Farfalle is an open-source AI-powered search engine that runs local LLMs or cloud models via a Next.js + FastAPI stack, with Docker-based setup and multiple external search providers.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Farfalle is an open-source AI-powered search engine that can run local LLMs (llama3, gemma, mistral, phi3) via Ollama, or use cloud models (OpenAI/gpt4-o) and other providers. It supports multiple search backends and a planning agent for improved results.
How it works
Frontend: Next.js. Backend: FastAPI. Search API integrates SearXNG, Tavily, Serper, and Bing. It can query local models through Ollama or custom LLMs via LiteLLM. It includes a chat history and expert search features.
Getting started
Prerequisites: Docker, Ollama (if running local models). Start Ollama server with a local model like llama3, mistral, gemma, or phi3. Quick Start commands:
git clone https://github.com/rashadphz/farfalle.git
cd farfalle && cp .env-template .env
Modify .env with your API keys (optional, not required if using Ollama).
docker-compose -f docker-compose.dev.yaml up -d
Visit http://localhost:3000 after the app starts. For custom setup, see custom-setup-instructions.md.
Getting started (API keys and deploy)
Obtain API keys from Tavily, Serper, OpenAI, Bing, and Groq if using external services. Deploy options include Render for backend and Vercel for frontend, with instructions to set NEXT_PUBLIC_API_URL to the backend URL when deploying.
Recent releases
Latest releases: none.
Traction
Stars: 3538 Forks: 320 Open issues: 53
Behind the repo
Not provided in the README excerpts.
Caveats
License: Apache-2.0. Created: 2024-04-25. Last push: 2024-09-27. Language: TypeScript. The project uses Docker-based deployment and supports local and cloud LLMs, with features like Chat History and Expert Search.