LightRAG/docs

LightRAG Enterprise Documentation

Version 2.0.0-enterprise | Graph-Enhanced Retrieval-Augmented Generation

╔═══════════════════════════════════════════════════════════════════════════╗
║                                                                           ║
║     ██╗     ██╗ ██████╗ ██╗  ██╗████████╗██████╗  █████╗  ██████╗       ║
║     ██║     ██║██╔════╝ ██║  ██║╚══██╔══╝██╔══██╗██╔══██╗██╔════╝       ║
║     ██║     ██║██║  ███╗███████║   ██║   ██████╔╝███████║██║  ███╗      ║
║     ██║     ██║██║   ██║██╔══██║   ██║   ██╔══██╗██╔══██║██║   ██║      ║
║     ███████╗██║╚██████╔╝██║  ██║   ██║   ██║  ██║██║  ██║╚██████╔╝      ║
║     ╚══════╝╚═╝ ╚═════╝ ╚═╝  ╚═╝   ╚═╝   ╚═╝  ╚═╝╚═╝  ╚═╝ ╚═════╝       ║
║                                                                           ║
║              ENTERPRISE EDITION                                          ║
║         Simple and Fast Graph-Enhanced RAG System                        ║
║                                                                           ║
╚═══════════════════════════════════════════════════════════════════════════╝

🔱 Fork Project | This is an enterprise-ready fork of LightRAG by Raphaël MANSUY.

Goal: Create an enterprise-ready version of LightRAG with production-grade features.

First Enterprise Feature: ✅ Multi-tenancy support with RBAC, tenant isolation, and knowledge base management.

---

Documentation Overview

Document	Description
Quick Start	Get up and running in 5 minutes
Architecture Overview	System design, data flow, and core concepts
API Reference	Complete REST API documentation
Storage Backends	Configure KV, vector, and graph storage
LLM Integration	LLM providers and embedding models
Deployment Guide	Docker, Kubernetes, and production setup
Configuration Reference	All environment variables and options
Multi-Tenancy	Tenant isolation and RBAC

---

Quick Links

Getting Started

# Install
pip install lightrag-hku

# Start server
export OPENAI_API_KEY=sk-xxx
python -m lightrag.api.lightrag_server

Python Usage

from lightrag import LightRAG, QueryParam

rag = LightRAG(working_dir="./rag_storage")
await rag.ainsert("Your document text...")
result = await rag.aquery("Your question?", param=QueryParam(mode="hybrid"))

REST API

# Insert document
curl -X POST http://localhost:9621/documents/text \
  -H "Content-Type: application/json" \
  -d '{"text": "Document content..."}'

# Query
curl -X POST http://localhost:9621/query \
  -H "Content-Type: application/json" \
  -d '{"query": "Your question?", "mode": "hybrid"}'

Docker

docker run -p 9621:9621 -e OPENAI_API_KEY=sk-xxx ghcr.io/hkuds/lightrag:latest

---

System Overview

┌─────────────────────────────────────────────────────────────────────────────┐
│                              LightRAG                                        │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                              │
│   Documents ──▶ Chunking ──▶ Entity Extraction ──▶ Knowledge Graph          │
│                    │                                      │                  │
│                    ▼                                      ▼                  │
│              Embeddings ──────────────────────▶ Hybrid Retrieval            │
│                                                       │                      │
│                                                       ▼                      │
│                              Query ──▶ LLM Generation ──▶ Response          │
│                                                                              │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                              │
│  Storage Backends:                                                           │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐                       │
│  │ KV Storage   │  │ Vector Store │  │ Graph Store  │                       │
│  │ JSON/Redis/  │  │ NanoVectorDB │  │ NetworkX/    │                       │
│  │ PostgreSQL/  │  │ pgvector/    │  │ Neo4j/AGE/   │                       │
│  │ MongoDB      │  │ Milvus/FAISS │  │ Memgraph     │                       │
│  └──────────────┘  └──────────────┘  └──────────────┘                       │
│                                                                              │
│  LLM Providers:                                                              │
│  OpenAI │ Anthropic │ Ollama │ Azure │ Bedrock │ HuggingFace │ ...         │
│                                                                              │
└─────────────────────────────────────────────────────────────────────────────┘

