Diff between Assistants Tools and Function Calling?

[ben22]

So I recently have made a fastapi/monorepo tool using AI to manage small “microservices” that make up the codebase. I want to reduce the costs of the API calls, I am just wondering if I switch to only using the function calling will the AI still be able to call other functions to try to fix it’s error when it does? The assistant continues and will solve itself, but it is just costly.

I want my code base and services continue to be self healing, I haven’t tested the function calling API. Wanting some insights before I go ham on the docs.

[merefield]

@ben22

to call other functions to try to fix it’s error when it does?

Can you elaborate on the use case?

What are you expecting these functions to do and what hints will you provide the Completion engine to know that it needs to call them?

Semantically a looping local bot with local functions calling LLM for completions is the same as the Assitant, but the latter is presently a bit wasteful of context and processing cycles (as I understand it).