Skill Detail

Build graph RAG context with Neo4j LLM Graph Builder

Convert a bounded document set into a Neo4j knowledge graph, inspect extracted nodes and relationships, and use it for graph-backed RAG.

Data Extraction & TransformationMulti-Framework
Data Extraction & Transformation Multi-Framework Security Reviewed
⭐ 4.7k GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill build-graph-rag-context-with-neo4j-llm-graph-builder Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Neo4j Database 5.23 or later with APOC, Python 3.12 backend, configured LLM provider credentials, optional front-end and Docker deployment paths from the upstream README.
Install & setup
Create backend/.env from backend/example.env and set NEO4J_URI, NEO4J_USERNAME, NEO4J_PASSWORD, and NEO4J_DATABASE. From backend, create a Python 3.12 virtual environment, install requirements with pip install -r requirements.txt -c constraints.txt, then run uvicorn score:app –reload. Follow the README for front-end and Docker-specific deployment paths.
Author
Neo4j Labs
Publisher
Open Source Organization
Last updated
Jun 4, 2026
Quick brief

Use Neo4j LLM Graph Builder when an agent workflow needs structured graph context from unstructured documents before retrieval, question answering, or analysis. The operator connects a Neo4j database, configures credentials and model settings, loads a bounded corpus, extracts entities and relationships, reviews the generated graph, and uses the graph or chat interface as a source for downstream RAG work. Invoke this instead of using a generic document parser when the important output is an inspectable knowledge graph with nodes, relationships, schema choices, and source metadata. The operator should validate the selected schema, spot-check extracted relationships, and keep source documents traceable before handing the graph to an agent. Start with a small corpus when cost, latency, or extraction quality is unknown. The scope boundary is graph construction and review for a controlled corpus. It is not a generic Neo4j listing, a replacement for data modeling, or an unsupervised ingestion pipeline for arbitrary private documents.