fix: extra newline token in stream mode

basic_demo/openai_api_server.py

@@ -484,7 +484,8 @@ async def predict_stream(model_id, gen_params):
                     object="chat.completion.chunk"
                 )
                 yield "{}".format(chunk.model_dump_json(exclude_unset=True))
-
+            if not has_send_first_chunk and output.startswith("\n"):
+                output = output[1:]
             send_msg = delta_text if has_send_first_chunk else output
             has_send_first_chunk = True
             message = DeltaMessage(