# GraphRAG

> GraphRAG is a retrieval-augmented generation approach that uses a language model to extract an entity-and-relationship graph from a document collection, summarizes communities of related entities, and answers questions from those summaries.

- Identifier: PTL-0133
- Category: Retrieval & Tool Use
- Canonical URL: https://protologue.com/t/graphrag/
- Also known as: Graph RAG, knowledge-graph RAG
- Introduced: 2024

## Description

Standard RAG retrieves a handful of passages, which works for questions answered in one place but fails on global questions about a whole corpus, such as its main themes. GraphRAG answers these by combining partial answers from many community summaries. Building the graph requires many LLM calls, so indexing is far more expensive than for embedding-based RAG.

## Broader terms

- [Retrieval-Augmented Generation](https://protologue.com/t/retrieval-augmented-generation/)

## Related terms

- [Agentic RAG](https://protologue.com/t/agentic-rag/)

## Sources

- Edge et al. (2024). From Local to Global: A Graph RAG Approach to Query-Focused Summarization. https://arxiv.org/abs/2404.16130

## Cite this entry

Protologue. (2026). GraphRAG. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0133). https://protologue.com/t/graphrag/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)