---

Query Modes

Mode	Description	Best For
naive	Basic vector similarity search	Simple lookups
local	Entity-focused retrieval	Specific facts
global	High-level community summaries	Broad questions
hybrid	Combined local + global	Balanced queries
mix	Full KG + vector integration	Complex reasoning

---

Deployment Options

Option	Use Case	Guide
Local	Development	pip install lightrag-hku
Docker	Staging/Production	Docker Guide
Kubernetes	Production/Scale	K8s Guide

Storage Topology

Environment	KV	Vector	Graph
Development	JSON	NanoVectorDB	NetworkX
Production	PostgreSQL	pgvector	Neo4j
High-Scale	Redis	Milvus	Neo4j

---

Configuration Quick Reference

Essential Environment Variables

# LLM
LLM_BINDING=openai
LLM_MODEL=gpt-4o-mini
OPENAI_API_KEY=sk-xxx

# Embedding
EMBEDDING_BINDING=openai
EMBEDDING_MODEL=text-embedding-ada-002
EMBEDDING_DIM=1536

# Storage
LIGHTRAG_KV_STORAGE=JsonKVStorage
LIGHTRAG_VECTOR_STORAGE=NanoVectorDBStorage
LIGHTRAG_GRAPH_STORAGE=NetworkXStorage

# Server
PORT=9621

See Configuration Reference for all options.

---

Feature Highlights

Knowledge Graph Integration

┌────────────────────────────────────────────────────┐
│              KNOWLEDGE GRAPH                        │
│                                                     │
│    [Person: Einstein] ─────developed──────▶        │
│           │                               │        │
│     born_in                        [Theory:        │
│           │                        Relativity]     │
│           ▼                               │        │
│    [Location: Germany]           describes_│       │
│                                           ▼        │
│                                  [Concept:         │
│                                   Spacetime]       │
└────────────────────────────────────────────────────┘

Multi-Tenant Isolation

Tenant A          Tenant B          Tenant C
    │                 │                 │
    ▼                 ▼                 ▼
┌────────┐       ┌────────┐       ┌────────┐
│ KB-1   │       │ KB-1   │       │ KB-1   │
│ KB-2   │       │ KB-2   │       └────────┘
│ KB-3   │       └────────┘
└────────┘

---

Resources

Enterprise Fork

• GitHub (Enterprise): https://github.com/raphaelmansuy/LightRAG
• Author: Raphaël MANSUY - Elitizon
• Issues: https://github.com/raphaelmansuy/LightRAG/issues

Original Project

• GitHub (Original): https://github.com/HKUDS/LightRAG
• PyPI: https://pypi.org/project/lightrag-hku/

---

Document Index

1. Quick Start Guide

   • Installation options
   • Python SDK basics
   • REST API basics
   • Common patterns

2. Architecture Overview

   • System design diagrams
   • Core concepts (entities, relations, chunks)
   • Data flow pipeline
   • Query execution flow

3. API Reference

   • Document endpoints
   • Query endpoints
   • Graph endpoints
   • Admin endpoints

4. Storage Backends

   • KV storage options
   • Vector storage options
   • Graph storage options
   • Configuration tables

5. LLM Integration

   • Provider configurations
   • Embedding models
   • Reranking options
   • Custom implementations

6. Deployment Guide

   • Local development
   • Docker deployment
   • Kubernetes/Helm
   • Production best practices

7. Configuration Reference

   • Environment variables
   • CLI arguments
   • QueryParam options
   • Complete .env example

8. Multi-Tenancy Guide

   • Tenant isolation
   • RBAC roles/permissions
   • TenantRAGManager
   • API endpoints

---

Original LightRAG built with ❤️ by HKUDS

Enterprise Edition maintained by Raphaël MANSUY @ Elitizon