neo4j/neo4j-graphrag-python
Neo4j GraphRAG for Python
What it solves
Standard Retrieval Augmented Generation (RAG) often relies solely on vector similarity, which can miss complex relationships between entities. This package enables developers to implement GraphRAG, combining the structured relational power of Neo4j knowledge graphs with the generative capabilities of LLMs to provide more accurate and context-aware responses.
How it works
The library provides tools to construct knowledge graphs from unstructured text or PDFs using LLM-driven pipelines. It supports various retrieval strategies, including vector similarity searches, graph traversals, and hybrid approaches that combine both. It integrates with major LLM providers (like OpenAI, Anthropic, and MistralAI) and vector databases (like Weaviate, Pinecone, and Qdrant) to facilitate end-to-end RAG workflows.
Who it’s for
Python developers building advanced GenAI applications who want to leverage structured knowledge graphs to enhance the accuracy and reasoning of their LLM-based systems.
Highlights
- Official first-party library from Neo4j with long-term support
- Automated knowledge graph construction pipelines from text and PDFs
- Support for hybrid retrieval combining vector search and graph traversal
- Extensive integration with popular LLM providers and vector databases