diff --git a/api/app/celery_app.py b/api/app/celery_app.py
index c087e1d7..ee32715b 100644
--- a/api/app/celery_app.py
+++ b/api/app/celery_app.py
@@ -44,8 +44,8 @@ celery_app.conf.update(
     task_ignore_result=False,
     
     # 超时设置
-    task_time_limit=1800,  # 30分钟硬超时
-    task_soft_time_limit=1500,  # 25分钟软超时
+    task_time_limit=3600,  # 60分钟硬超时
+    task_soft_time_limit=3000,  # 50分钟软超时
     
     # Worker 设置 (per-worker settings are in docker-compose command line)
     worker_prefetch_multiplier=1,  # Don't hoard tasks, fairer distribution
diff --git a/api/app/controllers/chunk_controller.py b/api/app/controllers/chunk_controller.py
index 620d8a1a..988aa706 100644
--- a/api/app/controllers/chunk_controller.py
+++ b/api/app/controllers/chunk_controller.py
@@ -441,14 +441,14 @@ async def retrieve_chunks(
     # 1 participle search, 2 semantic search, 3 hybrid search
     match retrieve_data.retrieve_type:
         case chunk_schema.RetrieveType.PARTICIPLE:
-            rs = vector_service.search_by_full_text(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.similarity_threshold)
+            rs = vector_service.search_by_full_text(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.similarity_threshold, file_names_filter=retrieve_data.file_names_filter)
             return success(data=rs, msg="retrieval successful")
         case chunk_schema.RetrieveType.SEMANTIC:
-            rs = vector_service.search_by_vector(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.vector_similarity_weight)
+            rs = vector_service.search_by_vector(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.vector_similarity_weight, file_names_filter=retrieve_data.file_names_filter)
             return success(data=rs, msg="retrieval successful")
         case _:
-            rs1 = vector_service.search_by_vector(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.vector_similarity_weight)
-            rs2 = vector_service.search_by_full_text(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.similarity_threshold)
+            rs1 = vector_service.search_by_vector(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.vector_similarity_weight, file_names_filter=retrieve_data.file_names_filter)
+            rs2 = vector_service.search_by_full_text(query=retrieve_data.query, top_k=retrieve_data.top_k, indices=indices, score_threshold=retrieve_data.similarity_threshold, file_names_filter=retrieve_data.file_names_filter)
             # Efficient deduplication
             seen_ids = set()
             unique_rs = []
diff --git a/api/app/schemas/chunk_schema.py b/api/app/schemas/chunk_schema.py
index cef9b9cb..ce8f70f2 100644
--- a/api/app/schemas/chunk_schema.py
+++ b/api/app/schemas/chunk_schema.py
@@ -46,6 +46,7 @@ class ChunkUpdate(BaseModel):
 class ChunkRetrieve(BaseModel):
     query: str
     kb_ids: list[uuid.UUID]
+    file_names_filter: list[str] | None = Field(None)
     similarity_threshold: float | None = Field(None)
     vector_similarity_weight: float | None = Field(None)
     top_k: int | None = Field(None)
diff --git a/api/app/tasks.py b/api/app/tasks.py
index 299d188b..093f081f 100644
--- a/api/app/tasks.py
+++ b/api/app/tasks.py
@@ -257,7 +257,7 @@ def parse_document(file_path: str, document_id: uuid.UUID):
                 progress_msg += f"{datetime.now().strftime('%H:%M:%S')} GraphRAG task result for task {task}:\n{result}\n"
                 return result
 
-            try:
+            def sync_task():
                 trio.run(
                     lambda: _run(
                         row=task,
@@ -272,6 +272,10 @@ def parse_document(file_path: str, document_id: uuid.UUID):
                         with_community=with_community,
                     )
                 )
+            try:
+                with ThreadPoolExecutor(max_workers=1) as executor:
+                    future = executor.submit(sync_task)
+                    future.result()  # Blocks until the task completes
             except Exception as e:
                 progress_msg += f"{datetime.now().strftime('%H:%M:%S')} GraphRAG task failed for task {task}:\n{str(e)}\n"
             progress_msg += f"{datetime.now().strftime('%H:%M:%S')} Knowledge Graph done ({time.time() - start_time}s)"