Tokenizer.ComponentName = %q, want %q", tokenizer.ComponentName, "Tokenizer")
+	}
+	// ParamsDefaults must contain "fields", "filename_embd_weight", "search_method".
+	if _, ok := tokenizer.ParamsDefaults["fields"]; !ok {
+		t.Errorf("Tokenizer.ParamsDefaults missing key %q", "fields")
+	}
+	if _, ok := tokenizer.ParamsDefaults["filename_embd_weight"]; !ok {
+		t.Errorf("Tokenizer.ParamsDefaults missing key %q", "filename_embd_weight")
+	}
+	if _, ok := tokenizer.ParamsDefaults["search_method"]; !ok {
+		t.Errorf("Tokenizer.ParamsDefaults missing key %q", "search_method")
+	}
+	if _, ok := tokenizer.ParamsDefaults["outputs"]; ok {
+		t.Errorf("Tokenizer.ParamsDefaults should NOT contain %q (excluded)", "outputs")
+	}
+
+	// --- No extra/leaked ---
+	if len(schemas) != 3 {
+		t.Errorf("expected 3 components, got %d: %v", len(schemas), componentIDs(schemas))
+	}
+}
+
+func TestExtractAllComponentParams_EmptyDSL(t *testing.T) {
+	dsl := map[string]any{
+		"components": map[string]any{},
+	}
+	dslJSON, _ := json.Marshal(dsl)
+	schemas, err := ExtractAllComponentParams(dslJSON)
+	if err != nil {
+		t.Fatalf("ExtractAllComponentParams: %v", err)
+	}
+	if len(schemas) != 0 {
+		t.Errorf("expected 0 components, got %d", len(schemas))
+	}
+}
+
+func TestExtractAllComponentParams_InvalidJSON(t *testing.T) {
+	_, err := ExtractAllComponentParams([]byte("not json"))
+	if err == nil {
+		t.Error("expected error for invalid JSON")
+	}
+}
+
+func TestExtractAllComponentParams_MissingComponents(t *testing.T) {
+	dsl := map[string]any{"globals": map[string]any{}}
+	dslJSON, _ := json.Marshal(dsl)
+	_, err := ExtractAllComponentParams(dslJSON)
+	if err == nil {
+		t.Error("expected error for DSL missing components")
+	}
+}
+
+func componentIDs(schemas []ComponentParamsSchema) []string {
+	ids := make([]string, len(schemas))
+	for i, s := range schemas {
+		ids[i] = s.CpnID
+	}
+	slices.Sort(ids)
+	return ids
+}
+
+func TestExtractAllComponentParams_AllBuiltinTemplates(t *testing.T) {
+	entries, err := fs.ReadDir(builtinTemplateFS, "template")
+	if err != nil {
+		t.Fatalf("read embedded template dir: %v", err)
+	}
+
+	var tested int
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+		name := entry.Name()
+		if !strings.HasPrefix(name, builtinTemplatePrefix) || !strings.HasSuffix(name, ".json") {
+			continue
+		}
+		tested++
+		raw, err := fs.ReadFile(builtinTemplateFS, filepath.Join("template", name))
+		if err != nil {
+			t.Errorf("%s: read: %v", name, err)
+			continue
+		}
+
+		// The template file wraps the DSL in {"dsl": {...}}.
+		var wrapper struct {
+			DSL json.RawMessage `json:"dsl"`
+		}
+		if err := json.Unmarshal(raw, &wrapper); err != nil {
+			t.Errorf("%s: unmarshal wrapper: %v", name, err)
+			continue
+		}
+		if len(wrapper.DSL) == 0 {
+			t.Errorf("%s: missing dsl key", name)
+			continue
+		}
+
+		schemas, err := ExtractAllComponentParams(wrapper.DSL)
+		if err != nil {
+			t.Errorf("%s: ExtractAllComponentParams: %v", name, err)
+			continue
+		}
+		if len(schemas) == 0 {
+			t.Errorf("%s: expected at least 1 component, got 0", name)
+			continue
+		}
+
+		// Every schema must have a non-empty CpnID and ComponentName.
+		for _, s := range schemas {
+			if s.CpnID == "" {
+				t.Errorf("%s: component %q has empty CpnID", name, s.ComponentName)
+			}
+			if s.ComponentName == "" {
+				t.Errorf("%s: cpnID %q has empty ComponentName", name, s.CpnID)
+			}
+			if s.ParamsDefaults == nil {
+				t.Errorf("%s: cpnID %q has nil ParamsDefaults", name, s.CpnID)
+			}
+			// "outputs" must never leak into ParamsDefaults.
+			if _, ok := s.ParamsDefaults["outputs"]; ok {
+				t.Errorf("%s: cpnID %q ParamsDefaults contains excluded key %q", name, s.CpnID, "outputs")
+			}
+		}
+
+		t.Logf("%s: %d components: %v", name, len(schemas), componentIDs(schemas))
+	}
+	if tested == 0 {
+		t.Fatal("no builtin templates found")
+	}
+	t.Logf("tested %d builtin templates", tested)
+}
diff --git a/internal/ingestion/service/doc_state.go b/internal/ingestion/service/doc_state.go
index 2ad3eefbf8..1167599d2a 100644
--- a/internal/ingestion/service/doc_state.go
+++ b/internal/ingestion/service/doc_state.go
@@ -59,7 +59,7 @@ func (u *docStateUpdater) apply(r *taskpkg.PipelineResult) {
 			common.Warn(fmt.Sprintf("failed to update document metadata: %v", err))
 		}
 	}
-	if err := u.docSvc.IncrementChunkNum(r.DocID, r.KbID, r.ChunkCount, r.TokenConsumption, 0); err != nil {
+	if err := u.docSvc.IncrementChunkNum(r.DocID, r.KbID, r.ChunkCount, r.TokenConsumption, r.Duration); err != nil {
 		common.Warn(fmt.Sprintf("failed to increment chunk num: %v", err))
 	}
 }
diff --git a/internal/ingestion/service/execute_task_ack_test.go b/internal/ingestion/service/execute_task_ack_test.go
index 6b6a659047..b7b34de665 100644
--- a/internal/ingestion/service/execute_task_ack_test.go
+++ b/internal/ingestion/service/execute_task_ack_test.go
@@ -170,13 +170,20 @@ func TestExecuteTask_HeartbeatsInProgressDuringLongTask(t *testing.T) {
 	go ingestor.executeTask(newAckTaskCtx(context.Background(), taskID, docID, handle))
 
 	<-started
-	time.Sleep(30 * time.Millisecond) // let the ticker fire a few times
-	close(proceed)                    // release the long task
 
+	// Poll for heartbeats with a generous deadline so the test is resilient
+	// to slow CI schedulers. The ticker fires every heartbeatInterval (5ms);
+	// the first tick may be delayed if the goroutine is not scheduled promptly.
+	heartbeatDeadline := time.Now().Add(2 * time.Second)
+	for handle.inProgress.Load() == 0 && time.Now().Before(heartbeatDeadline) {
+		time.Sleep(5 * time.Millisecond)
+	}
 	if handle.inProgress.Load() == 0 {
 		t.Fatal("expected InProgress heartbeats while runDocumentTask was blocked, got 0")
 	}
 
+	close(proceed) // release the long task — only after confirming heartbeats
+
 	// Poll for Ack completion (executeTask must finish MarkCompleted + Ack).
 	deadline := time.Now().Add(2 * time.Second)
 	for handle.acks.Load() == 0 && time.Now().Before(deadline) {
diff --git a/internal/ingestion/service/ingestion_service.go b/internal/ingestion/service/ingestion_service.go
index 2f44ee5746..755b62bf58 100644
--- a/internal/ingestion/service/ingestion_service.go
+++ b/internal/ingestion/service/ingestion_service.go
@@ -21,6 +21,7 @@ import (
 	"errors"
 	"fmt"
 	"ragflow/internal/utility"
+	"runtime"
 	"strings"
 	"sync"
 	"sync/atomic"
@@ -83,7 +84,7 @@ type Ingestor struct {
 
 func NewIngestor(name string, maxConcurrency int32, supportedTypes []string) *Ingestor {
 	if maxConcurrency <= 0 {
-		maxConcurrency = 1
+		maxConcurrency = int32(runtime.NumCPU())
 	}
 	ctx, cancel := context.WithCancel(context.Background())
 	id := utility.GenerateUUID()
diff --git a/internal/ingestion/task/pipeline_executor.go b/internal/ingestion/task/pipeline_executor.go
index 4d12c9979e..ca7639fff7 100644
--- a/internal/ingestion/task/pipeline_executor.go
+++ b/internal/ingestion/task/pipeline_executor.go
@@ -40,6 +40,7 @@ type PipelineResult struct {
 	Metadata         map[string]any
 	ChunkCount       int
 	TokenConsumption int
+	Duration         float64 // pipeline wall-clock seconds
 }
 
 type PipelineExecutor struct {
@@ -52,7 +53,7 @@ type PipelineExecutor struct {
 	loadDSLFunc     func(ctx context.Context, canvasID string) (string, string, error)
 	runPipelineFunc func(ctx context.Context, dsl string) (map[string]any, string, error)
 	progressSink    pipelinepkg.ProgressSink
-	requireResume   bool // when true, defaultRunPipeline passes WithRequireResume to the pipeline
+	requireResume   bool // when true, the pipeline run passes WithRequireResume
 }
 
 func validateTaskContext(taskCtx *TaskContext) error {
@@ -148,6 +149,7 @@ func (s *PipelineExecutor) Doc() *entity.Document     { return &s.taskCtx.Doc }
 func (s *PipelineExecutor) Tenant() *entity.Tenant    { return &s.taskCtx.Tenant }
 
 func (s *PipelineExecutor) Execute(ctx context.Context) (*PipelineResult, error) {
+	start := time.Now()
 	if err := ctx.Err(); err != nil {
 		return nil, err
 	}
@@ -170,7 +172,7 @@ func (s *PipelineExecutor) Execute(ctx context.Context) (*PipelineResult, error)
 		return nil, nil
 	}
 
-	result, err := s.processOutput(ctx, pipelineOutput)
+	result, err := s.processOutput(ctx, pipelineOutput, start)
 	if err != nil {
 		return nil, err
 	}
@@ -181,7 +183,7 @@ func (s *PipelineExecutor) Execute(ctx context.Context) (*PipelineResult, error)
 	return result, nil
 }
 
-func (s *PipelineExecutor) processOutput(ctx context.Context, pipelineOutput map[string]any) (*PipelineResult, error) {
+func (s *PipelineExecutor) processOutput(ctx context.Context, pipelineOutput map[string]any, start time.Time) (*PipelineResult, error) {
 	if pipelineOutput == nil {
 		return nil, nil
 	}
@@ -228,6 +230,7 @@ func (s *PipelineExecutor) processOutput(ctx context.Context, pipelineOutput map
 		Metadata:         metadata,
 		ChunkCount:       len(chunks),
 		TokenConsumption: embeddingTokenConsumption,
+		Duration:         time.Since(start).Seconds(),
 	}, nil
 }
 
@@ -281,6 +284,33 @@ func (s *PipelineExecutor) loadDSLFromCanvas(ctx context.Context, canvasID strin
 	return string(raw), canvasID, nil
 }
 
+// warnUnknownComponentParams logs a warning for any component id in the
+// parserConfig whose id is absent from the pipeline DSL. The runtime merge
+// (component params -> PatchDSL / override_params) silently drops such
+// entries, so we surface them here for operability. API-side validation
+// already rejects unknown ids on write; this is purely a defensive guard
+// for legacy/stale rows.
+func warnUnknownComponentParams(dsl string, parserConfig map[string]any) {
+	if len(parserConfig) == 0 {
+		return
+	}
+	schemas, err := pipelinepkg.ExtractAllComponentParams([]byte(dsl))
+	if err != nil {
+		common.Warn(fmt.Sprintf("warnUnknownComponentParams: cannot parse DSL to validate component params: %v", err))
+		return
+	}
+	dslCPNs := make(map[string]struct{}, len(schemas))
+	for _, s := range schemas {
+		dslCPNs[s.CpnID] = struct{}{}
+	}
+	for cpnID := range parserConfig {
+		if _, ok := dslCPNs[cpnID]; !ok {
+			common.Warn(fmt.Sprintf(
+				"parser_config references cpnID %q not present in the pipeline DSL; it will be ignored at runtime", cpnID))
+		}
+	}
+}
+
 func (s *PipelineExecutor) runPipelineWithDSL(ctx context.Context, dsl string) (map[string]any, string, error) {
 	if s == nil || s.taskCtx == nil {
 		return nil, dsl, fmt.Errorf("pipeline executor: nil task context")
@@ -288,6 +318,12 @@ func (s *PipelineExecutor) runPipelineWithDSL(ctx context.Context, dsl string) (
 
