quivr

This project addresses the challenge of continuously ingesting, versioning, searching, and alerting on evolving content such as news articles, emails, and PDFs. Unlike typical retriever-augmented generation (RAG) frameworks that focus primarily on integrating LLMs with vector stores, it provides a full pipeline with built-in version control, alerting on new matches, and support for multiple document sources including RSS feeds and Microsoft 365 mailboxes. It exposes a stable HTTP API with plugin-based architecture for custom file formats, embedding models, and alert logic, allowing engineering teams to integrate and extend it without building these components from scratch. The project is actively maintained with a recent commit in July 2025 and has amassed substantial interest with over 39,000 stars, but its growth signal is not yet established, indicating it may still be evolving. It is not suitable if you need a mature, widely adopted RAG framework with a large ecosystem, where alternatives like LangChain or LlamaIndex offer more integrations and community support.

Choose this project when you need a backend service that continuously collects and versions content from diverse sources while providing precise passage-level search and alerting capabilities via an HTTP API. For more mature or feature-rich RAG pipelines with broader LLM and vector store support, consider LangChain or LlamaIndex instead.

This is well suited for engineering teams building applications that require ongoing monitoring of changing document collections, such as internal knowledge bases that update with news, emails, or reports, and need to alert users when new relevant content arrives. Typical use cases include internal RAG search over corporate communications or external content monitoring with version history and alert notifications.

Latest commit: 2025-07-09. Activity data is based on daily RepoPi snapshots of the GitHub repository.