GraphRAG

As explained in Knowledge Graphs and LLMs in Action

GraphRAG is a retrieval-augmented generation approach that uses graph-structured knowledge to ground language-model outputs.

GraphRAG combines knowledge graphs with large language models to retrieve structured, relational context before generating an answer. It addresses vector-search failures such as noisy matches, missed entities, limited multihop reasoning, stale information, and weak transparency. A graph can combine curated facts, document text, metadata, and traceable source links, giving domain experts more control over the evidence used by the model. section 13.4.2 section 13.5

A practical GraphRAG pipeline can combine several retrieval paths:

• Represent entities, typed relationships, documents, and metadata in a knowledge graph. section 13.4.3
• Use the language model and graph schema to translate a natural-language question into a Cypher query. section 1.3 section 13.4.3
• Retrieve a connected subgraph, such as shortest paths between requested entities, when relationships provide the relevant context. section 13.4.3
• Use a graph-enhanced document retriever to select documents mentioning the required entities and preserve metadata and source traceability. section 13.4.3
• Give the retrieved graph and documents to the final model, then validate answers against the underlying evidence. section 13.4.5 section 1.3

Research-archive question answering

For “What did her fellow researchers say about Dorothy M. Wrinch?”, an agent first retrieves people connected to Wrinch, then finds the relevant source documents through a graph-enhanced retriever, and finally generates a factual answer from that context. Structured graph content can also answer questions such as shared research topics between Harvard University and Johns Hopkins University. section 13.4.5

Figure 13.5 Example of our Graph RAG agents’ internal steps. The agent starts with a thought that leads it to use the KG retriever to obtain a list of people who talked about Dorothy M. Wrinch. For each of them, it uses the KG-enhanced document retriever to identify the relevant documents. Finally, the agent uses all of this as a context for generating the final answer.

Shows the agent's thought, retrieval, observation, and answer-generation sequence for a grounded GraphRAG response. section 13.4.5

How Knowledge Graphs and LLMs in Action covers it

How deep it goes

Explains graph design, retrieval paths, agent orchestration, metadata, provenance, and implementation tooling rather than only defining the technique.

Examples

Uses a Lauritsen–cyclotron retrieval failure, Rockefeller Archive Center questions, graph retrievers, and a Neo4j implementation listing.

What it assumes

Assumes familiarity with language models, prompts, tools, information retrieval, and knowledge graphs while connecting them into an agent architecture.

What you can do after

Operate GraphRAG agents with KG retrievers and KG-enhanced document retrievers to produce more precise, grounded answers.

How it connects

Connects knowledge graphs, large language models, retrieval-augmented generation, vector search, Cypher, metadata, and provenance.

liveBookmanning.com

Read Knowledge Graphs and LLMs in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  The Art of AI Product DevelopmentA retrieval-augmented generation approach that uses graph-structured knowledge and entity relationships to retrieve contextual information.
• 

  liveBookmanning.com
  Essential GraphRAGGraphRAG is a retrieval-augmented generation approach that incorporates graph-structured knowledge into information retrieval and response generation.
• 

  liveBookmanning.com
  Domain-Specific Small Language ModelsA retrieval-augmented generation approach that uses knowledge graphs and community summaries to support multihop reasoning over document collections.
• 

  liveBookmanning.com
  AI Agents and ApplicationsGraphRAG is a retrieval-augmented generation pattern that uses graph-structured knowledge for retrieval, reasoning, and response generation.

Compare how 5 books explain GraphRAG →

Related concepts

• knowledge graphuses · A knowledge graph is a graph-structured representation of entities and their semantically meaningful relationships, often including associated attributes or facts.
• Leiden algorithmuses · The Leiden algorithm is a graph community-detection method that iteratively refines and aggregates partitions into well-connected communities while optimizing a quality function.
• retrieval-augmented generationkind of · Retrieval-augmented generation is a technique that retrieves relevant information from external sources and supplies it as context to a generative model when producing an output.

This explainer was generated by Manning's knowledge graph from the text of Knowledge Graphs and LLMs in Action. The section links point to the parts of the book it draws on.