LightRAG

Why deploy LightRAG on Render?

LightRAG is a Retrieval-Augmented Generation (RAG) system that combines vector search with an automatically-extracted knowledge graph. It solves the limitation of standard RAG systems by understanding relationships between entities in your documents, not just matching text similarity. The server provides a REST API and web UI for document ingestion and querying with multiple retrieval modes.

This template gives you a fully wired LightRAG server with persistent storage already configured on a Render Disk, so your knowledge graph survives redeploys without any volume mounting or backup scripts to set up yourself. Instead of manually configuring the REST API, Web UI, and storage backends, you get one-click deploy that handles the Docker build, disk provisioning, and environment variable structure—just plug in your OpenAI key and a server secret. Render's persistent disk means you skip the complexity of external object storage while keeping your indexed documents intact across deploys.

Architecture

What you can build

After deploying, you'll have a running LightRAG server with a web UI where you can upload documents and query them using graph-aware retrieval—meaning the system understands relationships between entities (people, concepts, events), not just keyword matches. The server exposes a REST API for programmatic access and persists your knowledge graph to disk across restarts. You'll need to bring your own OpenAI API key for the LLM and embedding calls.

Key features

• Knowledge graph RAG: Combines vector store with automatically-extracted knowledge graph to understand entity relationships, not just text similarity.
• Multiple retrieval modes: Supports naive, local, global, hybrid, and mix retrieval modes for comparing vector-only vs graph-aware query strategies.
• Built-in Web UI: Includes interactive interface at /webui for uploading documents, visualizing the knowledge graph, and running queries.
• Persistent file storage: Uses JsonKVStorage, NanoVectorDBStorage, and NetworkXStorage on a 1GB Render Disk that survives restarts and redeploys.
• Public demo mode: DEMO=true flag enables read-only, no-auth access with rate limiting per IP for safely sharing pre-built knowledge bases.

Use cases

• Researcher queries uploaded papers to find entity relationships across studies
• Startup founder builds searchable knowledge base from company documentation
• Developer deploys read-only demo of product docs for customer self-service
• Consultant analyzes contract PDFs to surface hidden clause connections

What's included

Prerequisites

• OpenAI API Key (for LLM): Your OpenAI API key used for the language model that powers LightRAG's chat completions and entity extraction.
• OpenAI API Key (for Embeddings): Your OpenAI API key used for generating text embeddings (can be the same key as the LLM key).
• LightRAG API Key: A strong secret you create to protect access to your deployed LightRAG server and Web UI.

Next steps

1. Open the Web UI at /webui and log in with your LIGHTRAG_API_KEY — You should see the LightRAG dashboard with Documents, Knowledge Graph, and Retrieval tabs
2. Upload a test document like book.txt from Project Gutenberg and watch the Documents tab — The file should move from Pending to Processing to Processed, and the Knowledge Graph tab should populate with extracted entities and relationships
3. Test a query on the Retrieval tab using hybrid mode with 'What are the top themes in this story?' — You should receive a graph-aware response that references specific entities and relationships from your uploaded document

Resources

• LightRAG official site