GraphRAG
A RAG variant that builds a knowledge graph of entities and relations from documents and retrieves along it, not only by text similarity.
A model reads the documents, extracts entities (people, companies, products) and the relations between them, and often writes summaries of related clusters. Questions that span many documents, such as “which suppliers are tied to delayed projects”, are answered from the graph, where plain RAG would find only scattered passages. The term was popularized by a Microsoft Research project published in 2024. The price is an expensive indexing step that has to be repeated as documents change.
Related terms
- RAG (retrieval-augmented generation): Finding relevant documents first and putting them into the prompt, so the model answers from them.
- Agentic RAG: RAG where the model decides what to search for, reads the results and searches again if needed, instead of one fixed retrieval step.
- Chunking (RAG): Splitting documents into smaller passages before indexing them for search, so retrieval returns the relevant part instead of a whole file.
- Hybrid search: Combining keyword search (such as BM25) with vector search, so results match both exact terms and meaning.
