I faced the same issue, and you can resolve it by making the chat treatment (e.g., OpenAI calls) asynchronous, so the 200 response is returned instantly.