gusye1234/nano-graphrag

Health v2 · maintenance only

Adoption not part of the score

3974 stars · 424 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

nano-graphrag is a lightweight, ~1100-line Python implementation of Microsoft's GraphRAG, designed to be small, fast, and easy to read or hack. It supports pluggable backends (faiss, neo4j, ollama), async operation, and multiple LLM providers including OpenAI, Azure OpenAI, and Amazon Bedrock.

Use cases

• build a graph-based RAG pipeline over my documents
• understand how GraphRAG works by reading a small codebase
• run GraphRAG locally with ollama instead of OpenAI
• customize a GraphRAG implementation for my own research
• extract knowledge graphs from text and query them with an LLM
• index a corpus and answer questions using graph-based retrieval

When to choose

• you want a minimal, hackable GraphRAG rather than the heavy official Microsoft implementation
• you need to modify or extend GraphRAG internals for research or prototyping
• you want to swap in your own LLM, embedding, or vector store backends

When to avoid

• you need a production-grade, fully supported GraphRAG with enterprise features
• you want a turnkey application with a UI rather than a Python library
• you need multi-user long-term memory RAG (the authors point to memobase instead)

Facets

library · maturity active

rag llm-inference search-engine nlp large-language-models artificial-intelligence python cross-platform graphrag knowledge-graph graph-based-rag lightweight easy-to-hack async retrieval-augmented-generation natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="gusye1234/nano-graphrag")