Nice and neat tool! I wonder which llm model is running in the backend and is there any way to run it locally/self-hosted?