// the find
The-Vibe-Company/quivr
Opiniated RAG for integrating GenAI in your apps 🧠 Focus on your product rather than the RAG. Easy integration in existing products with customisation! Any LLM: GPT4, Groq, Llama. Any Vectorstore: PGVector, Faiss. Any Files. Anyway you want.
Quivr is a Python library (installable as quivr-core) for wiring up RAG over your own files with a handful of lines of code — pick an LLM (OpenAI, Anthropic, Mistral, local via Ollama), pick a vectorstore (PGVector, FAISS), point it at files, and ask questions. It's aimed at developers who want RAG bolted into an existing product without building the retrieval pipeline themselves.
The Brain.from_files / brain.ask API genuinely is a 5-line RAG setup, and the YAML-driven workflow config (node graph: filter_history -> rewrite -> retrieve -> generate) lets you tweak the pipeline — add a reranker, change max_history, swap the reranker supplier — without touching the graph code. It's not locked into one LLM or vectorstore vendor, which is the main thing that kills most 'quick RAG' libraries in production. File parsing is offloaded to a separate Megaparse project, so PDF/docx/epub/odt handling isn't reinvented here, and there's an actual per-format test suite (pdf, docx, epub, odt, tika) rather than one smoke test.
The README is written for the quivr.com hosted product (Discord badges, partner logos, 'Second Brain' branding) but the repo itself is just the core library — there's a real mismatch between the 39k stars/marketing framing and what you actually get installing quivr-core. The 'opinionated' pipeline is a fixed node graph (filter_history/rewrite/retrieve/generate); if your retrieval flow doesn't fit that shape, you're fighting the framework rather than configuring it. Docs beyond the one basic_rag example are thin — reranker config, tool use, and the langgraph-based RetrievalConfig aren't well explained outside source code. Pulling in Megaparse/Tika/unstructured for ingestion means a fair amount of transitive dependency weight for what's pitched as a lightweight 'add RAG to your app' library.