What I'd really like is a non-intrusive LLM running that I can put on a third monitor which sees what I'm working on, while I work, and tries to mirror and make suggestions WITHOUT interjecting itself into my editor. Then I just want a mappable prompt command (like <leader>P) where I can provide feedback / suggestions (or maybe I just do that via mic).

This way I can choose to just ignore it for the most part, but if I see something interesting, I can refine it, prompt, cherry pick, etc. I really don't want autocomplete for anything.