Quivr

Quivr is an open-source, opinionated retrieval-augmented generation (RAG) framework hosted on GitHub (QuivrHQ/quivr), built to let developers integrate generative AI into their own applications without building a RAG pipeline from scratch. Users feed it files, including PDFs, text, and Markdown, with support for custom parsers, and query them conversationally through an LLM, effectively creating a personal or team second brain over their documents. Under the hood, Quivr's Brain engine supports multiple LLM providers, including OpenAI, Anthropic, Mistral, Gemma, and local models via Ollama, plus a choice of vector stores such as PGVector and Faiss. It offers configurable retrieval workflows, optional internet-search augmentation, and a Megaparse integration for more advanced file ingestion. The project is written primarily in Python with a React/TypeScript frontend, ships Docker-based deployment, and can be installed as a pip package with a minimal code example to stand up a working RAG system. Quivr is aimed at developers and technical teams who want a ready-made, customizable RAG backend rather than individual end users looking for a polished consumer chat app. It's backed by Y Combinator and Theodo and has built a large open-source following, with roughly 39,000 GitHub stars and an active commit history.