Integrate local LLM into starship, to be able to run them easily on command. Maybe have it open to multiple types of models as well.