 	parserConfig := map[string]interface{}(s.taskCtx.Doc.ParserConfig)
 	common.InjectExtractorLLMID(parserConfig, s.taskCtx.Tenant.LLMID)
+
+	// Surface component params whose cpnID is absent from the DSL. The
+	// runtime merge (PatchDSL / override_params) silently drops such entries;
+	// API-side validation already rejects unknown ids on write, so this is a
+	// defensive guard for legacy/stale rows.
+	warnUnknownComponentParams(dsl, parserConfig)
 	patchedDSL, err := pipelinepkg.PatchDSL(dsl, parserConfig)
 	if err != nil {
 		return nil, dsl, fmt.Errorf("patch dsl with parser_config: %w", err)
diff --git a/internal/ingestion/task/pipeline_executor_defaults_test.go b/internal/ingestion/task/pipeline_executor_defaults_test.go
new file mode 100644
index 0000000000..8ba0ced19b
--- /dev/null
+++ b/internal/ingestion/task/pipeline_executor_defaults_test.go
@@ -0,0 +1,184 @@
+//
+//  Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+//
+//  Licensed under the Apache License, Version 2.0 (the "License");
+//  you may not use this file except in compliance with the License.
+//  You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+//  Unless required by applicable law or agreed to in writing, software
+//  distributed under the License is distributed on an "AS IS" BASIS,
+//  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+//  See the License for the specific language governing permissions and
+//  limitations under the License.
+//
+
+package task
+
+import (
+	"encoding/json"
+	"reflect"
+	"testing"
+
+	"ragflow/internal/entity"
+	pipelinepkg "ragflow/internal/ingestion/pipeline"
+)
+
+// builtinComponentParamsGolden holds hardcoded golden values of
+// pipeline.ComponentParamsDefaults for every built-in ingestion template.
+// Each entry is the compact JSON serialization of the resolved default
+// component params (the "outputs" wire key is already excluded).
+//
+// These are NOT derived at runtime: they pin the exact default params that
+// each template's DSL currently bakes in. If a template's DSL default values
+// change, the matching entry here MUST be updated in lockstep, otherwise the
+// corresponding test method fails and flags the regression.
+//
+// Comparison is done per-component (see assertComponentsMatch), so the JSON
+// serialization order inside each entry is irrelevant.
+var builtinComponentParamsGolden = map[string]string{
+	"audio":        "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:SongsFillAir\":{\"audio\":{\"output_format\":\"text\",\"preprocess\":[\"main_content\"],\"suffix\":[\"aac\",\"aiff\",\"ape\",\"au\",\"da\",\"flac\",\"midi\",\"mp3\",\"ogg\",\"oggvorbis\",\"realaudio\",\"vqf\",\"wav\",\"wave\",\"wma\"]}},\"TokenChunker:BlueSkiesLaugh\":{},\"Tokenizer:KindEyesWatch\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"book":         "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"doc\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"doc\"]},\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"html\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"htm\",\"html\"]},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"remove_toc\":true,\"suffix\":[\"pdf\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"TitleChunker:GrumpyGarlicsBake\":{\"hierarchy\":5,\"include_heading_content\":true,\"levels\":[[\"^#[^#]\",\"^##[^#]\",\"^###[^#]\",\"^####[^#]\"],[\"第[零一二三四五六七八九十百0-9]+(分?编|部分)\",\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"第[零一二三四五六七八九十百0-9]+条\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\"],[\"第[0-9]+章\",\"第[0-9]+节\",\"[0-9]{1,2}[\\\\. 、]\",\"[0-9]{1,2}\\\\.[0-9]{1,2}($|[^a-zA-Z/%~.-])\",\"[0-9]{1,2}\\\\.[0-9]{1,2}\\\\.[0-9]{1,2}\"],[\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"[零一二三四五六七八九十百]+[ 、]\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\",\"[\\\\(（][0-9]{,2}[\\\\)）]\"],[\"PART (ONE|TWO|THREE|FOUR|FIVE|SIX|SEVEN|EIGHT|NINE|TEN)\",\"Chapter (I+V?|VI*|XI|IX|X)\",\"Section [0-9]+\",\"Article [0-9]+\"]],\"method\":\"hierarchy\"},\"Tokenizer:HotDonutsRing\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"email":        "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:BirdsFlutterHigh\":{\"email\":{\"fields\":[\"from\",\"to\",\"cc\",\"bcc\",\"date\",\"subject\",\"body\",\"attachments\"],\"output_format\":\"text\",\"preprocess\":[\"main_content\"],\"suffix\":[\"eml\"]}},\"TokenChunker:WarmBreadSmells\":{\"children_delimiters\":[],\"chunk_token_size\":512,\"delimiter_mode\":\"token_size\",\"delimiters\":[\"\\n\",\"!\",\"?\",\"。\",\"；\",\"！\",\"？\"],\"image_context_size\":0,\"overlapped_percent\":0,\"table_context_size\":0},\"Tokenizer:NiceWordsSpoken\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"general":      "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"doc\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"doc\"]},\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"html\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"htm\",\"html\"]},\"markdown\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"md\",\"markdown\",\"mdx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"spreadsheet\":{\"flatten_media_to_text\":false,\"output_format\":\"html\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"xls\",\"xlsx\",\"csv\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\",\"py\",\"js\",\"java\",\"c\",\"cpp\",\"h\",\"php\",\"go\",\"ts\",\"sh\",\"cs\",\"kt\",\"sql\"]}},\"TokenChunker:SixApplesFall\":{\"children_delimiters\":[],\"chunk_token_size\":512,\"delimiter_mode\":\"token_size\",\"delimiters\":[\"\\n\",\"!\",\"?\",\"。\",\"；\",\"！\",\"？\"],\"image_context_size\":0,\"overlapped_percent\":0,\"table_context_size\":0},\"Tokenizer:LegalReadersDecide\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"laws":         "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"doc\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"doc\"]},\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"html\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"htm\",\"html\"]},\"markdown\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"md\",\"markdown\",\"mdx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"TitleChunker:SpicyKeysKick\":{\"hierarchy\":2,\"include_heading_content\":false,\"levels\":[[\"^#[^#]\",\"^##[^#]\",\"^###[^#]\",\"^####[^#]\"],[\"第[零一二三四五六七八九十百0-9]+(分?编|部分)\",\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"第[零一二三四五六七八九十百0-9]+条\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\"],[\"第[0-9]+章\",\"第[0-9]+节\",\"[0-9]{1,2}[\\\\. 、]\",\"[0-9]{1,2}\\\\.[0-9]{1,2}($|[^a-zA-Z/%~.-])\",\"[0-9]{1,2}\\\\.[0-9]{1,2}\\\\.[0-9]{1,2}\"],[\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"[零一二三四五六七八九十百]+[ 、]\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\",\"[\\\\(（][0-9]{,2}[\\\\)）]\"],[\"PART (ONE|TWO|THREE|FOUR|FIVE|SIX|SEVEN|EIGHT|NINE|TEN)\",\"Chapter (I+V?|VI*|XI|IX|X)\",\"Section [0-9]+\",\"Article [0-9]+\"]],\"method\":\"hierarchy\"},\"Tokenizer:PublicJobsTake\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"manual":       "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"doc\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"doc\"]},\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}}},\"TitleChunker:NineInsectsFind\":{\"hierarchy\":0,\"include_heading_content\":false,\"levels\":[[\"^#[^#]\",\"^##[^#]\",\"^###[^#]\",\"^####[^#]\"],[\"第[零一二三四五六七八九十百0-9]+(分?编|部分)\",\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"第[零一二三四五六七八九十百0-9]+条\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\"],[\"第[0-9]+章\",\"第[0-9]+节\",\"[0-9]{1,2}[\\\\. 、]\",\"[0-9]{1,2}\\\\.[0-9]{1,2}($|[^a-zA-Z/%~.-])\",\"[0-9]{1,2}\\\\.[0-9]{1,2}\\\\.[0-9]{1,2}\"],[\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"[零一二三四五六七八九十百]+[ 、]\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\",\"[\\\\(（][0-9]{,2}[\\\\)）]\"],[\"PART (ONE|TWO|THREE|FOUR|FIVE|SIX|SEVEN|EIGHT|NINE|TEN)\",\"Chapter (I+V?|VI*|XI|IX|X)\",\"Section [0-9]+\",\"Article [0-9]+\"]],\"method\":\"group\"},\"Tokenizer:FunnyBalloonsGrin\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"one":          "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"OneChunker:DryDrinksVisit\":{},\"Parser:HipSignsRhyme\":{\"doc\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"doc\"]},\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"html\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"htm\",\"html\"]},\"markdown\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"md\",\"markdown\",\"mdx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"spreadsheet\":{\"flatten_media_to_text\":false,\"output_format\":\"html\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"xls\",\"xlsx\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"Tokenizer:FrankWeeksListen\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"paper":        "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"pdf\":{\"enable_multi_column\":true,\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}}},\"TitleChunker:SparklySchoolsTravel\":{\"hierarchy\":0,\"include_heading_content\":false,\"levels\":[[\"^#[^#]\",\"^##[^#]\",\"^###[^#]\",\"^####[^#]\"],[\"第[零一二三四五六七八九十百0-9]+(分?编|部分)\",\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"第[零一二三四五六七八九十百0-9]+条\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\"],[\"第[0-9]+章\",\"第[0-9]+节\",\"[0-9]{1,2}[\\\\. 、]\",\"[0-9]{1,2}\\\\.[0-9]{1,2}($|[^a-zA-Z/%~.-])\",\"[0-9]{1,2}\\\\.[0-9]{1,2}\\\\.[0-9]{1,2}\"],[\"第[零一二三四五六七八九十百0-9]+章\",\"第[零一二三四五六七八九十百0-9]+节\",\"[零一二三四五六七八九十百]+[ 、]\",\"[\\\\(（][零一二三四五六七八九十百]+[\\\\)）]\",\"[\\\\(（][0-9]{,2}[\\\\)）]\"],[\"PART (ONE|TWO|THREE|FOUR|FIVE|SIX|SEVEN|EIGHT|NINE|TEN)\",\"Chapter (I+V?|VI*|XI|IX|X)\",\"Section [0-9]+\",\"Article [0-9]+\"]],\"method\":\"group\"},\"Tokenizer:GreatCarsWash\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"picture":      "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:ViewsCaptureLight\":{\"image\":{\"output_format\":\"text\",\"parse_method\":\"ocr\",\"preprocess\":[\"main_content\"],\"suffix\":[\"bmp\",\"gif\",\"jpeg\",\"jpg\",\"png\",\"svg\",\"tif\",\"tiff\",\"webp\"]},\"video\":{\"output_format\":\"text\",\"preprocess\":[\"main_content\"],\"suffix\":[\"3gp\",\"3gpp\",\"avi\",\"flv\",\"mkv\",\"mov\",\"mp4\",\"mpeg\",\"mpg\",\"webm\",\"wmv\"]}},\"TokenChunker:BrightColorsGlow\":{},\"Tokenizer:SharpLensFocus\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"presentation": "{\"Extractor:AutoExtractDefault\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"\",\"llm_id\":\"\"},\"File\":{},\"Parser:HipSignsRhyme\":{\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"slides\":{\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pptx\",\"ppt\"]}},\"PresentationChunker:HappyHillsGlow\":{},\"Tokenizer:TallTreesDance\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"qa":           "{\"File\":{},\"Parser:HipSignsRhyme\":{\"docx\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"markdown\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"md\",\"markdown\",\"mdx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":false,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"spreadsheet\":{\"flatten_media_to_text\":false,\"output_format\":\"html\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"xls\",\"xlsx\",\"csv\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"QAChunker:TidyCloudsThink\":{},\"Tokenizer:ColdCloudsDream\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"resume":       "{\"Extractor:ThreeDrinksAct\":{\"auto_keywords\":0,\"auto_questions\":0,\"field_name\":\"metadata\",\"frequencyPenaltyEnabled\":true,\"frequency_penalty\":0.7,\"llm_id\":\"THUDM/GLM-4.1V-9B-Thinking@SILICONFLOW\",\"maxTokensEnabled\":false,\"max_tokens\":256,\"presencePenaltyEnabled\":true,\"presence_penalty\":0.4,\"prompts\":[{\"content\":\"Content: {TitleChunker:FlatMiceFix@chunks}\",\"role\":\"user\"}],\"sys_prompt\":\"Act as a precise resume metadata extractor. Extract stable, chunk-supported metadata from the provided resume content.\\n\\nRules:\\n1. Use only information explicitly stated in the content. Do not infer, guess, normalize, or add missing facts.\\n2. The input may be only one chunk of a resume. Extract only what this content directly supports.\\n3. Use only these field names:\\ncandidate_name, gender, phone, email, city, location, nationality, linkedin, github, website, highest_degree, degree_levels, school_names, majors, graduation_years, work_experience_years, current_job_title, job_titles, company_names, job_experience, industries, target_job_titles, target_locations, employment_types, skills, certificates, awards, summary_tags\\n4. Ignore detailed responsibilities, project descriptions, achievement narratives, self-evaluation, and other low-value local details.\\n5. Keep values in the same language as the source text whenever possible.\\n6. Remove duplicates and keep only concise, high-value metadata.\\n7. Return only fields that are explicitly supported by the content. Do not return empty or unsupported fields.\\n\\nField guidance:\\n- highest_degree: highest explicit degree level mentioned\\n- degree_levels: all explicit degree levels mentioned\\n- school_names: explicit school, college, or university names\\n- majors: explicit fields of study\\n- graduation_years: explicit graduation years only\\n- work_experience_years: only if explicitly stated\\n- current_job_title: only if explicitly current or most recent\\n- job_titles: explicit role titles\\n- company_names: explicit employer names\\n- job_experience: concise structured work entries explicitly supported by the content, preferably including title, company, and time information when available\\n- industries: explicit industry names only\\n- target_job_titles: explicit desired roles only\\n- target_locations: explicit desired work locations only\\n- skills: concise, core, search-useful skills explicitly mentioned\\n- certificates: explicit certificate names only\\n- awards: explicit award names only\\n- summary_tags: short, high-value tags strictly supported by the content\\n\\nReturn only the extracted metadata. Do not output explanatory text.\",\"temperature\":0.1,\"temperatureEnabled\":true,\"tenant_llm_id\":29,\"topPEnabled\":true,\"top_p\":0.3},\"File\":{},\"Parser:HipSignsRhyme\":{\"docx\":{\"flatten_media_to_text\":true,\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"docx\"],\"vlm\":{}},\"pdf\":{\"flatten_media_to_text\":true,\"output_format\":\"json\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"pdf\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"TitleChunker:FlatMiceFix\":{\"hierarchy\":1,\"include_heading_content\":false,\"levels\":[[\"^\\\\s*(?i:(?:\\\\d+[\\\\.\\\\)]\\\\s*)?(?:EDUCATION|ACADEMIC\\\\s*BACKGROUND|ACADEMIC\\\\s*HISTORY|EDUCATIONAL\\\\s*BACKGROUND|RELEVANT\\\\s*COURSEWORK|COURSEWORK|EXPERIENCE|WORK\\\\s*EXPERIENCE|PROFESSIONAL\\\\s*EXPERIENCE|RELEVANT\\\\s*EXPERIENCE|EMPLOYMENT\\\\s*HISTORY|CAREER\\\\s*HISTORY|INTERNSHIP\\\\s*EXPERIENCE|PROJECTS|PROJECT\\\\s*EXPERIENCE|ACADEMIC\\\\s*PROJECTS|PROFESSIONAL\\\\s*PROJECTS|SKILLS|TECHNICAL\\\\s*SKILLS|CORE\\\\s*COMPETENCIES|COMPETENCIES|QUALIFICATIONS|SUMMARY\\\\s*OF\\\\s*QUALIFICATIONS|CERTIFICATIONS|LICENSES|CERTIFICATES|AWARDS|HONORS|HONOURS|ACHIEVEMENTS|PUBLICATIONS|RESEARCH|RESEARCH\\\\s*EXPERIENCE|LEADERSHIP|LEADERSHIP\\\\s*EXPERIENCE|ACTIVITIES|EXTRACURRICULAR\\\\s*ACTIVITIES|ACTIVITIES\\\\s*(?:\\u0026|AND)\\\\s*SKILLS|INVOLVEMENT|CAMPUS\\\\s*INVOLVEMENT|VOLUNTEER\\\\s*EXPERIENCE|VOLUNTEERING|COMMUNITY\\\\s*SERVICE|LANGUAGES|INTERESTS|HOBBIES|PROFILE|PROFESSIONAL\\\\s*PROFILE|SUMMARY|PROFESSIONAL\\\\s*SUMMARY|CAREER\\\\s*SUMMARY|OBJECTIVE|CAREER\\\\s*OBJECTIVE|PERSONAL\\\\s*INFORMATION|CONTACT\\\\s*INFORMATION|ADDITIONAL\\\\s*INFORMATION|TRAINING))\\\\s*[:：]?\\\\s*$\"],[\"^\\\\s*(?:\\\\d+[\\\\.、\\\\)]\\\\s*)?(?:教育背景|教育经历|学历背景|学术背景|技术背景|工作经历|工作经验|实习经历|项目经历|项目经验|科研经历|研究经历|校园经历|实践经历|专业经历|职业经历|技能|专业技能|技能特长|核心技能|技术栈|个人技能|工作技能|职业技能|技能与评价|技能与自我评价|工作技能与自我评价|职业技能与自我评价|证书|资格证书|职业资格|资质证书|获奖情况|获奖经历|荣誉|荣誉奖项|奖项|科研成果|论文发表|发表论文|领导经历|学生工作|校园活动|社团经历|活动经历|志愿经历|志愿服务|社会实践|语言能力|语言|自我评价|个人评价|自我总结|个人总结|个人优势|个人简介|个人信息|基本信息|联系方式|求职意向|应聘意向|职业目标|求职目标|兴趣爱好|兴趣特长|培训经历|其他信息|附加信息)\\\\s*[:：]?\\\\s*$\"]],\"method\":\"hierarchy\",\"root_chunk_as_heading\":true},\"Tokenizer:KindHandsWin\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"table":        "{\"File\":{},\"Parser:HipSignsRhyme\":{\"spreadsheet\":{\"column_mode\":\"auto\",\"column_names\":[],\"column_roles\":{},\"flatten_media_to_text\":false,\"output_format\":\"html\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"xls\",\"xlsx\",\"csv\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"TableChunker:FastFoxesJump\":{},\"Tokenizer:DeepLakesShine\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+	"tag":          "{\"File\":{},\"Parser:HipSignsRhyme\":{\"spreadsheet\":{\"flatten_media_to_text\":false,\"output_format\":\"html\",\"parse_method\":\"DeepDOC\",\"preprocess\":[\"main_content\"],\"suffix\":[\"xls\",\"xlsx\",\"csv\"],\"vlm\":{}},\"text\\u0026code\":{\"output_format\":\"json\",\"preprocess\":[\"main_content\"],\"suffix\":[\"txt\"]}},\"TagChunker:NewNoonsGlow\":{},\"Tokenizer:OldOwlsWatch\":{\"fields\":\"text\",\"filename_embd_weight\":0.1,\"search_method\":[\"embedding\",\"full_text\"]}}",
+}
+
+// Per-template test methods. Each resolves default component params from a
+// built-in template DSL and verifies: (1) they match the hardcoded golden,
+// (2) they survive a JSON round-trip through entity.JSONMap (DB storage).
+func TestBuildComponentParams_Audio(t *testing.T)   { assertTemplateComponentParams(t, "audio") }
+func TestBuildComponentParams_Book(t *testing.T)    { assertTemplateComponentParams(t, "book") }
+func TestBuildComponentParams_Email(t *testing.T)   { assertTemplateComponentParams(t, "email") }
+func TestBuildComponentParams_General(t *testing.T) { assertTemplateComponentParams(t, "general") }
+func TestBuildComponentParams_Laws(t *testing.T)    { assertTemplateComponentParams(t, "laws") }
+func TestBuildComponentParams_Manual(t *testing.T)  { assertTemplateComponentParams(t, "manual") }
+func TestBuildComponentParams_One(t *testing.T)     { assertTemplateComponentParams(t, "one") }
+func TestBuildComponentParams_Paper(t *testing.T)   { assertTemplateComponentParams(t, "paper") }
+func TestBuildComponentParams_Picture(t *testing.T) { assertTemplateComponentParams(t, "picture") }
+func TestBuildComponentParams_Presentation(t *testing.T) {
+	assertTemplateComponentParams(t, "presentation")
+}
+func TestBuildComponentParams_Qa(t *testing.T)     { assertTemplateComponentParams(t, "qa") }
+func TestBuildComponentParams_Resume(t *testing.T) { assertTemplateComponentParams(t, "resume") }
+func TestBuildComponentParams_Table(t *testing.T)  { assertTemplateComponentParams(t, "table") }
+func TestBuildComponentParams_Tag(t *testing.T)    { assertTemplateComponentParams(t, "tag") }
+
+// assertTemplateComponentParams resolves the default component params for the
+// given built-in template and verifies two layers per-component:
+//
+//  1. Layer 1 (DSL parse): pipeline.ComponentParamsDefaults returns exactly the
+//     hardcoded golden.
+//  2. Layer 2 (storage round-trip): the parsed result survives JSON
+//     marshal→entity.JSONMap→unmarshal, simulating the GORM DB round-trip that
+//     runPipelineWithDSL's parserConfig reads from Doc.ParserConfig.
+func assertTemplateComponentParams(t *testing.T, ref string) {
+	t.Helper()
+
+	got, want := loadTemplateDefaults(t, ref)
+
+	// Layer 1: parsed defaults match the golden per-component.
+	assertComponentsMatch(t, "Layer1 parse", got, want)
+
+	// Layer 2: defaults survive a JSON round-trip through entity.JSONMap
+	// (the DB storage format — GORM deserializes into map[string]any, then
+	// runPipelineWithDSL reads it back as map[string]interface{}).
+	raw, err := json.Marshal(got)
+	if err != nil {
+		t.Fatalf("marshal %q: %v", ref, err)
+	}
+	var stored entity.JSONMap
+	if err := json.Unmarshal(raw, &stored); err != nil {
+		t.Fatalf("unmarshal into JSONMap %q: %v", ref, err)
+	}
+	var roundTripped map[string]map[string]any
+	if err := json.Unmarshal(raw, &roundTripped); err != nil {
+		t.Fatalf("unmarshal back %q: %v", ref, err)
+	}
+	assertComponentsMatch(t, "Layer2 round-trip", roundTripped, got)
+}
+
+// loadTemplateDefaults loads the builtin DSL for ref, parses the default
+// component params, and returns both the parsed result and the hardcoded
+// golden expectation.
+func loadTemplateDefaults(t *testing.T, ref string) (got, want map[string]map[string]any) {
+	t.Helper()
+
+	golden, ok := builtinComponentParamsGolden[ref]
+	if !ok {
+		t.Fatalf("missing golden entry for template %q", ref)
+	}
+	if err := json.Unmarshal([]byte(golden), &want); err != nil {
+		t.Fatalf("unmarshal golden %q: %v", ref, err)
+	}
+
+	dsl, err := pipelinepkg.LoadBuiltinDSL(ref)
+	if err != nil {
+		t.Fatalf("load builtin DSL %q: %v", ref, err)
+	}
+	got, err = pipelinepkg.ComponentParamsDefaults([]byte(dsl))
+	if err != nil {
+		t.Fatalf("ComponentParamsDefaults %q: %v", ref, err)
+	}
+	return
+}
+
+func assertComponentsMatch(t *testing.T, label string, got, want map[string]map[string]any) {
+	t.Helper()
+	for cpnID := range got {
+		if _, ok := want[cpnID]; !ok {
+			t.Errorf("[%s] unexpected component %q", label, cpnID)
+		}
+	}
+	for cpnID, wantParams := range want {
+		t.Run(cpnID, func(t *testing.T) {
+			gotParams, ok := got[cpnID]
+			if !ok {
+				t.Errorf("[%s] missing component %q", label, cpnID)
+				return
+			}
+			if !reflect.DeepEqual(gotParams, wantParams) {
+				t.Errorf("[%s] component %q params mismatch\n got=%#v\nwant=%#v", label, cpnID, gotParams, wantParams)
+			}
+		})
+	}
+}
+
+// TestBuildComponentParams_GoldenCoversAllTemplates keeps the hardcoded golden
+// table in lockstep with the built-in template registry. A renamed or added
+// template must get a matching golden entry, and a removed template must drop
+// its entry, otherwise this test fails. ("naive" is an alias for "general" and
+// is intentionally not a separate file/template.)
+func TestBuildComponentParams_GoldenCoversAllTemplates(t *testing.T) {
+	reg, err := pipelinepkg.DefaultRegistry()
+	if err != nil {
+		t.Fatalf("builtin registry: %v", err)
+	}
+	refs := reg.Refs()
+	if len(refs) == 0 {
+		t.Fatal("builtin registry returned no templates")
+	}
+	if len(refs) != len(builtinComponentParamsGolden) {
+		t.Fatalf("golden table has %d entries but registry has %d templates: %v",
+			len(builtinComponentParamsGolden), len(refs), refs)
+	}
+	refSet := make(map[string]struct{}, len(refs))
+	for _, r := range refs {
+		refSet[r] = struct{}{}
+	}
+	for g := range builtinComponentParamsGolden {
+		if _, ok := refSet[g]; !ok {
+			t.Errorf("golden entry %q has no matching builtin template", g)
+		}
+	}
+}
diff --git a/internal/ingestion/task/pipeline_executor_test.go b/internal/ingestion/task/pipeline_executor_test.go
index 1b82656809..55c6108911 100644
--- a/internal/ingestion/task/pipeline_executor_test.go
+++ b/internal/ingestion/task/pipeline_executor_test.go
@@ -261,7 +261,7 @@ func TestRecordPipelineLog_ValidJSONParsed(t *testing.T) {
 
