feat: basic boilerplate for 4-turbo #33
Conversation
|
So, I've added the same boilerplate for gpt-4-turbo as for other models, but during testing it always times out. It's not related to the page content size. I've noticed that if you comment out function calls (and e.g. add response_format: "json") it works just fine (regarding time). Also, if e.g. timeout: 10000 is added, it will always do it correctly on the third retry. I guess the problem might be langchain (there are some issues open about this model)? I'll be looking into this, but maybe @Patai5 you know something about it? |
|
@foxt451 That's weird, I'm testing the model currently and I'm having zero issues with it 🤷♂️ I guess they either fixed it since, or there might be a problem with your API key? It's taking very long to answer compared to other models, but I guess that is the price of 128k tokens :D Regarding the timeout you mentioned - it looks like langchain defaults to forever :D I guess we should set something there, how about 180 seconds? |
foxt451
changed the title
|
@Patai5 wow, have just rerun this and it worked fine this time. I guess merge then? (I've requested your review this time). Btw this model enables many capabilities like image/file sending, audio stuff, that could be added in other PRs. Especially to send images from the website |
|
I resolved the conflicts, hope I didn't screw anything :D |
|
Oops, my bad, sorry, I've somehow managed to lose this PR somewhere 😅 |
Related to #32