ChatGPT API

shouldn't it be as easy as just changing the model in LLMPredictor?

oh I see it is not same as completions endpoint.

There’s an additional parameter “messages”. not sure how easy/difficult that would be to implement

I tried, it didnt work

is the messages your entire conversation with the assistant?

I think so…which is kinda meh to me. Like what if you have many messages lol, I guess it’s up to us to determine how many messages we want to be in context

But Now that I think about it, each chat thread probably won’t be more than a small number of messages anyways…in theory

does it persist the conversation after you send a request to chat/completions endpoint or you can only create a conversation once and then that's it?

What I mean is that they don't return conversation-id unlike how they do it on browser.