Local search

As explained in Essential GraphRAG

Local search is a GraphRAG retrieval method that combines information from entities and their nearby graph structure with related source-text passages.

Local search answers entity-focused queries by combining semantically relevant graph entities with connected relationships, community summaries, and source-text chunks. Vector search identifies entry-point entities, while graph traversal gathers surrounding structured and unstructured context. section 7.3.2 section 7.3

A practical local-search pipeline:

• Embed entity summaries and store the vectors in a database vector index. section 7.3.2
• Embed the user query and retrieve the most similar entities as entry points. section 7.3.2
• Expand those entities through connected chunks, communities, relationships, and entity summaries. section 7.3.2
• Rank and limit the retrieved context before passing it to an LLM for the final response. section 7.3.2

Entity-focused questions

For “What are the healing properties of chamomile?”, local search retrieves the chamomile entity, related graph information, community summaries, and associated source chunks. A runnable example asks, “Who is Ulysses?” section 7.3.2

Figure 7.5 Local search

Use the local-search diagram to follow the flow from vector-selected entities through graph and text retrieval to answer generation. section 7.3.2

How Essential GraphRAG covers it

How deep it goes

Implements embeddings, vector indexing, Cypher expansion, context limits, and LLM generation, while contrasting local with global retrieval.

Examples

Uses chamomile and Ulysses questions, entity embeddings, a Neo4j vector index, and Cypher retrieval statements.

What it assumes

Assumes a completed graph-indexing process with entities, relationships, communities, summaries, and source chunks available.

What you can do after

Build it from scratch by embedding entities, querying a vector index, expanding graph context, and generating an LLM response.

How it connects

Connects vector search, knowledge graphs, graph traversal, community summaries, source chunks, embeddings, and global search.

liveBookmanning.com

Read Essential GraphRAG on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Advanced Algorithms and Data StructuresLocal search is an optimization technique that iteratively improves a candidate solution through nearby modifications.
• 

  liveBookmanning.com
  Optimization Algorithms: AI techniques for design, planning, and control problemsA local search algorithm iteratively moves from a feasible solution to a neighboring solution according to an improvement rule.

Compare how 3 books explain local search →

Related concepts

• tabu searchused by · Tabu search is a memory-guided local search technique that restricts recently used moves or solutions to discourage cycling and escape local optima.
• hillclimbingincludes · Hill climbing is an iterative local-search technique that moves from a candidate solution to a neighboring solution with a better objective value.
• knowledge graphuses · A knowledge graph is a graph-structured representation of entities and their semantically meaningful relationships, often including associated attributes or facts.
• global searchcontrasts with · A search strategy that explores broadly across the search space to seek a global optimum rather than concentrating only near one current solution.
• global searchcontrasts with · Global search is a GraphRAG retrieval technique that answers broad queries by aggregating and synthesizing information across community summaries or other corpus-level representations.

This explainer was generated by Manning's knowledge graph from the text of Essential GraphRAG. The section links point to the parts of the book it draws on.