nano-graphrag
gusye1234Small, hackable reimplementation of GraphRAG in roughly eleven hundred lines, written because the official implementation is hard to read or modify. It is portable, fully asynchronous and typed, with swappable graph, vector and model backends, and is meant as a reference and customization base rather than a production system.
Openness
5 high confidence- license
- MIT(OSI)
- source
- public
- core features withheld
- no
Fully MIT: the entire reimplementation is published, with no hosted tier, enterprise directory or license key, so the core is ungated as well as OSI-licensed.
- https://github.com/gusye1234/nano-graphrag recorded 2026-08-13
MIT license, full public source for the GraphRAG reimplementation, 4.0k stars and 424 forks; no hosted tier or enterprise directory
- https://raw.githubusercontent.com/gusye1234/nano-graphrag/main/LICENSE recorded 2026-09-29
The LICENSE file is the standard MIT text: "MIT License" and "Permission is hereby granted, free of charge, to any person obtaining a copy" with no added restrictions.
- https://raw.githubusercontent.com/gusye1234/nano-graphrag/main/readme.md recorded 2026-09-29
The readme describes a fully public codebase: "nano-graphrag" is "about **1100 lines of code**" and is "Small yet [**portable**](#Components)(faiss, neo4j, ollama...)"; searched the body for enterprise, license key and hosted tier: none.
Adoption
1 medium confidenceAdoption is measured by PyPI downloads of nano-graphrag over the trailing month, a first-party count that is more reliable here than the star total for a small reimplementation like this one. With no release in more than a year, that figure has been declining.
- https://pypistats.org/api/packages/nano-graphrag/recent recorded 2026-08-12
1,334 downloads in the trailing 30 days for nano-graphrag
Capability
3 medium confidenceNano-graphrag offers a small, readable core with global and local queries, community detection, pluggable stores and multiple LLM providers. It has less breadth and polish than GraphRAG, the fuller implementation it reimplements, and development has stalled.
- https://github.com/gusye1234/nano-graphrag recorded 2026-08-13
'a simple, easy-to-hack GraphRAG implementation'; global and local queries, community detection, pluggable stores, multi-provider LLMs; 155 commits, no benchmark figure
- https://raw.githubusercontent.com/gusye1234/nano-graphrag/main/readme.md recorded 2026-09-29
The readme documents "# Perform global graphrag search", "# Perform local graphrag search", "supports incremental insert", a built-in Neo4jStorage and DeepSeek and ollama LLM examples; it says the code is "about **1100 lines of code**".
Verified 2026-09-24