GraphRAG
GraphRAG is a retrieval approach where an LLM first builds a knowledge graph from a document collection and writes summaries of related groups of entities, then answers questions using those summaries and graph links.
Also called: graph RAG
By Divya Sharma · Checked Nov 15, 2026 · How we review
Standard retrieval-augmented generation fetches a few passages similar to the question. That works for "what is our refund window?" but fails for "what are the main themes across these thousands of interviews?", because no single passage holds the answer. GraphRAG was proposed by Microsoft researchers in 2024 to handle these whole-corpus questions.
How it works
Indexing has two stages. An LLM reads every chunk and extracts entities and relationships, building a knowledge graph. The graph is then split into communities of closely linked entities, and the LLM writes a summary for each. To answer a global question, each relevant community summary produces a partial answer, and those are combined into a final response. For narrow questions, a local mode starts from matching entities and follows their neighbors.
The technique and the repo
"GraphRAG" names both the method and Microsoft's open-source implementation, reviewed at GraphRAG. Related projects such as LightRAG use similar ideas with a lighter indexing step.
Trade-offs
Indexing is expensive because an LLM processes the whole corpus, often more than once, and the graph must be rebuilt or updated when documents change. Extraction quality limits answer quality. For simple lookups, plain RAG with hybrid search is cheaper and usually just as good. The knowledge graphs guide covers tools.
Related terms
- Knowledge graph
- A knowledge graph stores information as entities (people, products, places, concepts) connected by named relationships, so software can follow links between facts instead of searching through loose text.
- Retrieval-augmented generation
- Retrieval-augmented generation (RAG) is a pattern where a system first searches a collection of documents for passages relevant to a question, then gives those passages to a language model so its answer draws on them.
- Agentic RAG
- Agentic RAG is retrieval-augmented generation run by an AI agent that decides when to search, what to search for and which source to use, and can search again if the first results fall short.
- Hybrid search
- Hybrid search runs a keyword search and a vector search on the same query and merges the two ranked lists, so results can match on exact words as well as on meaning.
- Chunking
- Chunking is the step that splits long documents into smaller passages before they are embedded and indexed, so a retrieval system can find and return only the parts of a document that answer a query.