Different provider for FAST and SMART LLMs

[Speedway1]

For each of the models used, is it possible to specify their own LLM provider?

This is especially helpful with locally hosted models where the "smart" LLM can be hosted on a different server (more RAM, GPU capacity) than with "fast" LLM.

[assafelovic]

This is definitely a great and requested feature but will require some heavy changes. Atm the agent assumes a main LLM provider with the options of sub models. Happy to see suggested PRs for this or discuss design around this

[Speedway1]

OK brilliant. I will give it some thought and come back to you on this. Thank you for creating such a brilliant tool. It's absolutely fantastic.