diff --git a/src/bisheng-langchain/bisheng_langchain/chat_models/sensetime.py b/src/bisheng-langchain/bisheng_langchain/chat_models/sensetime.py
index 7929a9f17..1761957ab 100644
--- a/src/bisheng-langchain/bisheng_langchain/chat_models/sensetime.py
+++ b/src/bisheng-langchain/bisheng_langchain/chat_models/sensetime.py
@@ -215,7 +215,12 @@ def _completion_with_retry(**kwargs: Any) -> Any:
                 "max_new_tokens": self.max_tokens,
                 'stream': False#self.streaming
             }
-            
+
+            token = encode_jwt_token(self.access_key_id, self.secret_access_key)
+            if isinstance(token, bytes):
+                token = token.decode('utf-8')
+            self.client.headers.update({'Authorization': 'Bearer {}'.format(token)})
+
             response = self.client.post(url=url, json=params).json()
             return response
         rsp_dict = _completion_with_retry(**kwargs)
@@ -231,6 +236,12 @@ def _completion_with_retry(**kwargs: Any) -> Any:
     async def acompletion_with_retry(self, **kwargs: Any) -> Any:
         """Use tenacity to retry the async completion call."""
         retry_decorator = _create_retry_decorator(self)
+
+        token = encode_jwt_token(self.access_key_id, self.secret_access_key)
+        if isinstance(token, bytes):
+            token = token.decode('utf-8')
+        self.client.headers.update({'Authorization': 'Bearer {}'.format(token)})
+        
         if self.streaming:
             self.client.headers.update({'Accept': 'text/event-stream'})
         else:
@@ -411,4 +422,4 @@ def _get_invocation_params(self,
     @property
     def _llm_type(self) -> str:
         """Return type of chat model."""
-        return 'sense-chat'
\ No newline at end of file
+        return 'sense-chat'