LightRAG

Why we included this project

LightRAG grounds an LLM in your own documents by building a lightweight knowledge graph from the text and pairing graph-based retrieval with vector search. That combination pays off when a question spans several documents or asks how two facts relate, where plain chunk retrieval tends to come up short. It installs as a Python library and also ships a server with a web UI, and the storage layer is pluggable (Neo4j, PostgreSQL, and others) with swappable LLM and embedding providers, so it drops into an existing stack without much rework. Teams building internal knowledge assistants, Q&A over PDFs and Office files, or research copilots over a private corpus will find the setup path short, and the active community helps when you get stuck.