update

app/core/runner/llm_callback_handler.py

@@ -45,10 +45,6 @@ class LLMCallbackHandler:
             for chunk in response_stream:
                 logging.debug(chunk)
 
-                if chunk.usage:
-                    self.event_handler.pub_message_usage(chunk)
-                    continue
-
                 if not chunk.choices:
                     continue
 
@@ -93,6 +89,9 @@ class LLMCallbackHandler:
                     self.event_handler.pub_message_delta(
                         self.message.id, index, delta.content, delta.role
                     )
+                if chunk.usage:
+                    self.event_handler.pub_message_usage(chunk)
+                    continue
         except Exception as e:
             logging.error("handle_llm_response error: %s", e)
             raise e

app/core/runner/thread_runner.py

@@ -184,6 +184,7 @@ class ThreadRunner:
         )
         response_msg = llm_callback_handler.handle_llm_response(response_stream)
         message_creation_run_step = llm_callback_handler.step
+        print("444444444444444444444444455555555577777777777777777777777")
         logging.info("chat_response_message: %s", response_msg)
 
         if msg_util.is_tool_call(response_msg):