update

common/common.py

@@ -3,12 +3,20 @@ from enum import Enum
 
 
 def configure_logging(level=logging.INFO, config_file_path='./common.log'):
+    
+    formatter = logging.Formatter("[%(asctime)s.%(msecs)03d] %(module)30s:%(lineno)4d %(levelname)-7s - %(message)s")
+    
+    console_handler = logging.StreamHandler()
+    console_handler.setLevel(level) 
+    console_handler.setFormatter(formatter) 
+    
     logging.basicConfig(
         filename=config_file_path,
         filemode="a",
         level=level,
         datefmt="%Y-%m-%d %H:%M:%S",
         format="[%(asctime)s.%(msecs)03d] %(module)30s:%(lineno)4d %(levelname)-7s - %(message)s",
+        handlers=[console_handler]
     )
 
 

components/services/dialogue.py

@@ -68,7 +68,7 @@ class DialogueService:
         except Exception as e:
             logger.error(f"Error in _postprocess_qe: {e}")
             from_chat = self._get_search_query(history)
-            return QEResult(use_search=from_chat is not None, search_query=from_chat)
+            return QEResult(use_search=from_chat is not None, search_query=from_chat.content)
 
     def _get_qe_request(self, history: List[Message]) -> ChatRequest:
         """
@@ -121,7 +121,7 @@ class DialogueService:
 
         return QEResult(use_search=bool_var, search_query=second_part)
 
-    def _get_search_query(self, history: List[Message]) -> str | None:
+    def _get_search_query(self, history: List[Message]) -> Message | None:
         """
         Получает запрос для поиска на основе последнего сообщения пользователя.
         """

routes/llm.py

@@ -117,38 +117,48 @@ async def sse_generator(request: ChatRequest, llm_api: DeepInfraApi, system_prom
     """
     Генератор для стриминга ответа LLM через SSE.
     """
+    try:
+        qe_result = await dialogue_service.get_qe_result(request.history)    
+        
+    except Exception as e:
+        logger.error(f"Error in SSE chat stream while dialogue_service.get_qe_result: {str(e)}", stack_info=True)
+        yield "data: {\"event\": \"error\", \"data\":\""+str(e)+"\" }\n\n"
     
-    qe_result = await dialogue_service.get_qe_result(request.history)    
-    
-    if qe_result.use_search and qe_result.search_query is not None:
-        dataset = dataset_service.get_current_dataset()
-        if dataset is None:
-            raise HTTPException(status_code=400, detail="Dataset not found")
-        _, scores, chunk_ids = entity_service.search_similar(qe_result.search_query, dataset.id)
-        chunks = entity_service.chunk_repository.get_chunks_by_ids(chunk_ids)
-        text_chunks = entity_service.build_text(chunks, scores)
-        search_results_event = {
-            "event": "search_results",
-            "data": f"{text_chunks}"
-        }
-        yield f"data: {json.dumps(search_results_event, ensure_ascii=False)}\n\n"
+    try:   
+        if qe_result.use_search and qe_result.search_query is not None:
+            dataset = dataset_service.get_current_dataset()
+            if dataset is None:
+                raise HTTPException(status_code=400, detail="Dataset not found")
+            _, scores, chunk_ids = entity_service.search_similar(qe_result.search_query, dataset.id)
+            chunks = entity_service.chunk_repository.get_chunks_by_ids(chunk_ids)
+            text_chunks = entity_service.build_text(chunks, scores)
+            search_results_event = {
+                "event": "search_results",
+                "data": f"{text_chunks}"
+            }
+            yield f"data: {json.dumps(search_results_event, ensure_ascii=False)}\n\n"
 
-        # new_message = f'<search-results>\n{text_chunks}\n</search-results>\n{last_query.content}'
-        
-        try_insert_search_results(request, text_chunks)
-        
-        
-    # Сворачиваем историю в первое сообщение
-    collapsed_request = collapse_history_to_first_message(request)
-                
-    # Стриминг токенов ответа
-    async for token in llm_api.get_predict_chat_generator(collapsed_request, system_prompt, predict_params):
-        token_event = {"event": "token", "data": token}
-        # logger.info(f"Streaming token: {token}")
-        yield f"data: {json.dumps(token_event, ensure_ascii=False)}\n\n"
+            # new_message = f'<search-results>\n{text_chunks}\n</search-results>\n{last_query.content}'
+            
+            try_insert_search_results(request, text_chunks)
+    except Exception as e:
+        logger.error(f"Error in SSE chat stream while searching: {str(e)}", stack_info=True)
+        yield "data: {\"event\": \"error\", \"data\":\""+str(e)+"\" }\n\n"
+    try:          
+        # Сворачиваем историю в первое сообщение
+        collapsed_request = collapse_history_to_first_message(request)
+                    
+        # Стриминг токенов ответа
+        async for token in llm_api.get_predict_chat_generator(collapsed_request, system_prompt, predict_params):
+            token_event = {"event": "token", "data": token}
+            # logger.info(f"Streaming token: {token}")
+            yield f"data: {json.dumps(token_event, ensure_ascii=False)}\n\n"
 
-    # Финальное событие
-    yield "data: {\"event\": \"done\"}\n\n"
+        # Финальное событие
+        yield "data: {\"event\": \"done\"}\n\n"
+    except Exception as e:
+        logger.error(f"Error in SSE chat stream while generating response: {str(e)}", stack_info=True)
+        yield "data: {\"event\": \"error\", \"data\":\""+str(e)+"\" }\n\n"
     
 
 @router.post("/chat/stream")