Product / UX / Orchestration & agents

LightRAG

Lightweight graph-based RAG framework from HKU's Data Science Lab that combines a knowledge graph with vector embeddings through dual-level retrieval. It drops GraphRAG's expensive community-report step and supports incremental updates, so new data can be added without rebuilding the global index. It ships naive, local, global and hybrid query modes, pluggable storage, an API server and a web UI.

Verified 2026-08-13 via the HKUDS/LightRAG repository.

Openness

5 high confidence

5.0

license: MIT(OSI)
source: public
core-gated: ungated

Fully MIT: the repo publishes the full retrieval stack including the API server and web UI, with no hosted tier, enterprise directory or license key, so the core is ungated as well as OSI-licensed.

https://github.com/HKUDS/LightRAG recorded 2026-08-13
MIT license, full public source including the API server and web UI, 38.8k stars and 5.5k forks; no hosted tier or enterprise directory

Adoption

3 medium confidence

3.0

313,096 PyPI downloads in the trailing 30 days for lightrag-hku, which bands at level 3 (100K-1M) on the software and model adoption scale, corroborated by a large star count.

https://pypistats.org/api/packages/lightrag-hku/recent recorded 2026-08-12
313,096 downloads in the trailing 30 days for lightrag-hku

Capability

4 medium confidence

4.0

Deeper and more efficient than GraphRAG for graph-RAG; narrower than general orchestrators.

https://github.com/HKUDS/LightRAG recorded 2026-08-13
dual-level retrieval, knowledge-graph construction, incremental updates, naive/local/global/hybrid query modes, pluggable storage, API server and web UI; no benchmark figure

Verified 2026-08-12