Being able to use a different LLM would be good. Ollama, GPT, Claude, any of those really. They'd be quicker with responses and Ollama can be ran locally.
ya we're wroking on local stuff for the next update, the issue is local models are noticably worse in quality and are big downloads.