 func TestRunPipeline_NilOutput(t *testing.T) {
 	svc := mustNewPipelineExecutor(t, makeTaskCtx(), "flow-1", 0)
-	_, err := svc.processOutput(context.Background(), nil)
+	_, err := svc.processOutput(context.Background(), nil, time.Now())
 	if err != nil {
 		t.Errorf("expected nil error for nil output, got %v", err)
 	}
@@ -271,7 +271,7 @@ func TestRunPipeline_EmptyOutput(t *testing.T) {
 	svc := mustNewPipelineExecutor(t, makeTaskCtx(), "flow-1", 0).WithLogCreateFunc(
 		func(log *entity.PipelineOperationLog) error { return nil },
 	)
-	_, err := svc.processOutput(context.Background(), map[string]any{})
+	_, err := svc.processOutput(context.Background(), map[string]any{}, time.Now())
 	if err != nil {
 		t.Errorf("expected nil error for empty output, got %v", err)
 	}
@@ -281,7 +281,7 @@ func TestRunPipeline_NormalizedEmpty(t *testing.T) {
 	svc := mustNewPipelineExecutor(t, makeTaskCtx(), "flow-1", 0).WithLogCreateFunc(
 		func(log *entity.PipelineOperationLog) error { return nil },
 	)
-	_, err := svc.processOutput(context.Background(), map[string]any{"markdown": ""})
+	_, err := svc.processOutput(context.Background(), map[string]any{"markdown": ""}, time.Now())
 	if err != nil {
 		t.Errorf("expected nil error for empty normalized output, got %v", err)
 	}
@@ -299,7 +299,7 @@ func TestRunPipeline_FullFlow(t *testing.T) {
 			{"text": "world"},
 		},
 	}
-	_, err := svc.processOutput(context.Background(), output)
+	_, err := svc.processOutput(context.Background(), output, time.Now())
 	if err != nil {
 		t.Errorf("unexpected error: %v", err)
 	}
@@ -317,7 +317,7 @@ func TestRunPipeline_AlreadyHasVectors(t *testing.T) {
 			{"text": "hello", "q_768_vec": []float64{0.1, 0.2}},
 		},
 	}
-	_, err := svc.processOutput(context.Background(), output)
+	_, err := svc.processOutput(context.Background(), output, time.Now())
 	if err != nil {
 		t.Errorf("unexpected error: %v", err)
 	}
@@ -330,7 +330,7 @@ func TestRunPipeline_ContextCanceled(t *testing.T) {
 
 	_, err := svc.processOutput(ctx, map[string]any{
 		"chunks": []map[string]any{{"text": "hello"}},
-	})
+	}, time.Now())
 	if err == nil {
 		t.Error("expected context canceled error")
 	}
diff --git a/internal/ingestion/task/pipeline_real_integration_test.go b/internal/ingestion/task/pipeline_real_integration_test.go
index 9ba4890442..e7907818e8 100644
--- a/internal/ingestion/task/pipeline_real_integration_test.go
+++ b/internal/ingestion/task/pipeline_real_integration_test.go
@@ -1,6 +1,3 @@
-//go:build integration
-// +build integration
-
 package task
 
 import (
@@ -386,7 +383,7 @@ func TestRunPipeline_RealPipelineOutput_ProducesIndexFields(t *testing.T) {
 			return nil, nil
 		})
 
-	if _, err := svc.processOutput(context.Background(), pipelineOut); err != nil {
+	if _, err := svc.processOutput(context.Background(), pipelineOut, time.Now()); err != nil {
 		t.Fatalf("RunPipeline: %v", err)
 	}
 
@@ -430,7 +427,7 @@ func taskRepoRoot(t *testing.T) string {
 
 func mustLoadTaskRealIntegrationConfig(t *testing.T) *server.Config {
 	t.Helper()
-	if err := common.Init("info", common.FileOutput{}); err != nil {
+	if err := common.Init("info", common.FileOutput{}, ""); err != nil {
 		t.Fatalf("init common logger: %v", err)
 	}
 	server.SetLogger(zap.NewNop())
diff --git a/internal/parser/parser/pdf_parser_cgo.go b/internal/parser/parser/pdf_parser_cgo.go
index da775e2d00..3140340262 100644
--- a/internal/parser/parser/pdf_parser_cgo.go
+++ b/internal/parser/parser/pdf_parser_cgo.go
@@ -6,6 +6,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"ragflow/internal/common"
 
 	deepdocpdf "ragflow/internal/deepdoc/parser/pdf"
 	deepdoctype "ragflow/internal/deepdoc/parser/type"
@@ -15,6 +16,7 @@ func (p *PDFParser) ParseWithResult(filename string, data []byte) ParseResult {
 	if err := p.validateParseMethod(); err != nil {
 		return ParseResult{Err: err}
 	}
+	common.Info(fmt.Sprintf("------------file: %s, parse_method: %s", filename, p.ParseMethod))
 	switch normalizePDFParseMethod(p.ParseMethod) {
 	case "plain_text":
 		return parsePDFWithPlainText(filename, data, p)
diff --git a/internal/router/router.go b/internal/router/router.go
index 0f81a9d018..dbe8548335 100644
--- a/internal/router/router.go
+++ b/internal/router/router.go
@@ -167,6 +167,7 @@ func (r *Router) Setup(engine *gin.Engine) {
 		// Query: ?type=builtin returns built-in templates (default).
 		if r.pipelineHandler != nil {
 			apiNoAuth.GET("/pipelines", r.pipelineHandler.ListPipelines)
+			apiNoAuth.GET("/pipelines/:id", r.pipelineHandler.GetPipeline)
 		}
 
 		// searchbots
@@ -318,7 +319,6 @@ func (r *Router) Setup(engine *gin.Engine) {
 			// Document routes
 			documents := v1.Group("/documents")
 			{
-				documents.POST("", r.documentHandler.CreateDocument)
 				documents.POST("/upload", r.documentHandler.UploadInfo)
 				documents.GET("", r.documentHandler.ListDocuments)
 				documents.GET("/artifact/:filename", r.documentHandler.GetDocumentArtifact)
diff --git a/internal/service/chunk/chunk.go b/internal/service/chunk/chunk.go
index dcabe1517b..63d14d444b 100644
--- a/internal/service/chunk/chunk.go
+++ b/internal/service/chunk/chunk.go
@@ -631,7 +631,10 @@ func (s *ChunkService) StopParsing(userID, datasetID string, req service.StopPar
 			return nil, common.CodeDataError, fmt.Errorf("You don't own the document %s", docID)
 		}
 
-		task, _ := dao.NewIngestionTaskDAO().GetByDocumentID(docID)
+		task, err := dao.NewIngestionTaskDAO().GetByDocumentID(docID)
+		if err != nil {
+			return nil, common.CodeServerError, fmt.Errorf("get ingestion task for %s: %w", docID, err)
+		}
 		if task == nil || task.Status == common.COMPLETED ||
 			task.Status == common.STOPPED || task.Status == common.FAILED {
 			return &service.StopParsingResponse{
diff --git a/internal/service/dataset.go b/internal/service/dataset.go
index 4ef5bdf51f..5992ca17e6 100644
--- a/internal/service/dataset.go
+++ b/internal/service/dataset.go
@@ -1548,21 +1548,6 @@ func (d *DatasetService) SearchDatasets(req *SearchDatasetsRequest, userID strin
 	}, nil
 }
 
-// AutoMetadataField mirrors the REST dataset auto metadata field schema.
-type AutoMetadataField struct {
-	Name           string      `json:"name"`
-	Type           string      `json:"type"`
-	Description    *string     `json:"description,omitempty"`
-	Examples       interface{} `json:"examples,omitempty"`
-	RestrictValues bool        `json:"restrict_values,omitempty"`
-}
-
-// AutoMetadataConfig mirrors the REST dataset auto metadata schema.
-type AutoMetadataConfig struct {
-	Enabled *bool               `json:"enabled,omitempty"`
-	Fields  []AutoMetadataField `json:"fields,omitempty"`
-}
-
 // MetadataConfigField mirrors one field in the dataset metadata config API.
 type MetadataConfigField struct {
 	Key         string   `json:"key"`
@@ -1579,17 +1564,14 @@ type MetadataConfigRequest struct {
 
 // CreateDatasetRequest represents the request for creating a dataset.
 type CreateDatasetRequest struct {
-	Name               string                 `json:"name" binding:"required"`
-	Avatar             *string                `json:"avatar,omitempty"`
-	Description        *string                `json:"description,omitempty"`
-	EmbeddingModel     *string                `json:"embedding_model,omitempty"`
-	Permission         *string                `json:"permission,omitempty"`
-	ChunkMethod        *string                `json:"chunk_method,omitempty"`
-	ParseType          *int                   `json:"parse_type,omitempty"`
-	PipelineID         *string                `json:"pipeline_id,omitempty"`
-	ParserConfig       map[string]interface{} `json:"parser_config,omitempty"`
-	AutoMetadataConfig *AutoMetadataConfig    `json:"auto_metadata_config,omitempty"`
-	Ext                map[string]interface{} `json:"ext,omitempty"`
+	Name           string  `json:"name" binding:"required"`
+	EmbeddingModel *string `json:"embedding_model,omitempty"`
+	Permission     *string `json:"permission,omitempty"`
+	ParserID       *string `json:"parser_id,omitempty"`
+	PipelineID     *string `json:"pipeline_id,omitempty"`
+	// ParseType indicates pipeline selection mode: 1 = BuiltIn (parser_id),
+	// 2 = Pipeline (pipeline_id). nil means unspecified (backward compat).
+	ParseType *int `json:"parse_type,omitempty"`
 }
 
 // ListDatasets lists datasets with pagination and filtering.
@@ -1693,42 +1675,42 @@ func (d *DatasetService) CreateDataset(req *CreateDatasetRequest, tenantID strin
 		return nil, common.CodeDataError, errors.New("Tenant not found.")
 	}
 
+	// parse_type explicitly signals the pipeline mode (1 = BuiltIn, 2 = Pipeline).
+	// When absent (nil), infer intent from which field is present for backward compat.
+	isPipelineMode := req.ParseType != nil && *req.ParseType == 2
+	isBuiltinMode := req.ParseType != nil && *req.ParseType == 1
+
+	if isBuiltinMode && req.PipelineID != nil {
+		// BuiltIn mode: discard pipeline_id so only parser_id matters.
+		req.PipelineID = nil
+	}
+	if isPipelineMode && req.ParserID != nil {
+		// Pipeline mode: ignore parser_id.
+		req.ParserID = nil
+	}
+
+	if req.ParseType == nil && req.ParserID != nil && req.PipelineID != nil {
+		return nil, common.CodeDataError, errors.New("parser_id and pipeline_id are mutually exclusive")
+	}
+
 	parserID := ""
 	permission := "me"
 	embeddingModel := ""
-	parserConfig := req.ParserConfig
 	pipelineID := req.PipelineID
-	description := req.Description
-	avatar := req.Avatar
-	var language *string
 
-	if req.Description != nil && len(*req.Description) > 65535 {
-		return nil, common.CodeDataError, errors.New("String should have at most 65535 characters")
-	}
-	if req.Avatar != nil {
-		if len(*req.Avatar) > 65535 {
-			return nil, common.CodeDataError, errors.New("String should have at most 65535 characters")
-		}
-		if err := validateDatasetAvatar(*req.Avatar); err != nil {
-			return nil, common.CodeDataError, err
-		}
-	}
 	if req.Permission != nil {
 		permission = strings.TrimSpace(*req.Permission)
 		if permission != "me" && permission != "team" {
 			return nil, common.CodeDataError, errors.New("Input should be 'me' or 'team'")
 		}
 	}
-	if req.ChunkMethod != nil {
-		parserID = strings.TrimSpace(*req.ChunkMethod)
+	if req.ParserID != nil {
+		parserID = strings.TrimSpace(*req.ParserID)
 		if err := validateParserID(parserID); err != nil {
 			return nil, common.CodeDataError, err
 		}
 		pipelineID = nil
 	}
-	if req.ParseType != nil && (*req.ParseType < 0 || *req.ParseType > 64) {
-		return nil, common.CodeDataError, fmt.Errorf("Input should be between 0 and 64")
-	}
 	if req.PipelineID != nil {
 		normalizedPipelineID, err := normalizeDatasetPipelineID(*req.PipelineID)
 		if err != nil {
@@ -1742,151 +1724,27 @@ func (d *DatasetService) CreateDataset(req *CreateDatasetRequest, tenantID strin
 			return nil, common.CodeDataError, err
 		}
 	}
-	if err := validateDatasetParserConfigSize(parserConfig); err != nil {
-		return nil, common.CodeDataError, err
-	}
 
