Working with gpt-3.5-turbo-instruct — possible to use GPT-4-turbo or GPT-4o (especially 4o)?

[nknakhr]

Referring to past issues, I was able to get the code working by replacing all instances of "text-davinci-00x" (x = 1 to 3) in defunct_run_gpt_prompt.py, gpt_structure.py, and run_gpt_prompt.py with "gpt-3.5-turbo-instruct". I really appreciate the insights shared by others — they were very helpful.

As far as I understand, this setup now works fine with gpt-3.5-turbo-instruct.
That said, has anyone tried running this with newer models like gpt-4-turbo or gpt-4o (especially 4o)?

If you're aware of any changes needed in the code (such as adjustments in API call structure or response handling), I’d be very grateful for your advice or pointers.

Thank you!

[pj0620]

It would be cool to use another library to abstract away direct model calls. That way, you can just plug and play any model it supports. For example, adding LangChain as a dependency and using a generic interface like BaseChatModel at the core would make it easy to swap between different models and providers with minimal changes.

All of the integrations LangChain supports (linked below) would work out of the box:

• OpenAI
• Anthropic
• Local Models