actually we just released support for Ollama so it's possible to use local models now :)