-	// ext mirrors the Python REST implementation and overrides known top-level fields.
-	for key, value := range req.Ext {
-		switch key {
-		case "name":
-			nameValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Dataset name must be string.")
-			}
-			nameValue = strings.TrimSpace(nameValue)
-			if nameValue == "" {
-				return nil, common.CodeDataError, errors.New("Dataset name can't be empty.")
-			}
-			if len(nameValue) > entity.DatasetNameLimit {
-				return nil, common.CodeDataError, fmt.Errorf("Dataset name length is %d which is large than %d", len(nameValue), entity.DatasetNameLimit)
-			}
-			name = nameValue
-		case "description":
-			descriptionValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Description must be string.")
-			}
-			if len(descriptionValue) > 65535 {
-				return nil, common.CodeDataError, errors.New("String should have at most 65535 characters")
-			}
-			description = &descriptionValue
-		case "avatar":
-			avatarValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Avatar must be string.")
-			}
-			if len(avatarValue) > 65535 {
-				return nil, common.CodeDataError, errors.New("String should have at most 65535 characters")
-			}
-			if err := validateDatasetAvatar(avatarValue); err != nil {
-				return nil, common.CodeDataError, err
-			}
-			avatar = &avatarValue
-		case "language":
-			languageValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Language must be string.")
-			}
-			languageValue = strings.TrimSpace(languageValue)
-			language = &languageValue
-		case "permission":
-			permissionValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Permission must be string.")
-			}
-			permissionValue = strings.TrimSpace(permissionValue)
-			if permissionValue != "me" && permissionValue != "team" {
-				return nil, common.CodeDataError, errors.New("Input should be 'me' or 'team'")
-			}
-			permission = permissionValue
-		case "embedding_model", "embd_id":
-			embeddingModelValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Embedding model identifier must follow <model_name>@<provider> format")
-			}
-			embeddingModelValue = strings.TrimSpace(embeddingModelValue)
-			if err := validateDatasetEmbeddingModel(embeddingModelValue); err != nil {
-				return nil, common.CodeDataError, err
-			}
-			embeddingModel = embeddingModelValue
-		case "chunk_method", "parser_id":
-			parserIDValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, parserIDError()
-			}
-			parserIDValue = strings.TrimSpace(parserIDValue)
-			if err := validateParserID(parserIDValue); err != nil {
-				return nil, common.CodeDataError, err
-			}
-			parserID = parserIDValue
-			pipelineID = nil
-		case "pipeline_id":
-			pipelineIDValue, ok := value.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("pipeline_id must be 32 hex characters")
-			}
-			normalizedPipelineID, err := normalizeDatasetPipelineID(pipelineIDValue)
-			if err != nil {
-				return nil, common.CodeDataError, err
-			}
-			pipelineID = normalizedPipelineID
-		case "parser_config":
-			parserConfigValue, ok := value.(map[string]interface{})
-			if !ok {
-				return nil, common.CodeDataError, errors.New("parser_config must be valid JSON")
-			}
-			if err := validateDatasetParserConfigSize(parserConfigValue); err != nil {
-				return nil, common.CodeDataError, err
-			}
-			parserConfig = parserConfigValue
+	// Reject references to a custom canvas the caller does not own or share.
+	// The handler passes the caller's user id in tenantID for CreateDataset.
+	if pipelineID != nil && strings.TrimSpace(*pipelineID) != "" {
+		if ok, err := canvasAccessibleForUser(tenantID, strings.TrimSpace(*pipelineID)); err != nil {
+			return nil, common.CodeServerError, err
+		} else if !ok {
+			return nil, common.CodeDataError, errors.New("canvas is not accessible")
 		}
 	}
 
-	// parser_id wins when it is present; otherwise parse_type and pipeline_id must arrive together.
-	if parserID == "" {
-		if req.ParseType == nil && pipelineID == nil {
-			parserID = "naive"
-		} else if req.ParseType == nil || pipelineID == nil {
-			missingFields := make([]string, 0, 2)
-			if req.ParseType == nil {
-				missingFields = append(missingFields, "parse_type")
-			}
-			if pipelineID == nil {
-				missingFields = append(missingFields, "pipeline_id")
-			}
-			return nil, common.CodeDataError, fmt.Errorf("parser_id omitted -> required fields missing: %s", strings.Join(missingFields, ", "))
-		}
+	// Resolve component params defaults from the DSL template. parser_config
+	// stores component params directly: {cpnID: {param: value}}.
+	parserConfig, cpErr := resolveComponentParamsDefaults(parserID, pipelineID)
+	if cpErr != nil {
+		common.Warn("failed to resolve component params defaults for dataset",
+			zap.String("parserID", parserID), zap.Error(cpErr))
+		parserConfig = entity.JSONMap{}
 	}
 
-	if req.AutoMetadataConfig != nil {
-		parserConfig = applyAutoMetadataConfig(parserConfig, req.AutoMetadataConfig)
-	}
-
-	// Build parser_config. When pipeline_id is set, seed from pipeline DSL defaults.
-	var parserConfigMap map[string]interface{}
-	if pipelineID != nil {
-		if parserConfig == nil {
-			parserConfig = make(map[string]interface{})
-		}
-		canvas, err := dao.NewUserCanvasDAO().GetByID(*pipelineID)
-		if err != nil || canvas == nil {
-			return nil, common.CodeDataError, fmt.Errorf("pipeline %s not found", *pipelineID)
-		}
-		pipelineDefaults := common.ExtractPipelineDefaults(canvas.DSL)
-		if pipelineDefaults != nil {
-			parserConfigMap = common.DeepMergeMaps(pipelineDefaults, parserConfig)
-		} else {
-			parserConfigMap = common.DeepMergeMaps(nil, parserConfig)
-		}
-		common.InjectExtractorLLMID(parserConfigMap, tenant.LLMID)
-	} else {
-		parserConfigMap = common.GetParserConfig(parserID, parserConfig)
-		parserConfigMap["llm_id"] = tenant.LLMID
-	}
-	if err := validateDatasetParserConfigSize(parserConfigMap); err != nil {
-		return nil, common.CodeDataError, err
-	}
+	var parserConfigMap map[string]interface{} = parserConfig
 
 	embdID := tenant.EmbdID
 	tenantEmbdID := ptrStringValue(tenant.TenantEmbdID)
@@ -1908,7 +1766,6 @@ func (d *DatasetService) CreateDataset(req *CreateDatasetRequest, tenantID strin
 	}
 
 	kbID := utility.GenerateToken()
-
 	status := string(entity.StatusValid)
 	// Deduplicate name within tenant
 	duplicateName, err := common.DuplicateName(func(n, tid string) bool {
@@ -1933,16 +1790,6 @@ func (d *DatasetService) CreateDataset(req *CreateDatasetRequest, tenantID strin
 		Status:       &status,
 	}
 
-	if description != nil {
-		kb.Description = description
-	}
-	if avatar != nil {
-		kb.Avatar = avatar
-	}
-	if language != nil {
-		kb.Language = language
-	}
-
 	if err = d.kbDAO.Create(kb); err != nil {
 		return nil, common.CodeServerError, errors.New("Failed to save dataset")
 	}
@@ -2080,21 +1927,21 @@ type DatasetConnectorRequest struct {
 }
 
 type UpdateDatasetRequest struct {
-	Name               *string                    `json:"name,omitempty"`
-	Avatar             *string                    `json:"avatar,omitempty"`
-	Description        *string                    `json:"description,omitempty"`
-	Language           *string                    `json:"language,omitempty"`
-	Connectors         *[]DatasetConnectorRequest `json:"connectors,omitempty"`
-	EmbdID             *string                    `json:"embd_id,omitempty"`
-	EmbeddingModel     *string                    `json:"embedding_model,omitempty"`
-	Permission         *string                    `json:"permission,omitempty"`
-	ParserID           *string                    `json:"parser_id,omitempty"`
-	ChunkMethod        *string                    `json:"chunk_method,omitempty"`
-	Pagerank           *int64                     `json:"pagerank,omitempty"`
-	ParserConfig       map[string]interface{}     `json:"parser_config,omitempty"`
-	PipelineID         *string                    `json:"pipeline_id,omitempty"`
-	AutoMetadataConfig *AutoMetadataConfig        `json:"auto_metadata_config,omitempty"`
-	Ext                map[string]interface{}     `json:"ext,omitempty"`
+	Name           *string                    `json:"name,omitempty"`
+	Avatar         *string                    `json:"avatar,omitempty"`
+	Description    *string                    `json:"description,omitempty"`
+	Language       *string                    `json:"language,omitempty"`
+	Connectors     *[]DatasetConnectorRequest `json:"connectors,omitempty"`
+	EmbdID         *string                    `json:"embd_id,omitempty"`
+	EmbeddingModel *string                    `json:"embedding_model,omitempty"`
+	Permission     *string                    `json:"permission,omitempty"`
+	ParserID       *string                    `json:"parser_id,omitempty"`
+	Pagerank       *int64                     `json:"pagerank,omitempty"`
+	ParserConfig   map[string]interface{}     `json:"parser_config,omitempty"`
+	PipelineID     *string                    `json:"pipeline_id,omitempty"`
+	// ParseType indicates pipeline selection mode: 1 = BuiltIn (parser_id),
+	// 2 = Pipeline (pipeline_id). nil means unspecified (backward compat).
+	ParseType *int `json:"parse_type,omitempty"`
 }
 
 // UpdateDataset Update a dataset
@@ -2118,7 +1965,6 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 	}
 
 	updates := make(map[string]interface{})
-	extUpdates := normalizeDatasetUpdateExt(req.Ext)
 
 	if req.Name != nil {
 		name := strings.TrimSpace(*req.Name)
@@ -2159,6 +2005,19 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 		}
 		updates["permission"] = permission
 	}
+
+	// parse_type explicitly signals the pipeline mode (1 = BuiltIn, 2 = Pipeline).
+	// When absent (nil), existing mutual-exclusivity behavior is preserved.
+	isPipelineMode := req.ParseType != nil && *req.ParseType == 2
+	isBuiltinMode := req.ParseType != nil && *req.ParseType == 1
+
+	if isBuiltinMode && req.PipelineID != nil {
+		req.PipelineID = nil
+	}
+	if isPipelineMode && req.ParserID != nil {
+		req.ParserID = nil
+	}
+
 	if req.PipelineID != nil {
 		pipelineID, err := normalizeDatasetPipelineID(*req.PipelineID)
 		if err != nil {
@@ -2169,52 +2028,23 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 		}
 	}
 
-	for key, value := range extUpdates {
-		if _, exists := updates[key]; !exists {
-			updates[key] = value
-		}
-	}
-
 	parserID, parserIDProvided, err := datasetUpdateParserID(req)
 	if err != nil {
 		return nil, common.CodeDataError, err
 	}
-	if !parserIDProvided {
-		if extParserID, ok := updates["parser_id"]; ok {
-			parserIDValue, ok := extParserID.(string)
-			if !ok {
-				return nil, common.CodeDataError, parserIDError()
-			}
-			parserID = strings.TrimSpace(parserIDValue)
-			if err := validateParserID(parserID); err != nil {
-				return nil, common.CodeDataError, err
-			}
-			parserIDProvided = true
-		}
-	}
 	if parserIDProvided {
 		updates["parser_id"] = parserID
 	}
 
+	// When parse_type is absent, parser_id and pipeline_id are mutually exclusive.
+	if req.ParseType == nil && parserIDProvided && req.PipelineID != nil {
+		return nil, common.CodeDataError, errors.New("parser_id and pipeline_id are mutually exclusive")
+	}
+
 	embdID, embdIDProvided, err := datasetUpdateEmbeddingID(req)
 	if err != nil {
 		return nil, common.CodeDataError, err
 	}
-	if !embdIDProvided {
-		if extEmbdID, ok := updates["embd_id"]; ok {
-			embdIDValue, ok := extEmbdID.(string)
-			if !ok {
-				return nil, common.CodeDataError, errors.New("Embedding model identifier must follow <model_name>@<provider> format")
-			}
-			embdID = strings.TrimSpace(embdIDValue)
-			if embdID != "" {
-				if err := validateDatasetEmbeddingModel(embdID); err != nil {
-					return nil, common.CodeDataError, err
-				}
-			}
-			embdIDProvided = true
-		}
-	}
 	if embdIDProvided {
 		tenantEmbdID := ptrStringValue(kb.TenantEmbdID)
 		if embdID == "" {
@@ -2236,19 +2066,37 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 		updates["tenant_embd_id"] = stringPtrIfNotEmpty(tenantEmbdID)
 	}
 
-	if req.AutoMetadataConfig != nil {
-		req.ParserConfig = applyAutoMetadataConfig(req.ParserConfig, req.AutoMetadataConfig)
-	}
 	if req.ParserConfig != nil {
 		if err := validateDatasetParserConfigSize(req.ParserConfig); err != nil {
 			return nil, common.CodeDataError, err
 		}
 		if len(req.ParserConfig) > 0 {
-			merged := common.DeepMergeMaps(kb.ParserConfig, req.ParserConfig)
-			if err := validateDatasetParserConfigSize(merged); err != nil {
-				return nil, common.CodeDataError, err
+			// Resolve the effective pipeline and load its DSL schema.
+			effectiveParserID := kb.ParserID
+			if parserIDProvided {
+				effectiveParserID = parserID
+			}
+			effectivePipelineID := kb.PipelineID
+			if req.PipelineID != nil {
+				if normalized, err := normalizeDatasetPipelineID(*req.PipelineID); err == nil {
+					effectivePipelineID = normalized
+				}
+			} else if parserIDProvided && kb.PipelineID != nil {
+				effectivePipelineID = nil
+			}
+
+			isCanvas := effectivePipelineID != nil && strings.TrimSpace(*effectivePipelineID) != ""
+			dslJSON, dslErr := loadPipelineDSL(isCanvas, effectiveParserID, effectivePipelineID)
+			if dslErr != nil {
+				common.Warn("failed to load pipeline DSL for building parser_config",
+					zap.String("parserID", effectiveParserID), zap.Error(dslErr))
+			}
+			if dslJSON != nil {
+				// Start from DSL defaults, overlay the cleaned incoming overrides.
+				// Unknown cpnIDs, unknown params, and legacy flat fields are
+				// dropped. Full replace — no merge with the stored config.
+				updates["parser_config"] = buildParserConfig(dslJSON, map[string]interface{}(req.ParserConfig))
 			}
-			updates["parser_config"] = entity.JSONMap(merged)
 		}
 	}
 
@@ -2273,12 +2121,17 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 
 	if parserIDProvided && parserID != kb.ParserID {
 		if _, ok := updates["parser_config"]; !ok {
-			updates["parser_config"] = entity.JSONMap(common.GetParserConfig(parserID, nil))
+			if resolved, cpErr := resolveComponentParamsDefaults(parserID, nil); cpErr != nil {
+				common.Warn("failed to resolve component params defaults on parser_id switch",
+					zap.String("parserID", parserID), zap.Error(cpErr))
+			} else if resolved != nil {
+				updates["parser_config"] = resolved
+			}
 		}
 	}
 	if kb.PipelineID != nil && parserIDProvided {
 		if _, ok := updates["pipeline_id"]; !ok {
-			updates["pipeline_id"] = ""
+			updates["pipeline_id"] = nil // clear to NULL, not empty string
 		}
 	}
 
@@ -2290,28 +2143,17 @@ func (d *DatasetService) UpdateDataset(datasetID, tenantID string, req UpdateDat
 		if parserIDProvided {
 			cfgParserID = parserID
 		}
-		parserConfigMap := common.GetParserConfig(cfgParserID, nil)
-		if upPipelineID, ok := updates["pipeline_id"].(string); ok && upPipelineID != "" {
-			canvas, err := dao.NewUserCanvasDAO().GetByID(upPipelineID)
-			if err != nil || canvas == nil {
-				return nil, common.CodeDataError, fmt.Errorf("pipeline %s not found", upPipelineID)
-			}
-			pipelineDefaults := common.ExtractPipelineDefaults(canvas.DSL)
-			if pipelineDefaults != nil {
-				parserConfigMap = common.DeepMergeMaps(pipelineDefaults, parserConfigMap)
-			}
+		cfgPipelineID, _ := updates["pipeline_id"].(string)
+		var cpPipelineID *string
+		if cfgPipelineID != "" {
+			cpPipelineID = &cfgPipelineID
 		}
-		if req.ParserConfig != nil && len(req.ParserConfig) > 0 {
-			parserConfigMap = common.DeepMergeMaps(parserConfigMap, req.ParserConfig)
+		if cpDefaults, cpErr := resolveComponentParamsDefaults(cfgParserID, cpPipelineID); cpErr != nil {
+			common.Warn("failed to resolve component params defaults on pipeline change",
+				zap.String("parserID", cfgParserID), zap.Error(cpErr))
+		} else if cpDefaults != nil {
+			updates["parser_config"] = cpDefaults
 		}
-		tenant, err := d.tenantDAO.GetByID(tenantID)
-		if err == nil && tenant != nil {
-			common.InjectExtractorLLMID(parserConfigMap, tenant.LLMID)
-		}
-		if err := validateDatasetParserConfigSize(parserConfigMap); err != nil {
-			return nil, common.CodeDataError, err
-		}
-		updates["parser_config"] = entity.JSONMap(parserConfigMap)
 	}
 
 	if nameValue, ok := updates["name"].(string); ok && strings.ToLower(nameValue) != strings.ToLower(kb.Name) {
@@ -2379,10 +2221,6 @@ func datasetUpdateParserID(req UpdateDatasetRequest) (string, bool, error) {
 		parserID = strings.TrimSpace(*req.ParserID)
 		provided = true
 	}
-	if req.ChunkMethod != nil {
-		parserID = strings.TrimSpace(*req.ChunkMethod)
-		provided = true
-	}
 	if !provided {
 		return "", false, nil
 	}
@@ -3114,30 +2952,6 @@ func (d *DatasetService) verifyEmbeddingAvailability(embdID string, tenantID str
 	return true, ""
 }
 
