Support for LLMs with a local backend

[nucleartux]

ollama for example

[nucleartux]

I implemented support for Ollama, but now it works only with llama3.1 model.
I think of implementing support for choosing between multiple models, but not sure about the UX right now.