Farfalle is a self-hostable AI search engine that answers questions by combining web search with large language models. It supports running local LLMs via Ollama (llama3, mistral, gemma, phi3), cloud models through Groq and OpenAI, or any custom LLM via LiteLLM. The application uses a FastAPI backend with a Next.js frontend, integrating with search providers like SearXNG, Tavily, and Serper to retrieve source content before generating answers.