-func applyAutoMetadataConfig(parserConfig map[string]interface{}, config *AutoMetadataConfig) map[string]interface{} {
-	if parserConfig == nil {
-		parserConfig = make(map[string]interface{})
-	}
-
-	fields := make([]map[string]interface{}, 0, len(config.Fields))
-	for _, field := range config.Fields {
-		fields = append(fields, map[string]interface{}{
-			"name":            field.Name,
-			"type":            field.Type,
-			"description":     field.Description,
-			"examples":        field.Examples,
-			"restrict_values": field.RestrictValues,
-		})
-	}
-	parserConfig["metadata"] = fields
-	enableMetadata := true
-	if config.Enabled != nil {
-		enableMetadata = *config.Enabled
-	}
-	parserConfig["enable_metadata"] = enableMetadata
-	return parserConfig
-}
-
 func parserConfigValueOrEmptyList(parserConfig map[string]interface{}, key string) interface{} {
 	if parserConfig == nil {
 		return []interface{}{}
@@ -3191,7 +3005,7 @@ func datasetListItemToMap(kb *entity.KnowledgebaseListItem) map[string]interface
 		"document_count":  kb.DocNum,
 		"token_num":       kb.TokenNum,
 		"chunk_count":     kb.ChunkNum,
-		"chunk_method":    kb.ParserID,
+		"parser_id":       kb.ParserID,
 		"embedding_model": kb.EmbdID,
 		"nickname":        kb.Nickname,
 	}
@@ -3228,7 +3042,7 @@ func datasetToMap(kb *entity.Knowledgebase) map[string]interface{} {
 		"chunk_count":              kb.ChunkNum,
 		"similarity_threshold":     kb.SimilarityThreshold,
 		"vector_similarity_weight": kb.VectorSimilarityWeight,
-		"chunk_method":             kb.ParserID,
+		"parser_id":                kb.ParserID,
 		"parser_config":            kb.ParserConfig,
 		"pagerank":                 kb.Pagerank,
 		"create_time":              kb.CreateTime,
@@ -3329,3 +3143,48 @@ func (d *DatasetService) RenameTag(datasetID, userID, fromTag, toTag string) (ma
 func (d *DatasetService) GetFieldMap(ids []string) (map[string]interface{}, error) {
 	return d.kbDAO.GetFieldMap(ids)
 }
+
+// resolveComponentParamsDefaults loads the DSL for the target pipeline and
+// returns the component params defaults as an entity.JSONMap {cpnID: {param: value}}.
+// For builtin templates the DSL is loaded from the embedded registry; for custom
+// canvas pipelines it is loaded from the canvas row in the database.
+func resolveComponentParamsDefaults(parserID string, pipelineID *string) (entity.JSONMap, error) {
+	isCanvas := pipelineID != nil && strings.TrimSpace(*pipelineID) != ""
+	var cp map[string]map[string]any
+	var err error
+	if isCanvas {
+		dslJSON, lerr := loadCanvasDSLJSON(strings.TrimSpace(*pipelineID))
+		if lerr != nil {
+			return nil, fmt.Errorf("load canvas DSL: %w", lerr)
+		}
+		cp, err = pipelinepkg.ComponentParamsDefaults(dslJSON)
+	} else {
+		registry, regErr := pipelinepkg.DefaultRegistry()
+		if regErr != nil {
+			return nil, fmt.Errorf("builtin registry: %w", regErr)
+		}
+		if !registry.IsValid(parserID) {
+			return nil, fmt.Errorf("unknown builtin parser_id: %q", parserID)
+		}
+		dslStr, dslErr := pipelinepkg.LoadBuiltinDSL(parserID)
+		if dslErr != nil {
+			return nil, fmt.Errorf("load builtin DSL: %w", dslErr)
+		}
+		cp, err = pipelinepkg.ComponentParamsDefaults([]byte(dslStr))
+	}
+	if err != nil {
+		return nil, err
+	}
+	out := make(entity.JSONMap, len(cp))
+	for k, v := range cp {
+		out[k] = v
+	}
+	return out, nil
+}
+
+// canvasAccessibleForUser reports whether the given canvas is owned by or
+// team-shared with the caller.
+func canvasAccessibleForUser(userID, canvasID string) (bool, error) {
+	tenantIDs, _ := dao.NewUserTenantDAO().GetTenantIDsByUserID(userID)
+	return dao.NewUserCanvasDAO().Accessible(canvasID, userID, tenantIDs), nil
+}
diff --git a/internal/service/dataset_create_test.go b/internal/service/dataset_create_test.go
new file mode 100644
index 0000000000..7c6297ca82
--- /dev/null
+++ b/internal/service/dataset_create_test.go
@@ -0,0 +1,207 @@
+//
+//  Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+//
+//  Licensed under the Apache License, Version 2.0 (the "License");
+//  you may not use this file except in compliance with the License.
+//  You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+//  Unless required by applicable law or agreed to in writing, software
+//  distributed under the License is distributed on an "AS IS" BASIS,
+//  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+//  See the License for the specific language governing permissions and
+//  limitations under the License.
+//
+
+package service
+
+import (
+	"strings"
+	"testing"
+
+	"ragflow/internal/common"
+	"ragflow/internal/dao"
+	"ragflow/internal/entity"
+)
+
+// testDatasetCreateService builds a DatasetService for CreateDataset tests.
+// CreateDataset resolves the tenant via d.tenantDAO, so it must be wired here
+// (unlike the update path which does not touch it).
+// insertCreateDatasetTenant seeds a tenant with status="1" (required by
+// TenantDAO.GetByID, which CreateDataset calls) for the given tenant id.
+func insertCreateDatasetTenant(t *testing.T, tenantID string) {
+	t.Helper()
+	var existing entity.Tenant
+	if err := dao.DB.Where("id = ?", tenantID).First(&existing).Error; err != nil {
+		tn := &entity.Tenant{
+			ID:           tenantID,
+			LLMID:        "llm-default",
+			EmbdID:       "embd-default",
+			TenantEmbdID: sptr("embd-1"),
+			ASRID:        "asr-default",
+			Status:       sptr("1"),
+		}
+		if err := dao.DB.Create(tn).Error; err != nil {
+			t.Fatalf("insert test tenant: %v", err)
+		}
+	}
+}
+
+func testDatasetCreateService(t *testing.T) *DatasetService {
+	t.Helper()
+
+	return &DatasetService{
+		kbDAO:        dao.NewKnowledgebaseDAO(),
+		documentDAO:  dao.NewDocumentDAO(),
+		connectorDAO: dao.NewConnectorDAO(),
+		tenantDAO:    dao.NewTenantDAO(),
+	}
+}
+
+func TestCreateDataset_NoComponentParams(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertCreateDatasetTenant(t, "tenant-1")
+
+	chunkMethod := "naive"
+	result, code, err := testDatasetCreateService(t).CreateDataset(&CreateDatasetRequest{
+		Name:     "ds-no-cp",
+		ParserID: &chunkMethod,
+	}, "tenant-1")
+	if err != nil {
+		t.Fatalf("CreateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	if result["parser_id"] != strings.TrimSpace(chunkMethod) {
+		t.Fatalf("expected parser_id %q, got %#v", chunkMethod, result["parser_id"])
+	}
+}
+
+func TestCreateDataset_ComponentParamsPopulated(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertCreateDatasetTenant(t, "tenant-1")
+
+	chunkMethod := "naive"
+	result, code, err := testDatasetCreateService(t).CreateDataset(&CreateDatasetRequest{
+		Name:     "ds-with-cp-defaults",
+		ParserID: &chunkMethod,
+	}, "tenant-1")
+	if err != nil {
+		t.Fatalf("CreateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+
+	parserConfig, ok := result["parser_config"].(entity.JSONMap)
+	if !ok {
+		t.Fatalf("parser_config is not entity.JSONMap, got %T", result["parser_config"])
+	}
+	if len(parserConfig) == 0 {
+		t.Fatal("parser_config is empty, expected DSL component params defaults")
+	}
+
+	// Verify at least one component from the general/naive template has defaults.
+	// The general template has TokenChunker, Tokenizer, Parser, and File components.
+	found := false
+	for cpnID := range parserConfig {
+		if strings.Contains(cpnID, "TokenChunker") {
+			found = true
+			break
+		}
+	}
+	if !found {
+		t.Errorf("parser_config does not contain any TokenChunker: %v", parserConfig)
+	}
+}
+
+func TestCreateDataset_ParseTypeBuiltinClearsPipelineID(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertCreateDatasetTenant(t, "tenant-1")
+	// Seed a canvas so it exists, but parse_type=1 should ignore it.
+	seedDatasetUpdateCanvas(t, "abcdef0123456789abcdef0123456789", "tenant-1",
+		datasetUpdateCanvasDSL("Parser:HipSignsRhyme", "chunk_token_num"))
+
+	chunkMethod := "naive"
+	pipelineID := "ABCDEF0123456789ABCDEF0123456789"
+	parseType := 1
+
+	result, code, err := testDatasetCreateService(t).CreateDataset(&CreateDatasetRequest{
+		Name:       "ds-builtin-clears-pipeline",
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+		ParseType:  &parseType,
+	}, "tenant-1")
+	if err != nil {
+		t.Fatalf("CreateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	// parse_type=1 clears pipeline_id → only parser_id should be persisted.
+	if result["parser_id"] != chunkMethod {
+		t.Fatalf("expected parser_id %q, got %#v", chunkMethod, result["parser_id"])
+	}
+	if pid, ok := result["pipeline_id"]; ok && pid != nil && pid != "" {
+		t.Fatalf("expected pipeline_id to be cleared for BuiltIn mode, got %#v", pid)
+	}
+}
+
+func TestCreateDataset_ParseTypePipelineIgnoresParserID(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertCreateDatasetTenant(t, "tenant-1")
+	seedDatasetUpdateCanvas(t, "abcdef0123456789abcdef0123456789", "tenant-1",
+		datasetUpdateCanvasDSL("Parser:CustomP", "chunk_token_num"))
+
+	chunkMethod := "book"
+	pipelineID := "ABCDEF0123456789ABCDEF0123456789"
+	parseType := 2
+
+	result, code, err := testDatasetCreateService(t).CreateDataset(&CreateDatasetRequest{
+		Name:       "ds-pipeline-ignores-parser",
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+		ParseType:  &parseType,
+	}, "tenant-1")
+	if err != nil {
+		t.Fatalf("CreateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	// parse_type=2 ignores parser_id → pipeline_id should be persisted;
+	// parser_id should fall back to the default ("naive") since it wasn't set.
+	if result["pipeline_id"] != strings.ToLower(pipelineID) {
+		t.Fatalf("expected pipeline_id %q, got %#v", strings.ToLower(pipelineID), result["pipeline_id"])
+	}
+}
+
+func TestCreateDataset_RejectsBothWithoutParseType(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertCreateDatasetTenant(t, "tenant-1")
+
+	chunkMethod := "naive"
+	pipelineID := "abcdef0123456789abcdef0123456789"
+	_, code, err := testDatasetCreateService(t).CreateDataset(&CreateDatasetRequest{
+		Name:       "ds-both-no-parse-type",
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+		// ParseType deliberately nil
+	}, "tenant-1")
+	if err == nil {
+		t.Fatal("expected mutual-exclusivity error when both set without parse_type")
+	}
+	if code != common.CodeDataError {
+		t.Fatalf("expected data error code, got %d", code)
+	}
+	if !strings.Contains(err.Error(), "mutually exclusive") {
+		t.Fatalf("expected error to mention 'mutually exclusive', got: %v", err)
+	}
+}
diff --git a/internal/service/dataset_update_test.go b/internal/service/dataset_update_test.go
index aab189338c..368af05e04 100644
--- a/internal/service/dataset_update_test.go
+++ b/internal/service/dataset_update_test.go
@@ -17,6 +17,7 @@
 package service
 
 import (
+	"encoding/json"
 	"strings"
 	"testing"
 
@@ -39,24 +40,14 @@ func TestDatasetServiceUpdateDatasetUpdatesFields(t *testing.T) {
 	permission := string(entity.TenantPermissionTeam)
 	chunkMethod := string(entity.ParserTypeBook)
 	embeddingModel := "BAAI/bge-large-zh-v1.5@Builtin"
-	pipelineID := "ABCDEF0123456789ABCDEF0123456789"
 
 	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
 		Name:           &name,
 		Description:    &description,
 		Language:       &language,
 		Permission:     &permission,
-		ChunkMethod:    &chunkMethod,
+		ParserID:       &chunkMethod,
 		EmbeddingModel: &embeddingModel,
-		PipelineID:     &pipelineID,
-		ParserConfig: map[string]interface{}{
-			"parent_child": map[string]interface{}{
-				"use_parent_child": true,
-			},
-			"ext": map[string]interface{}{
-				"delimiter": "\n\n",
-			},
-		},
 	})
 	if err != nil {
 		t.Fatalf("UpdateDataset failed: %v", err)
@@ -67,15 +58,12 @@ func TestDatasetServiceUpdateDatasetUpdatesFields(t *testing.T) {
 	if result["name"] != "Renamed Dataset" {
 		t.Fatalf("expected trimmed name in response, got %#v", result["name"])
 	}
-	if result["chunk_method"] != chunkMethod {
-		t.Fatalf("expected chunk method %q, got %#v", chunkMethod, result["chunk_method"])
+	if result["parser_id"] != chunkMethod {
+		t.Fatalf("expected parser_id %q, got %#v", chunkMethod, result["parser_id"])
 	}
 	if result["embedding_model"] != embeddingModel {
 		t.Fatalf("expected embedding model %q, got %#v", embeddingModel, result["embedding_model"])
 	}
-	if result["pipeline_id"] != strings.ToLower(pipelineID) {
-		t.Fatalf("expected normalized pipeline id, got %#v", result["pipeline_id"])
-	}
 	if connectors, ok := result["connectors"].([]*dao.ConnectorDatasetListItem); !ok || len(connectors) != 0 {
 		t.Fatalf("expected empty connector list, got %#v", result["connectors"])
 	}
@@ -99,14 +87,92 @@ func TestDatasetServiceUpdateDatasetUpdatesFields(t *testing.T) {
 	if persisted.EmbdID != embeddingModel {
 		t.Fatalf("expected embd id %q, got %q", embeddingModel, persisted.EmbdID)
 	}
-	if persisted.PipelineID == nil || *persisted.PipelineID != strings.ToLower(pipelineID) {
-		t.Fatalf("expected normalized pipeline id persisted, got %#v", persisted.PipelineID)
+	// parser_config stores DSL runtime component params directly.
+	if _, ok := persisted.ParserConfig["Parser:HipSignsRhyme"].(map[string]interface{}); !ok {
+		t.Fatalf("expected Parser:HipSignsRhyme in parser_config, got %#v", persisted.ParserConfig)
 	}
-	if pc, ok := persisted.ParserConfig["parent_child"].(map[string]interface{}); !ok || pc["use_parent_child"] != true {
-		t.Fatalf("expected parent_child preserved as nested, got %#v", persisted.ParserConfig)
+}
+
+func TestUpdateDataset_RejectsSimultaneousParserIDAndPipelineID(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertDatasetUpdateKB(t, "kb-1", "tenant-1", "Original")
+
+	chunkMethod := "book"
+	pipelineID := "abcdef0123456789abcdef0123456789"
+
+	_, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+	})
+	if err == nil {
+		t.Fatal("expected mutual-exclusivity error when both parser_id and pipeline_id are set")
 	}
-	if pc, ok := persisted.ParserConfig["ext"].(map[string]interface{}); !ok || pc["delimiter"] != "\n\n" {
-		t.Fatalf("expected ext preserved as nested, got %#v", persisted.ParserConfig)
+	if code != common.CodeDataError {
+		t.Fatalf("expected data error code, got %d", code)
+	}
+	if !strings.Contains(err.Error(), "mutually exclusive") {
+		t.Fatalf("expected error to mention 'mutually exclusive', got: %v", err)
+	}
+}
+
+func TestUpdateDataset_ParseTypeBuiltinClearsPipelineID(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertDatasetUpdateKB(t, "kb-1", "tenant-1", "Original")
+	seedDatasetUpdateCanvas(t, "abcdef0123456789abcdef0123456789", "tenant-1",
+		datasetUpdateCanvasDSL("Parser:HipSignsRhyme", "chunk_token_num"))
+
+	chunkMethod := "book"
+	pipelineID := "ABCDEF0123456789ABCDEF0123456789"
+	parseType := 1
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+		ParseType:  &parseType,
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	// parse_type=1 clears pipeline_id → only parser_id should be set.
+	if result["parser_id"] != chunkMethod {
+		t.Fatalf("expected parser_id %q, got %#v", chunkMethod, result["parser_id"])
+	}
+	if pid, ok := result["pipeline_id"]; ok && pid != nil && pid != "" {
+		t.Fatalf("expected pipeline_id to be cleared for BuiltIn mode, got %#v", pid)
+	}
+}
+
+func TestUpdateDataset_ParseTypePipelineIgnoresParserID(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertDatasetUpdateKB(t, "kb-1", "tenant-1", "Original")
+	seedDatasetUpdateCanvas(t, "abcdef0123456789abcdef0123456789", "tenant-1",
+		datasetUpdateCanvasDSL("Parser:CustomP", "chunk_token_num"))
+
+	chunkMethod := "book"
+	pipelineID := "ABCDEF0123456789ABCDEF0123456789"
+	parseType := 2
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserID:   &chunkMethod,
+		PipelineID: &pipelineID,
+		ParseType:  &parseType,
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	// parse_type=2 ignores parser_id → pipeline_id should be set;
+	// parser_id should keep the original value.
+	if result["pipeline_id"] != strings.ToLower(pipelineID) {
+		t.Fatalf("expected pipeline_id %q, got %#v", strings.ToLower(pipelineID), result["pipeline_id"])
 	}
 }
 
