fix: stabilize knowledge compilation navigation updates (#17345)

api/apps/restful_apis/agent_api.py

@@ -1135,6 +1135,9 @@ async def rerun_agent(tenant_id):
     if 0 < doc["progress"] < 1:
         return get_data_error_result(message=f"`{doc['name']}` is processing...")
 
+    from rag.advanced_rag.knowlege_compile.dataset_nav import remove_dataset_nav_doc_sync
+
+    remove_dataset_nav_doc_sync(tenant_id, doc["kb_id"], doc["id"])
     if settings.docStoreConn.index_exist(search.index_name(tenant_id), doc["kb_id"]):
         settings.docStoreConn.delete({"doc_id": doc["id"]}, search.index_name(tenant_id), doc["kb_id"])
     doc["progress_msg"] = ""

api/apps/restful_apis/document_api.py

@@ -1470,6 +1470,9 @@ def _run_sync(user_id: str, req):
         DocumentService.update_by_id(doc_id, info)
         if req.get("delete", False):
             TaskService.filter_delete([Task.doc_id == doc_id])
+            from rag.advanced_rag.knowlege_compile.dataset_nav import remove_dataset_nav_doc_sync
+
+            remove_dataset_nav_doc_sync(doc_tenant_id, doc.kb_id, doc.id)
             if settings.docStoreConn.index_exist(search.index_name(doc_tenant_id), doc.kb_id):
                 settings.docStoreConn.delete({"doc_id": doc_id}, search.index_name(doc_tenant_id), doc.kb_id)
 
@@ -1580,6 +1583,9 @@ async def parse_documents(tenant_id, dataset_id):
 
                 DocumentService.update_by_id(doc_id, info)
                 TaskService.filter_delete([Task.doc_id == doc_id])
+                from rag.advanced_rag.knowlege_compile.dataset_nav import remove_dataset_nav_doc_sync
+
+                remove_dataset_nav_doc_sync(tenant_id, doc.kb_id, doc.id)
                 if settings.docStoreConn.index_exist(search.index_name(tenant_id), doc.kb_id):
                     settings.docStoreConn.delete({"doc_id": doc_id}, search.index_name(tenant_id), doc.kb_id)
 

api/apps/services/dataset_api_service.py

@@ -2452,7 +2452,10 @@ def _nav_item(row: dict) -> dict:
         payload = json.loads(row.get("content_with_weight") or "{}")
     except Exception:
         payload = {}
-    is_cluster = (row.get("type_kwd") or payload.get("type")) == "nav_cluster"
+    row_type = row.get("type_kwd")
+    if isinstance(row_type, (list, tuple, set)):
+        row_type = next(iter(row_type), None)
+    is_cluster = (row_type or payload.get("type")) == "nav_cluster"
     return {
         "name": row.get("name") or "",
         "description": payload.get("description") or "",
@@ -2551,7 +2554,8 @@ async def delete_nav(dataset_id: str, tenant_id: str):
     index_nm, _ = pack
 
     try:
-        deleted = settings.docStoreConn.delete(
+        deleted = await thread_pool_exec(
+            settings.docStoreConn.delete,
             {"compile_kwd": [_NAV_COMPILE_KWD]},
             index_nm,
             dataset_id,
@@ -2592,16 +2596,17 @@ async def delete_nav_node(dataset_id: str, tenant_id: str, name: str):
         if not frontier:
             break
         try:
-            res = settings.docStoreConn.search(
-                select_fields=["name"],
-                highlight_fields=[],
-                condition={"compile_kwd": [_NAV_COMPILE_KWD], "parent_kwd": frontier},
-                match_expressions=[],
-                order_by=OrderByExpr(),
-                offset=0,
-                limit=10000,
-                index_names=index_nm,
-                knowledgebase_ids=[dataset_id],
+            res = await thread_pool_exec(
+                settings.docStoreConn.search,
+                ["name"],
+                [],
+                {"compile_kwd": [_NAV_COMPILE_KWD], "parent_kwd": frontier},
+                [],
+                OrderByExpr(),
+                0,
+                10000,
+                index_nm,
+                [dataset_id],
             )
             rows = settings.docStoreConn.get_fields(res, ["name"]) or {}
         except Exception:
@@ -2616,7 +2621,8 @@ async def delete_nav_node(dataset_id: str, tenant_id: str, name: str):
         frontier = nxt
 
     try:
-        deleted = settings.docStoreConn.delete(
+        deleted = await thread_pool_exec(
+            settings.docStoreConn.delete,
             {"compile_kwd": [_NAV_COMPILE_KWD], "name": list(names)},
             index_nm,
             dataset_id,

api/db/services/document_service.py

@@ -492,6 +492,20 @@ class DocumentService(CommonService):
         except Exception as e:
             logging.warning(f"Failed to delete thumbnail for document {doc.id}: {e}")
 
+        # Prune this doc's line from the KB's tree-kind navigation before the
+        # broad doc_id delete below removes the nav_doc row needed to locate
+        # and update its parent cluster.
+        try:
+            from rag.advanced_rag.knowlege_compile.dataset_nav import (
+                remove_dataset_nav_doc_sync,
+            )
+
+            remove_dataset_nav_doc_sync(tenant_id, doc.kb_id, doc.id)
+        except Exception as e:
+            logging.warning(
+                f"Failed to prune dataset_nav for document {doc.id}: {e}",
+            )
+
         # Delete chunks from doc store - this is critical, log errors
         try:
             settings.docStoreConn.delete({"doc_id": doc.id}, chunk_index_name, doc.kb_id)
@@ -507,20 +521,6 @@ class DocumentService(CommonService):
         except Exception as e:
             logging.warning(f"Failed to clean up artifact products for document {doc.id}: {e}")
 
-        # Prune this doc's line from the KB's tree-kind navigation
-        # markdown (best-effort — the markdown is a downstream artifact,
-        # and failure here must not block the document delete).
-        try:
-            from rag.advanced_rag.knowlege_compile.dataset_nav import (
-                remove_dataset_nav_doc_sync,
-            )
-
-            remove_dataset_nav_doc_sync(tenant_id, doc.kb_id, doc.id)
-        except Exception as e:
-            logging.warning(
-                f"Failed to prune dataset_nav for document {doc.id}: {e}",
-            )
-
         # Delete document metadata (non-critical, log and continue)
         try:
             DocMetadataService.delete_document_metadata(doc.id, doc.kb_id, tenant_id)