LlamaCPP with GPU Support

[SerenityNrrd]

Allow us to use Custom Models like Gorilla or other llama2 models instead of OPENAI... or simply allow us to change the OpenAI base url

[neph1]

That is my goal with this fork: https://github.com/neph1/SuperAGI
It currently supports KoboldCpp (because that's what I use), but adding other backends should be trivial.
I haven't been able to keep it updated with all the changes from upstream, lately...

Here's the discussion:
#669