@@ -466,3 +532,238 @@ func insertDatasetUpdateTenantModel(t *testing.T, id, providerID, instanceID, mo
 		t.Fatalf("insert test tenant model: %v", err)
 	}
 }
+
+// seedDatasetUpdateCanvas migrates user_canvas on the active test DB and
+// inserts a canvas row with the given DSL.
+func seedDatasetUpdateCanvas(t *testing.T, id, userID string, dslJSON []byte) {
+	t.Helper()
+	if err := dao.DB.AutoMigrate(&entity.UserCanvas{}); err != nil {
+		t.Fatalf("migrate user_canvas: %v", err)
+	}
+	var dslMap map[string]any
+	if err := json.Unmarshal(dslJSON, &dslMap); err != nil {
+		t.Fatalf("unmarshal seed dsl: %v", err)
+	}
+	canvas := &entity.UserCanvas{
+		ID:             id,
+		UserID:         userID,
+		Tags:           "",
+		Permission:     "me",
+		CanvasCategory: "agent_canvas",
+		DSL:            entity.JSONMap(dslMap),
+	}
+	if err := dao.DB.Create(canvas).Error; err != nil {
+		t.Fatalf("seed canvas: %v", err)
+	}
+}
+
+// datasetUpdateCanvasDSL builds a minimal canvas DSL declaring a Parser
+// component with the given cpnID and param keys.
+func datasetUpdateCanvasDSL(cpnID string, paramKeys ...string) []byte {
+	params := map[string]any{"outputs": map[string]any{}}
+	for _, k := range paramKeys {
+		params[k] = map[string]any{}
+	}
+	dsl := map[string]any{
+		"components": map[string]any{
+			cpnID: map[string]any{
+				"obj": map[string]any{"component_name": "Parser", "params": params},
+			},
+		},
+	}
+	raw, _ := json.Marshal(dsl)
+	return raw
+}
+
+// --- Step 3: component_params validation wired into UpdateDataset ---
+
+// insertDatasetUpdateCanvasKB seeds a KB bound to a custom canvas pipeline
+// (PipelineID set, ParserID empty) for the canvas validation tests.
+func insertDatasetUpdateCanvasKB(t *testing.T, id, tenantID, name, pipelineID string) {
+	t.Helper()
+
+	pid := pipelineID
+	kb := &entity.Knowledgebase{
+		ID:           id,
+		TenantID:     tenantID,
+		Name:         name,
+		EmbdID:       "BAAI/bge-large-zh-v1.5@Builtin",
+		CreatedBy:    tenantID,
+		Permission:   string(entity.TenantPermissionMe),
+		ParserID:     "",
+		PipelineID:   &pid,
+		ParserConfig: entity.JSONMap{"chunk_token_num": float64(128)},
+		Status:       sptr(string(entity.StatusValid)),
+	}
+	if err := dao.DB.Create(kb).Error; err != nil {
+		t.Fatalf("insert test canvas kb: %v", err)
+	}
+}
+
+func TestUpdateDataset_StripsUnknownParam_Builtin(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertDatasetUpdateKB(t, "kb-1", "tenant-1", "Original")
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserConfig: map[string]interface{}{
+			"Parser:HipSignsRhyme": map[string]interface{}{
+				"no_such_param": 1,
+			},
+		},
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset should succeed (unknown params are stripped): %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+
+	persisted, err := dao.NewKnowledgebaseDAO().GetByID("kb-1")
+	if err != nil {
+		t.Fatalf("get updated kb: %v", err)
+	}
+	cp := map[string]interface{}(persisted.ParserConfig)
+	rhyme, ok := cp["Parser:HipSignsRhyme"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("expected Parser:HipSignsRhyme persisted, got %#v", cp)
+	}
+	if _, exists := rhyme["no_such_param"]; exists {
+		t.Fatal("expected no_such_param to be stripped")
+	}
+	if result["parser_id"] != string(entity.ParserTypeNaive) {
+		t.Fatalf("expected parser_id preserved, got %#v", result["parser_id"])
+	}
+}
+
+func TestUpdateDataset_AcceptsValidComponentParams_Builtin(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	insertDatasetUpdateKB(t, "kb-1", "tenant-1", "Original")
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserConfig: map[string]interface{}{
+			"Parser:HipSignsRhyme": map[string]interface{}{
+				"pdf": map[string]interface{}{"parse_method": "deepdoc"},
+			},
+		},
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	if result["parser_id"] != string(entity.ParserTypeNaive) {
+		t.Fatalf("expected parser_id preserved, got %#v", result["parser_id"])
+	}
+
+	persisted, err := dao.NewKnowledgebaseDAO().GetByID("kb-1")
+	if err != nil {
+		t.Fatalf("get updated kb: %v", err)
+	}
+	cp := map[string]interface{}(persisted.ParserConfig)
+	if len(cp) == 0 {
+		t.Fatalf("expected component_params persisted, got %#v", persisted.ParserConfig)
+	}
+	rhyme, ok := cp["Parser:HipSignsRhyme"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("expected Parser:HipSignsRhyme persisted, got %#v", cp)
+	}
+	pdf, ok := rhyme["pdf"].(map[string]interface{})
+	if !ok || pdf["parse_method"] != "deepdoc" {
+		t.Fatalf("expected pdf setup persisted, got %#v", rhyme)
+	}
+}
+
+func TestUpdateDataset_StripsCanvasUnknownParam(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	dsl := datasetUpdateCanvasDSL("Parser:CustomRhyme", "pdf")
+	seedDatasetUpdateCanvas(t, "canvas-1", "tenant-1", dsl)
+	insertDatasetUpdateCanvasKB(t, "kb-1", "tenant-1", "Original", "canvas-1")
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserConfig: map[string]interface{}{
+			"Parser:NoSuch": map[string]interface{}{
+				"pdf": map[string]interface{}{},
+			},
+		},
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset should succeed (unknown cpnID is stripped): %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+
+	persisted, err := dao.NewKnowledgebaseDAO().GetByID("kb-1")
+	if err != nil {
+		t.Fatalf("get updated kb: %v", err)
+	}
+	cp := map[string]interface{}(persisted.ParserConfig)
+	if _, exists := cp["Parser:NoSuch"]; exists {
+		t.Fatal("expected unknown cpnID Parser:NoSuch to be stripped")
+	}
+	// The valid cpnID from the canvas DSL should still be present.
+	if _, exists := cp["Parser:CustomRhyme"]; !exists {
+		t.Fatal("expected Parser:CustomRhyme (from canvas DSL defaults) to be present")
+	}
+	if result["pipeline_id"] != "canvas-1" {
+		t.Fatalf("expected pipeline_id preserved, got %#v", result["pipeline_id"])
+	}
+}
+
+func TestUpdateDataset_AcceptsValidCanvasComponentParams(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	dsl := datasetUpdateCanvasDSL("Parser:CustomRhyme", "pdf")
+	seedDatasetUpdateCanvas(t, "canvas-1", "tenant-1", dsl)
+	insertDatasetUpdateCanvasKB(t, "kb-1", "tenant-1", "Original", "canvas-1")
+
+	result, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserConfig: map[string]interface{}{
+			"Parser:CustomRhyme": map[string]interface{}{
+				"pdf": map[string]interface{}{},
+			},
+		},
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+	if result["pipeline_id"] != "canvas-1" {
+		t.Fatalf("expected pipeline_id preserved, got %#v", result["pipeline_id"])
+	}
+}
+
+// TestUpdateDataset_SwitchCanvasToBuiltinValidatesAgainstBuiltin covers the
+// effective-ref edge where a request parser_id clears the existing canvas
+// pipeline: the override must be validated against the new builtin template,
+// not the stale canvas.
+func TestUpdateDataset_SwitchCanvasToBuiltinValidatesAgainstBuiltin(t *testing.T) {
+	db := setupDatasetUpdateTestDB(t)
+	pushServiceDB(t, db)
+	dsl := datasetUpdateCanvasDSL("Parser:CustomRhyme", "pdf")
+	seedDatasetUpdateCanvas(t, "canvas-1", "tenant-1", dsl)
+	insertDatasetUpdateCanvasKB(t, "kb-1", "tenant-1", "Original", "canvas-1")
+
+	chunkMethod := "naive"
+	_, code, err := testDatasetUpdateService(t).UpdateDataset("kb-1", "tenant-1", UpdateDatasetRequest{
+		ParserID: &chunkMethod,
+		ParserConfig: map[string]interface{}{
+			// Valid for the "general" builtin template, not the canvas.
+			"Parser:HipSignsRhyme": map[string]interface{}{
+				"pdf": map[string]interface{}{"parse_method": "deepdoc"},
+			},
+		},
+	})
+	if err != nil {
+		t.Fatalf("UpdateDataset failed: %v", err)
+	}
+	if code != common.CodeSuccess {
+		t.Fatalf("expected success code, got %d", code)
+	}
+}
diff --git a/internal/service/document.go b/internal/service/document.go
index d3343bf8da..2928e55b4c 100644
--- a/internal/service/document.go
+++ b/internal/service/document.go
@@ -38,6 +38,7 @@ import (
 	"ragflow/internal/engine"
 	enginetypes "ragflow/internal/engine/types"
 	"ragflow/internal/entity"
+	pipelinepkg "ragflow/internal/ingestion/pipeline"
 	"ragflow/internal/storage"
 	"ragflow/internal/tokenizer"
 	"ragflow/internal/utility"
@@ -85,15 +86,6 @@ func NewDocumentService() *DocumentService {
 }
 
 // CreateDocumentRequest create document request
-type CreateDocumentRequest struct {
-	Name      string `json:"name" binding:"required"`
-	KbID      string `json:"kb_id" binding:"required"`
-	ParserID  string `json:"parser_id" binding:"required"`
-	CreatedBy string `json:"created_by" binding:"required"`
-	Type      string `json:"type"`
-	Source    string `json:"source"`
-}
-
 // UpdateDocumentRequest update document request
 type UpdateDocumentRequest struct {
 	Name        *string  `json:"name"`
@@ -102,8 +94,6 @@ type UpdateDocumentRequest struct {
 	ChunkNum    *int64   `json:"chunk_num"`
 	Progress    *float64 `json:"progress"`
 	ProgressMsg *string  `json:"progress_msg"`
-	// FIXME: need to confirm below field
-	ProcessDuration *float64 `json:"progress_duration"`
 }
 
 // DocumentResponse document response
@@ -147,7 +137,6 @@ type ArtifactResponse struct {
 
 type UpdateDatasetDocumentRequest struct {
 	Name         *string        `json:"name"`
-	ChunkMethod  *string        `json:"chunk_method"`
 	ParserID     *string        `json:"parser_id"`
 	ChunkCount   *int64         `json:"chunk_count"`
 	TokenCount   *int64         `json:"token_count"`
@@ -163,7 +152,7 @@ type UpdateDatasetDocumentResponse struct {
 	ID              string                 `json:"id"`
 	Thumbnail       *string                `json:"thumbnail,omitempty"`
 	DatasetID       string                 `json:"dataset_id"`
-	ChunkMethod     string                 `json:"chunk_method"`
+	ParserID        string                 `json:"parser_id"`
 	PipelineID      *string                `json:"pipeline_id,omitempty"`
 	ParserConfig    map[string]interface{} `json:"parser_config"`
 	SourceType      string                 `json:"source_type"`
@@ -551,27 +540,6 @@ func (s *DocumentService) DownloadDocument(datasetID, docID string) (*DownloadDo
 }
 
 // CreateDocument create document
-func (s *DocumentService) CreateDocument(req *CreateDocumentRequest) (*entity.Document, error) {
-	document := &entity.Document{
-		ID:           utility.GenerateUUID(),
-		Name:         &req.Name,
-		KbID:         req.KbID,
-		ParserID:     req.ParserID,
-		ParserConfig: entity.JSONMap{},
-		CreatedBy:    req.CreatedBy,
-		Type:         req.Type,
-		SourceType:   req.Source,
-		Suffix:       ".doc",
-		Status:       func() *string { s := "1"; return &s }(),
-	}
-
-	if err := s.InsertDocument(document); err != nil {
-		return nil, fmt.Errorf("failed to create document: %w", err)
-	}
-
-	return document, nil
-}
-
 // GetDocumentByID get document by ID
 func (s *DocumentService) GetDocumentByID(id string) (*DocumentResponse, error) {
 	document, err := s.documentDAO.GetByID(id)
@@ -1222,29 +1190,6 @@ func (s *DocumentService) StartParseDocuments(doc *entity.Document, kb *entity.K
 		}
 	}
 
-	if opts.ApplyKB {
-		if doc.ParserConfig == nil {
-			doc.ParserConfig = entity.JSONMap{}
-		}
-		config := map[string]interface{}{
-			"llm_id":          kb.ParserConfig["llm_id"],
-			"enable_metadata": false,
-			"metadata":        map[string]interface{}{},
-		}
-		if value, ok := kb.ParserConfig["enable_metadata"]; ok {
-			config["enable_metadata"] = value
-		}
-		if value, ok := kb.ParserConfig["metadata"]; ok {
-			config["metadata"] = value
-		}
-		if err := s.updateDocumentParserConfig(doc.ID, config); err != nil {
-			return err
-		}
-		for key, value := range config {
-			doc.ParserConfig[key] = value
-		}
-	}
-
 	if _, err := s.IngestDocuments(doc.KbID, userID, []string{doc.ID}); err != nil {
 		return err
 	}
@@ -2201,7 +2146,7 @@ func (s *DocumentService) UpdateDatasetDocument(userID, datasetID, documentID st
 		return nil, common.CodeServerError, err
 	}
 
-	if code, err := s.validateDatasetDocumentUpdate(doc, req, present); err != nil {
+	if code, err := s.validateDatasetDocumentUpdate(datasetID, documentID, userID, doc, req, present); err != nil {
 		return nil, code, err
 	}
 
@@ -2218,24 +2163,61 @@ func (s *DocumentService) UpdateDatasetDocument(userID, datasetID, documentID st
 	}
 
 	if present["parser_config"] && req.ParserConfig != nil {
-		if err := s.updateDocumentParserConfig(doc.ID, req.ParserConfig); err != nil {
-			return nil, common.CodeDataError, err
+		// Resolve effective pipeline to load the DSL for cleaning.
+		isCanvas := kb.PipelineID != nil && strings.TrimSpace(*kb.PipelineID) != ""
+		if req.PipelineID != nil {
+			isCanvas = strings.TrimSpace(*req.PipelineID) != ""
+		}
+		if req.ParserID != nil {
+			isCanvas = false
+		}
+		effParserID := kb.ParserID
+		if req.ParserID != nil {
+			effParserID = strings.TrimSpace(*req.ParserID)
+		}
+		effPipelineID := kb.PipelineID
+		if req.PipelineID != nil {
+			effPipelineID = req.PipelineID
+		}
+		if req.ParserID != nil && req.PipelineID == nil && kb.PipelineID != nil {
+			effPipelineID = nil
+		}
+
+		dslJSON, err := loadPipelineDSL(isCanvas, effParserID, effPipelineID)
+		if err != nil {
+			common.Warn("cleanAndUpdateDocumentParserConfig: failed to load DSL, falling back to merge",
+				zap.Error(err))
+			if err := s.updateDocumentParserConfig(doc.ID, req.ParserConfig); err != nil {
+				return nil, common.CodeDataError, err
+			}
+		} else {
+			cleaned := buildParserConfig(dslJSON, map[string]interface{}(req.ParserConfig))
+			if err := s.documentDAO.UpdateByID(doc.ID, map[string]interface{}{
+				"parser_config": cleaned,
+			}); err != nil {
+				return nil, common.CodeDataError, err
+			}
 		}
 	}
 
-	if req.PipelineID != nil && *req.PipelineID != "" {
-		if err := s.resetDocumentForReparse(doc, kb.TenantID, nil, req.PipelineID); err != nil {
-			return nil, common.CodeDataError, err
+	if present["pipeline_id"] {
+		if req.PipelineID != nil && strings.TrimSpace(*req.PipelineID) != "" {
+			if err := s.resetDocumentForReparse(doc, kb.TenantID, nil, req.PipelineID); err != nil {
+				return nil, common.CodeDataError, err
+			}
+		} else {
+			// Explicitly cleared: drop the custom canvas so the worker falls
+			// back to the built-in template, matching validation.
+			empty := ""
+			if err := s.resetDocumentForReparse(doc, kb.TenantID, nil, &empty); err != nil {
+				return nil, common.CodeDataError, err
+			}
 		}
 	} else if present["parser_id"] && req.ParserID != nil && strings.TrimSpace(*req.ParserID) != "" {
 		parserID := strings.TrimSpace(*req.ParserID)
 		if err := s.resetDocumentForReparse(doc, kb.TenantID, &parserID, nil); err != nil {
 			return nil, common.CodeDataError, err
 		}
-	} else if req.ChunkMethod != nil && *req.ChunkMethod != "" {
-		if err := s.updateChunkMethod(doc, kb.TenantID, *req.ChunkMethod, req.ParserConfig, present["parser_config"]); err != nil {
-			return nil, common.CodeDataError, err
-		}
 	}
 
 	if present["enabled"] && req.Enabled != nil {
@@ -2260,7 +2242,7 @@ func (s *DocumentService) UpdateDatasetDocument(userID, datasetID, documentID st
 	return s.toUpdateDatasetDocumentResponse(updatedDoc, metaFields), common.CodeSuccess, nil
 }
 
