Hi! The beta version of the assistants api seems a bit complicated.
I know some people want to recreate a whole chat with chat history, so they may need the thread object etc…
But lots of use cases are just call->response. We optimized our system to call your api as a little function, in->out, with a templated consistant predictable response, already QA, no need to go back and forth. If it fails we retry, but normally it doesn’t fail.
And you the API runs all the tools in the background, and just returns the final result. Maybe the “final result” can be a list of messages, or the Final result can be streamed for people who implemented api streaming,
But it’s still always we send 1 API call with a json, and we get 1 response with a json. 100x simpler to implement.
Thank you soooo much for everything, you absolutely rock!!