LightRAG
Graph-based RAG with dual-level knowledge retrieval
About
RAG framework that builds a knowledge graph from documents, enabling retrieval at both local (specific facts) and global (thematic) levels. Outperforms naive RAG on complex questions requiring reasoning across multiple document sections. EMNLP 2025.
Choose LightRAG when…
- •Your queries require reasoning across multiple documents or topics
- •You want graph-based retrieval instead of flat vector search
- •You need both fact-level and concept-level retrieval in one system
Builder Slot
The pipeline layer that connects LLM calls, retrieval, and data processing into a workflow
Stack Genome Detection
AIchitect's Genome scanner detects LightRAG in your project via these signals:
lightrag-hkuIntegrates with (2)
LightRAG supports Qdrant as a vector backend for the embedding side of its hybrid graph-plus-vector index.
→ Run LightRAG's graph-RAG pipeline on Qdrant for the vector layer.
LightRAG uses OpenAI models (embeddings and chat) to build the knowledge graph and to answer queries on top of it.
→ Run graph-RAG with OpenAI as the model backbone without writing custom extraction code.
Alternatives to consider (1)
Pricing
Recent Activity
Crossed 25,000 stars ⭐
3 weeks ago
Crossed 10,000 stars ⭐
3 weeks ago
Crossed 5,000 stars ⭐
3 weeks ago
Crossed 1,000 stars ⭐
3 weeks ago
In 1 stack
Badge
Add to your GitHub README
[](https://www.aichitect.dev/tool/lightrag)Explore the full AI landscape
See how LightRAG fits into the bigger picture — browse all 207 tools and their relationships.