support streaming for functions (#6115)

langchain/chat_models/openai.py

@@ -321,16 +321,27 @@ def _generate(
             inner_completion = ""
             role = "assistant"
             params["stream"] = True
+            function_call: Optional[dict] = None
             for stream_resp in self.completion_with_retry(
                 messages=message_dicts, **params
             ):
                 role = stream_resp["choices"][0]["delta"].get("role", role)
-                token = stream_resp["choices"][0]["delta"].get("content", "")
+                token = stream_resp["choices"][0]["delta"].get("content") or ""
                 inner_completion += token
+                _function_call = stream_resp["choices"][0]["delta"].get("function_call")
+                if _function_call:
+                    if function_call is None:
+                        function_call = _function_call
+                    else:
+                        function_call["arguments"] += _function_call["arguments"]
                 if run_manager:
                     run_manager.on_llm_new_token(token)
             message = _convert_dict_to_message(
-                {"content": inner_completion, "role": role}
+                {
+                    "content": inner_completion,
+                    "role": role,
+                    "function_call": function_call,
+                }
             )
             return ChatResult(generations=[ChatGeneration(message=message)])
         response = self.completion_with_retry(messages=message_dicts, **params)