LightRAG
LightRAG knowledge-graph RAG server with web UI, backed by Postgres and Qdrant, on your own server.
LightRAG
LightRAG (HKU Data Intelligence Lab, MIT license) is a lightweight knowledge-graph RAG framework. It indexes your documents into a knowledge graph plus vector embeddings and answers questions over both.
What you get
- LightRAG v1.5.7 API server and web UI on port 9621
- Postgres 16 for key-value, document status and graph storage
- Qdrant for vector storage
- Health checks on every service and persistent volumes for data
Security
Upstream LightRAG runs with no authentication by default. This stack turns it on:
the web UI needs a login (user admin), the API needs an X-API-Key header, and only
/health is public. The admin password, API key, JWT secret and Postgres password are
generated fresh for every install.
What you provide
An OpenAI-compatible endpoint for the LLM and the embedding model (OpenAI by default, or any compatible provider): set the API keys, model names and embedding dimension at install.
Supported Clouds