-func (s *DocumentService) validateDatasetDocumentUpdate(doc *entity.Document, req *UpdateDatasetDocumentRequest, present map[string]bool) (common.ErrorCode, error) {
+func (s *DocumentService) validateDatasetDocumentUpdate(datasetID, documentID, userID string, doc *entity.Document, req *UpdateDatasetDocumentRequest, present map[string]bool) (common.ErrorCode, error) {
 	if req == nil {
 		return common.CodeDataError, errors.New("Invalid request payload")
 	}
@@ -2280,18 +2262,6 @@ func (s *DocumentService) validateDatasetDocumentUpdate(doc *entity.Document, re
 		}
 	}
 
-	if present["chunk_method"] {
-		if req.ChunkMethod == nil || strings.TrimSpace(*req.ChunkMethod) == "" {
-			return common.CodeDataError, errors.New("`chunk_method` (empty string) is not valid")
-		}
-		chunkMethod := strings.TrimSpace(*req.ChunkMethod)
-		if err := validateParserID(chunkMethod); err != nil {
-			return common.CodeDataError, fmt.Errorf("`chunk_method` %s doesn't exist", chunkMethod)
-		}
-		if doc.Type == "visual" || isPresentationFile(doc.Name) {
-			return common.CodeDataError, errors.New("Not supported yet!")
-		}
-	}
 	if present["parser_id"] && req.ParserID != nil {
 		parserID := strings.TrimSpace(*req.ParserID)
 		if (doc.Type == "visual" && parserID != "picture") || (isPresentationFile(doc.Name) && parserID != "presentation") {
@@ -2446,6 +2416,138 @@ func (s *DocumentService) updateDocumentNameOnly(doc *entity.Document, tenantID,
 	)
 }
 
+// loadCanvasDSLJSON returns the DSL JSON for a custom canvas pipeline. The
+// canvas's dsl column holds the same component-graph structure that built-in
+// templates use, so it can be validated by the same schema extractor. It is a
+// package-level function so both document and knowledge-base updates reuse it.
+func loadCanvasDSLJSON(canvasID string) ([]byte, error) {
+	if strings.TrimSpace(canvasID) == "" {
+		return nil, fmt.Errorf("empty canvas id")
+	}
+	canvas, err := dao.NewUserCanvasDAO().GetByID(canvasID)
+	if err != nil {
+		if errors.Is(err, dao.ErrUserCanvasNotFound) {
+			return nil, fmt.Errorf("canvas %s not found", canvasID)
+		}
+		return nil, fmt.Errorf("load canvas %s: %w", canvasID, err)
+	}
+	if len(canvas.DSL) == 0 {
+		return nil, fmt.Errorf("canvas %s has no DSL", canvasID)
+	}
+	raw, err := json.Marshal(canvas.DSL)
+	if err != nil {
+		return nil, fmt.Errorf("marshal canvas %s DSL: %w", canvasID, err)
+	}
+	return raw, nil
+}
+
+// loadPipelineDSL loads the DSL JSON for a pipeline identified by parserID
+// (built-in) or pipelineID (custom canvas). When both are provided, isCanvas
+// selects which one to use.
+func loadPipelineDSL(isCanvas bool, parserID string, pipelineID *string) ([]byte, error) {
+	if isCanvas {
+		return loadCanvasDSLJSON(strings.TrimSpace(*pipelineID))
+	}
+	registry, err := pipelinepkg.DefaultRegistry()
+	if err != nil {
+		return nil, fmt.Errorf("builtin pipeline registry: %w", err)
+	}
+	if !registry.IsValid(parserID) {
+		return nil, fmt.Errorf("unknown builtin parser_id: %s", parserID)
+	}
+	dslStr, err := pipelinepkg.LoadBuiltinDSL(parserID)
+	if err != nil {
+		return nil, fmt.Errorf("load builtin DSL for %q: %w", parserID, err)
+	}
+	return []byte(dslStr), nil
+}
+
+// cleanComponentParams filters rawConfig against the DSL schema given by dslJSON.
+// Keys containing ':' are treated as component IDs; they are kept only when both
+// the cpnID AND the param name exist in the DSL schema. Keys without ':' (legacy
+// flat fields such as chunk_token_num, image_context_size) are dropped with a
+// warning — they do not belong in the new component-params world.
+func cleanComponentParams(dslJSON []byte, rawConfig map[string]interface{}) map[string]interface{} {
+	schemas, err := pipelinepkg.ExtractAllComponentParams(dslJSON)
+	if err != nil {
+		common.Warn("cleanComponentParams: failed to extract DSL schema, returning input as-is",
+			zap.Error(err))
+		return rawConfig
+	}
+
+	validCPNs := make(map[string]map[string]struct{}, len(schemas))
+	for _, s := range schemas {
+		keys := make(map[string]struct{}, len(s.ParamsDefaults))
+		for k := range s.ParamsDefaults {
+			keys[k] = struct{}{}
+		}
+		validCPNs[s.CpnID] = keys
+	}
+
+	result := make(map[string]interface{}, len(rawConfig))
+	for key, val := range rawConfig {
+		if !strings.Contains(key, ":") {
+			common.Warn("cleanComponentParams: dropping legacy flat field",
+				zap.String("key", key))
+			continue
+		}
+		validKeys, ok := validCPNs[key]
+		if !ok {
+			common.Warn("cleanComponentParams: dropping unknown cpnID",
+				zap.String("cpnID", key))
+			continue
+		}
+		params, ok := val.(map[string]any)
+		if !ok {
+			continue
+		}
+		cleaned := make(map[string]any, len(params))
+		for pk, pv := range params {
+			if _, ok := validKeys[pk]; ok {
+				cleaned[pk] = pv
+			} else {
+				common.Warn("cleanComponentParams: dropping unknown param",
+					zap.String("cpnID", key), zap.String("param", pk))
+			}
+		}
+		if len(cleaned) > 0 {
+			result[key] = cleaned
+		}
+	}
+	return result
+}
+
+// buildParserConfig builds the final parser_config by starting from the DSL
+// defaults for every component, then overlaying the cleaned incoming overrides.
+// This ensures all components from the current pipeline are present while
+// stripping stale params from other pipelines.
+func buildParserConfig(dslJSON []byte, rawConfig map[string]interface{}) entity.JSONMap {
+	cleaned := cleanComponentParams(dslJSON, rawConfig)
+	defaults, err := pipelinepkg.ComponentParamsDefaults(dslJSON)
+	if err != nil {
+		common.Warn("buildParserConfig: failed to extract DSL defaults, using cleaned only",
+			zap.Error(err))
+		return entity.JSONMap(cleaned)
+	}
+	result := make(entity.JSONMap, len(defaults))
+	for cpnID, params := range defaults {
+		base := make(map[string]interface{}, len(params))
+		for k, v := range params {
+			base[k] = v
+		}
+		if over, ok := cleaned[cpnID]; ok {
+			if om, ok := over.(map[string]any); ok {
+				result[cpnID] = common.DeepMergeMaps(base, map[string]interface{}(om))
+			} else {
+				result[cpnID] = base
+			}
+		} else {
+			result[cpnID] = base
+		}
+	}
+	return result
+}
+
 func (s *DocumentService) updateDocumentParserConfig(documentID string, config map[string]any) error {
 	if len(config) == 0 {
 		return nil
@@ -2598,22 +2700,6 @@ func (s *DocumentService) decrementDocumentAndKBCountersForReparse(doc *entity.D
 	return decremented, err
 }
 
-func (s *DocumentService) updateChunkMethod(doc *entity.Document, tenantID string, chunkMethod string, parserConfig map[string]any, hasParserConfig bool) error {
-	chunkMethod = strings.TrimSpace(chunkMethod)
-	if !strings.EqualFold(doc.ParserID, chunkMethod) {
-		if err := s.resetDocumentForReparse(doc, tenantID, &chunkMethod, nil); err != nil {
-			return err
-		}
-	}
-	if !hasParserConfig {
-		defaultConfig := common.GetParserConfig(chunkMethod, nil)
-		if err := s.updateDocumentParserConfig(doc.ID, defaultConfig); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
 func (s *DocumentService) updateDocumentStatusOnly(doc *entity.Document, kb *entity.Knowledgebase, status int) error {
 	statusStr := strconv.Itoa(status)
 	if doc.Status != nil && *doc.Status == statusStr {
@@ -2646,7 +2732,7 @@ func (s *DocumentService) toUpdateDatasetDocumentResponse(doc *entity.Document,
 		ID:              doc.ID,
 		Thumbnail:       doc.Thumbnail,
 		DatasetID:       doc.KbID,
-		ChunkMethod:     doc.ParserID,
+		ParserID:        doc.ParserID,
 		PipelineID:      doc.PipelineID,
 		ParserConfig:    map[string]interface{}(doc.ParserConfig),
 		SourceType:      doc.SourceType,
@@ -2991,11 +3077,15 @@ func (s *DocumentService) newDatasetDocument(kb *entity.Knowledgebase, tenantID,
 	if i := strings.LastIndex(filename, "."); i >= 0 {
 		suffix = filename[i+1:]
 	}
+	parserID := selectUploadParser(utility.FileType(filetype), filename, kb.ParserID)
+	if kb.PipelineID != nil {
+		parserID = "" // canvas pipeline mode — parser_id not applicable
+	}
 	loc := location
 	doc := &entity.Document{
 		ID:           docID,
 		KbID:         kb.ID,
-		ParserID:     selectUploadParser(utility.FileType(filetype), filename, kb.ParserID),
+		ParserID:     parserID,
 		PipelineID:   kb.PipelineID,
 		ParserConfig: parserConfig,
 		CreatedBy:    tenantID,
@@ -3012,11 +3102,24 @@ func (s *DocumentService) newDatasetDocument(kb *entity.Knowledgebase, tenantID,
 		hash := contentHashHex(blob)
 		doc.ContentHash = &hash
 	}
+
+	// When the document's builtin parser_id differs from the KB's (e.g. visual→picture,
+	// aural→audio), re-resolve component_params defaults from the document's own DSL
+	// template so cpnIDs and param keys match the pipeline that will actually execute.
+	if kb.PipelineID == nil && parserID != kb.ParserID {
+		if cp, err := resolveComponentParamsDefaults(parserID, nil); err != nil {
+			common.Warn("newDatasetDocument: resolve component_params defaults",
+				zap.String("parserID", parserID), zap.Error(err))
+		} else if cp != nil {
+			doc.ParserConfig = cp
+		}
+	}
+
 	return doc
 }
 
 // docToRawMap serialises a freshly created Document into the raw key shape the
-// handler remaps (chunk_num→chunk_count, kb_id→dataset_id, parser_id→chunk_method).
+// handler remaps (chunk_num→chunk_count, kb_id→dataset_id).
 func docToRawMap(doc *entity.Document) map[string]interface{} {
 	m := map[string]interface{}{
 		"id":            doc.ID,
diff --git a/internal/service/document_test.go b/internal/service/document_test.go
index 938d3fa47b..0320b50376 100644
--- a/internal/service/document_test.go
+++ b/internal/service/document_test.go
@@ -567,29 +567,6 @@ func insertTestFile(t *testing.T, id, parentID, name string, location *string) {
 	}
 }
 
-func TestCreateDocumentIncrementsKBDocNum(t *testing.T) {
-	db := setupServiceTestDB(t)
-	pushServiceDB(t, db)
-	insertTestKB(t, "kb-create", "tenant-1", 0, 0, 0)
-
-	svc := testDocumentService(t)
-	doc, err := svc.CreateDocument(&CreateDocumentRequest{
-		Name:      "created.txt",
-		KbID:      "kb-create",
-		ParserID:  "naive",
-		CreatedBy: "tenant-1",
-		Type:      "doc",
-		Source:    "local",
-	})
-	if err != nil {
-		t.Fatalf("CreateDocument failed: %v", err)
-	}
-	if doc == nil || doc.KbID != "kb-create" {
-		t.Fatalf("unexpected doc: %+v", doc)
-	}
-	assertKBDocNum(t, "kb-create", 1)
-}
-
 func TestDeleteDocumentFull_Basic(t *testing.T) {
 	db := setupServiceTestDB(t)
 	pushServiceDB(t, db)
@@ -812,8 +789,8 @@ func TestUploadLocalDocuments_MirrorsPythonCoreFields(t *testing.T) {
 	if doc["location"] != "nested/path/deck(1).pptx" {
 		t.Fatalf("location=%v, want nested/path/deck(1).pptx", doc["location"])
 	}
-	if doc["parser_id"] != "presentation" {
-		t.Fatalf("parser_id=%v, want presentation", doc["parser_id"])
+	if doc["parser_id"] != "" {
+		t.Fatalf("parser_id=%v, want empty (canvas pipeline mode)", doc["parser_id"])
 	}
 	if doc["content_hash"] != "06b05ab6733a618578af5f94892f3950" {
 		t.Fatalf("content_hash=%v", doc["content_hash"])
@@ -860,7 +837,7 @@ func TestUploadEmptyDocument_CreatesVirtualDocumentAndFileLink(t *testing.T) {
 	if code != common.CodeSuccess {
 		t.Fatalf("code=%v, want success", code)
 	}
-	if got["type"] != "virtual" || got["parser_id"] != "manual" || got["size"] != int64(0) {
+	if got["type"] != "virtual" || got["parser_id"] != "" || got["size"] != int64(0) {
 		t.Fatalf("unexpected doc map: %v", got)
 	}
 
@@ -1723,7 +1700,7 @@ func TestUpdateDatasetDocumentRenameUpdatesDocumentAndFile(t *testing.T) {
 	}
 }
 
-func TestUpdateDatasetDocumentChunkMethodResetsForReparse(t *testing.T) {
+func TestUpdateDatasetDocumentParserIDResetsForReparse(t *testing.T) {
 	db := setupServiceTestDB(t)
 	pushServiceDB(t, db)
 	insertTestKB(t, "kb-1", "tenant-1", 1, 10, 5)
@@ -1732,12 +1709,12 @@ func TestUpdateDatasetDocumentChunkMethodResetsForReparse(t *testing.T) {
 	chunkMethod := "manual"
 	svc := testDocumentService(t)
 	resp, code, err := svc.UpdateDatasetDocument("tenant-1", "kb-1", "doc-1", &UpdateDatasetDocumentRequest{
-		ChunkMethod: &chunkMethod,
-	}, map[string]bool{"chunk_method": true})
+		ParserID: &chunkMethod,
+	}, map[string]bool{"parser_id": true})
 	if err != nil {
 		t.Fatalf("UpdateDatasetDocument failed: code=%v err=%v", code, err)
 	}
-	if resp.ChunkMethod != chunkMethod || resp.Run != "UNSTART" || resp.TokenCount != 0 || resp.ChunkCount != 0 {
+	if resp.ParserID != chunkMethod || resp.Run != "UNSTART" || resp.TokenCount != 0 || resp.ChunkCount != 0 {
 		t.Fatalf("response = %+v, want method=%s run=UNSTART counts=0", resp, chunkMethod)
 	}
 
@@ -1754,37 +1731,6 @@ func TestUpdateDatasetDocumentChunkMethodResetsForReparse(t *testing.T) {
 	}
 }
 
-func TestUpdateDatasetDocumentParserIDResetsForReparse(t *testing.T) {
-	db := setupServiceTestDB(t)
-	pushServiceDB(t, db)
-	insertTestKB(t, "kb-1", "tenant-1", 1, 10, 5)
-	insertNamedTestDoc(t, "doc-1", "kb-1", "doc.txt", 10, 5)
-
-	parserID := "manual"
-	svc := testDocumentService(t)
-	resp, code, err := svc.UpdateDatasetDocument("tenant-1", "kb-1", "doc-1", &UpdateDatasetDocumentRequest{
-		ParserID: &parserID,
-	}, map[string]bool{"parser_id": true})
-	if err != nil {
-		t.Fatalf("UpdateDatasetDocument failed: code=%v err=%v", code, err)
-	}
-	if resp.ChunkMethod != parserID || resp.Run != "UNSTART" || resp.TokenCount != 0 || resp.ChunkCount != 0 {
-		t.Fatalf("response = %+v, want parser_id=%s run=UNSTART counts=0", resp, parserID)
-	}
-
-	doc, _ := dao.NewDocumentDAO().GetByID("doc-1")
-	if doc.ParserID != parserID {
-		t.Fatalf("parser_id = %q, want %q", doc.ParserID, parserID)
-	}
-	if doc.TokenNum != 0 || doc.ChunkNum != 0 {
-		t.Fatalf("doc counters = token:%d chunk:%d, want zero", doc.TokenNum, doc.ChunkNum)
-	}
-	kb, _ := dao.NewKnowledgebaseDAO().GetByID("kb-1")
-	if kb.TokenNum != 0 || kb.ChunkNum != 0 {
-		t.Fatalf("kb counters = token:%d chunk:%d, want zero", kb.TokenNum, kb.ChunkNum)
-	}
-}
-
 func TestResetDocumentForReparseSkipsSecondCounterDecrement(t *testing.T) {
 	db := setupServiceTestDB(t)
 	pushServiceDB(t, db)
@@ -2326,7 +2272,7 @@ func TestMergeFieldValuesKeepsNumericValues(t *testing.T) {
 	}
 }
 
