concurrent execution of chatgpt_api.py handle_post_chat_completions produce chaos

[liaoweiguo]

any hints?

[liaoweiguo]

streaming mode

[liaoweiguo]

没办法，强制单线程先：

  async def handle_post_chat_completions(self, request):
    async with self._lock:
      data = await request.json()
      if DEBUG >= 2: print(f"[ChatGPTAPI] Handling chat completions request from {request.remote}: {data}")
      stream = data.get("stream", False)

[d6638219]

I've also appeared

[Evanev7]

Should be fixed in 1.0!