Skip to main content

LightRAG

Graph-based RAG framework (HKUDS) that extracts entities and relations from documents via LLM, builds a persistent knowledge graph, and supports multiple retrieval modes (naive, local, global, hybrid, mix). Used in this wiki as the graph indexing layer alongside qmd. See Wiki Indexing Pipeline: qmd and LightRAG for operational details and Local Wiki RAG: LightRAG Graph Stack for the two-path stack. Stub — expand when a dedicated source is ingested.

Linked pages