-func TestUpdateDatasetDocumentPipelineIDTakesPrecedenceOverChunkMethod(t *testing.T) {
+func TestUpdateDatasetDocumentPipelineIDTakesPrecedenceOverParserID(t *testing.T) {
 	db := setupServiceTestDB(t)
 	pushServiceDB(t, db)
 	insertTestKB(t, "kb-1", "tenant-1", 1, 10, 5)
@@ -2336,17 +2282,17 @@ func TestUpdateDatasetDocumentPipelineIDTakesPrecedenceOverChunkMethod(t *testin
 	chunkMethod := "manual"
 	svc := testDocumentService(t)
 	resp, code, err := svc.UpdateDatasetDocument("tenant-1", "kb-1", "doc-1", &UpdateDatasetDocumentRequest{
-		PipelineID:  &pipelineID,
-		ChunkMethod: &chunkMethod,
-	}, map[string]bool{"pipeline_id": true, "chunk_method": true})
+		PipelineID: &pipelineID,
+		ParserID:   &chunkMethod,
+	}, map[string]bool{"pipeline_id": true, "parser_id": true})
 	if err != nil {
 		t.Fatalf("UpdateDatasetDocument failed: code=%v err=%v", code, err)
 	}
 	if resp.PipelineID == nil || *resp.PipelineID != pipelineID {
 		t.Fatalf("pipeline_id = %v, want %q", resp.PipelineID, pipelineID)
 	}
-	if resp.ChunkMethod != "naive" {
-		t.Fatalf("chunk_method = %q, want original naive", resp.ChunkMethod)
+	if resp.ParserID != "naive" {
+		t.Fatalf("parser_id = %q, want original naive", resp.ParserID)
 	}
 }
 
diff --git a/internal/service/document_upload_helpers.go b/internal/service/document_upload_helpers.go
index b8a6e6a8c4..f277f84913 100644
--- a/internal/service/document_upload_helpers.go
+++ b/internal/service/document_upload_helpers.go
@@ -2,37 +2,21 @@ package service
 
 import (
 	"encoding/hex"
-	"path/filepath"
-	"regexp"
-	"strings"
 
+	pipelinepkg "ragflow/internal/ingestion/pipeline"
 	"ragflow/internal/utility"
 
 	"github.com/zeebo/xxh3"
 )
 
-var (
-	presentationUploadPattern = regexp.MustCompile(`(?i)\.(ppt|pptx|pages)$`)
-	emailUploadPattern        = regexp.MustCompile(`(?i)\.(msg|eml)$`)
-)
-
-// selectUploadParser mirrors Python FileService.get_parser.
+// selectUploadParser resolves the parser_id for a document by delegating to
+// the builtin pipeline registry, which owns the file-type-to-parser mapping.
 func selectUploadParser(docType utility.FileType, filename, defaultParser string) string {
-	switch docType {
-	case utility.FileTypeVISUAL:
-		return "picture"
-	case utility.FileTypeAURAL:
-		return "audio"
-	}
-	base := filepath.Base(strings.TrimSpace(filename))
-	switch {
-	case presentationUploadPattern.MatchString(base):
-		return "presentation"
-	case emailUploadPattern.MatchString(base):
-		return "email"
-	default:
+	registry, err := pipelinepkg.DefaultRegistry()
+	if err != nil || registry == nil {
 		return defaultParser
 	}
+	return registry.DefaultParserID(string(docType), filename, defaultParser)
 }
 
 // contentHashHex mirrors Python xxhash.xxh128(blob).hexdigest().
diff --git a/internal/service/file2document.go b/internal/service/file2document.go
index da130c638d..efe8b15270 100644
--- a/internal/service/file2document.go
+++ b/internal/service/file2document.go
@@ -215,6 +215,20 @@ func (s *File2DocumentService) convertFiles(fileIDs, kbIDs []string, userID stri
 			}
 			if kb.PipelineID != nil {
 				doc.PipelineID = kb.PipelineID
+				doc.ParserID = "" // canvas pipeline mode — parser_id not applicable
+			}
+
+			// When the document's builtin parser_id differs from the KB's
+			// (e.g. visual→picture, aural→audio), re-resolve component_params
+			// defaults from the document's own DSL template so cpnIDs and
+			// param keys match the pipeline that will actually execute.
+			if kb.PipelineID == nil && parserID != kb.ParserID {
+				if cp, err := resolveComponentParamsDefaults(parserID, nil); err != nil {
+					common.Warn("convertFiles: resolve component_params defaults",
+						zap.String("parserID", parserID), zap.Error(err))
+				} else if cp != nil {
+					doc.ParserConfig = cp
+				}
 			}
 
 			// InsertDocument creates the row and increments KB doc_num in one
diff --git a/internal/tokenizer/tokenizer_concurrent_test.go b/internal/tokenizer/tokenizer_concurrent_test.go
index 877da7d30c..24261a97d0 100644
--- a/internal/tokenizer/tokenizer_concurrent_test.go
+++ b/internal/tokenizer/tokenizer_concurrent_test.go
@@ -30,7 +30,7 @@ import (
 
 func init() {
 	// Initialize logger for tests
-	if err := common.Init("info", common.FileOutput{}); err != nil {
+	if err := common.Init("info", common.FileOutput{}, ""); err != nil {
 		fmt.Printf("Failed to initialize logger: %v\n", err)
 	}
 }
diff --git a/test/testcases/restful_api/test_datasets.py b/test/testcases/restful_api/test_datasets.py
index bad4c9ce98..ceffa74445 100644
--- a/test/testcases/restful_api/test_datasets.py
+++ b/test/testcases/restful_api/test_datasets.py
@@ -28,6 +28,7 @@ from test.testcases.utils.file_utils import create_image_file, create_txt_file
 
 
 ARGUMENT_ERROR_CODE = 102 if IS_GO_PROXY else 101
+PARSER_ID_FIELD = "parser_id" if IS_GO_PROXY else "chunk_method"
 
 
 def _skip_go_ignored_null(payload, field):
@@ -159,7 +160,7 @@ def test_dataset_update_language_connectors_avatar_and_description_contract(rest
         json={
             "name": "dataset_update_lang_connectors",
             "description": "",
-            "chunk_method": "naive",
+            PARSER_ID_FIELD: "naive",
             "language": "English",
             "connectors": [],
             "avatar": avatar_value,
@@ -210,12 +211,12 @@ def test_dataset_update_chunk_method_contract(rest_client, clear_datasets, chunk
 
     update_res = rest_client.put(
         f"/datasets/{dataset_id}",
-        json={"chunk_method": chunk_method},
+        json={PARSER_ID_FIELD: chunk_method},
     )
     assert update_res.status_code == 200
     update_payload = update_res.json()
     assert update_payload["code"] == 0, update_payload
-    assert update_payload["data"]["chunk_method"] == chunk_method, update_payload
+    assert update_payload["data"][PARSER_ID_FIELD] == chunk_method, update_payload
 
 
 @pytest.mark.p1
@@ -362,9 +363,9 @@ def test_dataset_update_parser_config_valid_matrix_contract(rest_client, clear_d
 @pytest.mark.parametrize(
     "name, update_payload",
     [
-        ("parser_config_empty", {"chunk_method": "qa", "parser_config": {}}),
-        ("parser_config_none", {"chunk_method": "qa", "parser_config": None}),
-        ("parser_config_unset", {"chunk_method": "qa"}),
+        ("parser_config_empty", {PARSER_ID_FIELD: "qa", "parser_config": {}}),
+        ("parser_config_none", {PARSER_ID_FIELD: "qa", "parser_config": None}),
+        ("parser_config_unset", {PARSER_ID_FIELD: "qa"}),
     ],
     ids=["parser_config_empty", "parser_config_none", "parser_config_unset"],
 )
@@ -895,21 +896,21 @@ def test_dataset_update_chunk_method_invalid_contract(rest_client, clear_dataset
 
     expected_chunk_message = "Input should be 'naive', 'book', 'email', 'laws', 'manual', 'one', 'paper', 'picture', 'presentation', 'qa', 'table', 'tag' or 'resume'"
     for chunk_method in ("", "unknown", []):
-        res = rest_client.put(f"/datasets/{dataset_id}", json={"chunk_method": chunk_method})
+        res = rest_client.put(f"/datasets/{dataset_id}", json={PARSER_ID_FIELD: chunk_method})
         assert res.status_code == 200
         payload = res.json()
         assert payload["code"] == ARGUMENT_ERROR_CODE, payload
         if IS_GO_PROXY and not isinstance(chunk_method, str):
-            assert "cannot unmarshal" in payload["message"] and ".chunk_method" in payload["message"], payload
+            assert "cannot unmarshal" in payload["message"] and f".{PARSER_ID_FIELD}" in payload["message"], payload
         elif IS_GO_PROXY:
             assert payload["message"].startswith("Input should be 'audio', 'book'") and payload["message"].endswith("or 'tag'"), payload
         else:
             assert expected_chunk_message in payload["message"], payload
 
-    none_res = rest_client.put(f"/datasets/{dataset_id}", json={"chunk_method": None})
+    none_res = rest_client.put(f"/datasets/{dataset_id}", json={PARSER_ID_FIELD: None})
     assert none_res.status_code == 200
     none_payload = none_res.json()
-    _skip_go_ignored_null(none_payload, "chunk_method")
+    _skip_go_ignored_null(none_payload, PARSER_ID_FIELD)
     assert none_payload["code"] == ARGUMENT_ERROR_CODE, none_payload
     if IS_GO_PROXY:
         assert none_payload["message"].startswith("Input should be 'audio', 'book'") and none_payload["message"].endswith("or 'tag'"), none_payload
@@ -1072,7 +1073,7 @@ def test_dataset_update_field_unset_and_unsupported_contract(rest_client, clear_
     assert after_list_payload["data"][0]["description"] == original_data["description"], after_list_payload
     assert after_list_payload["data"][0]["embedding_model"] == original_data["embedding_model"], after_list_payload
     assert after_list_payload["data"][0]["permission"] == original_data["permission"], after_list_payload
-    assert after_list_payload["data"][0]["chunk_method"] == original_data["chunk_method"], after_list_payload
+    assert after_list_payload["data"][0][PARSER_ID_FIELD] == original_data[PARSER_ID_FIELD], after_list_payload
     assert after_list_payload["data"][0]["pagerank"] == original_data["pagerank"], after_list_payload
     assert after_list_payload["data"][0]["parser_config"] == original_data["parser_config"], after_list_payload
 
@@ -1142,6 +1143,8 @@ def test_dataset_create_name_and_case_insensitive_contract(rest_client, clear_da
 
 @pytest.mark.p2
 def test_dataset_create_avatar_and_description_contract(rest_client, clear_datasets):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept avatar/description")
     avatar = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/w8AAgMBgN6J1tQAAAAASUVORK5CYII="
     payload = {
         "name": "dataset_avatar_description",
@@ -1176,11 +1179,11 @@ def test_dataset_create_avatar_and_description_contract(rest_client, clear_datas
     ids=["naive", "book", "email", "laws", "manual", "one", "paper", "picture", "presentation", "qa", "table", "tag"],
 )
 def test_dataset_create_chunk_method_contract(rest_client, clear_datasets, name, chunk_method):
-    res = rest_client.post("/datasets", json={"name": name, "chunk_method": chunk_method})
+    res = rest_client.post("/datasets", json={"name": name, PARSER_ID_FIELD: chunk_method})
     assert res.status_code == 200
     payload = res.json()
     assert payload["code"] == 0, payload
-    assert payload["data"]["chunk_method"] == chunk_method, payload
+    assert payload["data"][PARSER_ID_FIELD] == chunk_method, payload
 
 
 @pytest.mark.p2
@@ -1277,6 +1280,8 @@ def test_dataset_create_embedding_model_format_contract(rest_client, clear_datas
 
 @pytest.mark.p2
 def test_dataset_create_parser_config_missing_raptor_and_graphrag(rest_client, clear_datasets):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     payload = {
         "name": "test_parser_config_missing_fields",
         "parser_config": {"chunk_token_num": 1024},
@@ -1429,6 +1434,8 @@ def test_dataset_create_concurrent_contract(rest_client, clear_datasets):
     ],
 )
 def test_dataset_create_parser_config_valid_matrix_contract(rest_client, clear_datasets, name, parser_config):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     payload = {"name": name, "parser_config": parser_config}
     res = rest_client.post("/datasets", json=payload)
     assert res.status_code == 200
@@ -1459,6 +1466,8 @@ def test_dataset_create_parser_config_valid_matrix_contract(rest_client, clear_d
     ids=["only_raptor", "only_graphrag", "both_fields"],
 )
 def test_dataset_create_parser_config_bugfix_contract(rest_client, clear_datasets, name, parser_config, expected_raptor, expected_graphrag):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     res = rest_client.post("/datasets", json={"name": name, "parser_config": parser_config})
     assert res.status_code == 200
     body = res.json()
@@ -1478,9 +1487,11 @@ def test_dataset_create_parser_config_bugfix_contract(rest_client, clear_dataset
     ids=["qa", "manual", "paper", "book", "laws", "presentation"],
 )
 def test_dataset_create_parser_config_different_chunk_methods_contract(rest_client, clear_datasets, chunk_method):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     payload = {
         "name": f"test_parser_config_{chunk_method}",
-        "chunk_method": chunk_method,
+        PARSER_ID_FIELD: chunk_method,
         "parser_config": {"chunk_token_num": 512},
     }
     res = rest_client.post("/datasets", json=payload)
@@ -1550,6 +1561,8 @@ def test_dataset_create_content_type_and_payload_bad_contract(rest_client):
 
 @pytest.mark.p2
 def test_dataset_create_avatar_contract(rest_client, clear_datasets, tmp_path):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept avatar/description")
     exceed_res = rest_client.post(
         "/datasets",
         json={"name": "avatar_exceeds_limit_length", "avatar": "a" * 65536},
@@ -1594,6 +1607,8 @@ def test_dataset_create_avatar_contract(rest_client, clear_datasets, tmp_path):
 
 @pytest.mark.p2
 def test_dataset_create_description_contract(rest_client, clear_datasets):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept avatar/description")
     exceeds_limit_res = rest_client.post(
         "/datasets",
         json={"name": "description_exceeds_limit_length", "description": "a" * 65536},
@@ -1657,18 +1672,18 @@ def test_dataset_create_permission_and_chunk_method_contract(rest_client, clear_
     ]
     expected_chunk_message = "Input should be 'naive', 'book', 'email', 'laws', 'manual', 'one', 'paper', 'picture', 'presentation', 'qa', 'table', 'tag' or 'resume'"
     for name, chunk_method in chunk_method_invalid_cases:
-        res = rest_client.post("/datasets", json={"name": name, "chunk_method": chunk_method})
+        res = rest_client.post("/datasets", json={"name": name, PARSER_ID_FIELD: chunk_method})
         assert res.status_code == 200
         payload = res.json()
         assert payload["code"] == ARGUMENT_ERROR_CODE, payload
         if IS_GO_PROXY and not isinstance(chunk_method, str):
-            assert "cannot unmarshal" in payload["message"] and ".chunk_method" in payload["message"], payload
+            assert "cannot unmarshal" in payload["message"] and f".{PARSER_ID_FIELD}" in payload["message"], payload
         elif IS_GO_PROXY:
             assert payload["message"].startswith("Input should be 'audio', 'book'") and payload["message"].endswith("or 'tag'"), payload
         else:
             assert expected_chunk_message in payload["message"], payload
 
-    chunk_method_none_res = rest_client.post("/datasets", json={"name": "chunk_method_none", "chunk_method": None})
+    chunk_method_none_res = rest_client.post("/datasets", json={"name": "chunk_method_none", PARSER_ID_FIELD: None})
     assert chunk_method_none_res.status_code == 200
     chunk_method_none_payload = chunk_method_none_res.json()
     assert chunk_method_none_payload["code"] == ARGUMENT_ERROR_CODE, chunk_method_none_payload
@@ -1681,11 +1696,13 @@ def test_dataset_create_permission_and_chunk_method_contract(rest_client, clear_
     assert chunk_method_unset_res.status_code == 200
     chunk_method_unset_payload = chunk_method_unset_res.json()
     assert chunk_method_unset_payload["code"] == 0, chunk_method_unset_payload
-    assert chunk_method_unset_payload["data"]["chunk_method"] == "naive", chunk_method_unset_payload
+    assert chunk_method_unset_payload["data"][PARSER_ID_FIELD] == "naive", chunk_method_unset_payload
 
 
 @pytest.mark.p2
 def test_dataset_create_parser_config_invalid_contract(rest_client, clear_datasets):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     invalid_cases = [
         ("auto_keywords_min_limit", {"auto_keywords": -1}, "Input should be greater than or equal to 0"),
         ("auto_keywords_max_limit", {"auto_keywords": 33}, "Input should be less than or equal to 32"),
@@ -1754,6 +1771,8 @@ def test_dataset_create_parser_config_invalid_contract(rest_client, clear_datase
 
 @pytest.mark.p2
 def test_dataset_create_parser_config_defaults_and_extra_fields_contract(rest_client, clear_datasets):
+    if IS_GO_PROXY:
+        pytest.skip("Go CreateDataset does not accept parser_config")
     empty_res = rest_client.post("/datasets", json={"name": "parser_config_empty", "parser_config": {}})
     assert empty_res.status_code == 200
     empty_payload = empty_res.json()