LightRAG/docs/DockerDeployment.md

LightRAG Docker Deployment

A lightweight Knowledge Graph Retrieval-Augmented Generation system with multiple LLM backend support.

🚀 Preparation

Clone the repository:

# Linux/MacOS
git clone https://github.com/HKUDS/LightRAG.git
cd LightRAG

# Windows PowerShell
git clone https://github.com/HKUDS/LightRAG.git
cd LightRAG

Configure your environment:

# Linux/MacOS
cp .env.example .env
# Edit .env with your preferred configuration

# Windows PowerShell
Copy-Item .env.example .env
# Edit .env with your preferred configuration

LightRAG can be configured using environment variables in the .env file:

Server Configuration

• HOST: Server host (default: 0.0.0.0)
• PORT: Server port (default: 9621)

LLM Configuration

• LLM_BINDING: LLM backend to use (lollms/ollama/openai)
• LLM_BINDING_HOST: LLM server host URL
• LLM_MODEL: Model name to use

Embedding Configuration

• EMBEDDING_BINDING: Embedding backend (lollms/ollama/openai)
• EMBEDDING_BINDING_HOST: Embedding server host URL
• EMBEDDING_MODEL: Embedding model name

RAG Configuration

• MAX_ASYNC: Maximum async operations
• MAX_TOKENS: Maximum token size
• EMBEDDING_DIM: Embedding dimensions

🐳 Docker Deployment

Docker instructions work the same on all platforms with Docker Desktop installed.

Start LightRAG server:

docker-compose up -d

LightRAG Server uses the following paths for data storage:

data/
├── rag_storage/    # RAG data persistence
└── inputs/         # Input documents

Updates

To update the Docker container:

docker-compose pull
docker-compose down
docker-compose up

Offline deployment

Software packages requiring transformers, torch, or cuda will is not preinstalled in the dokcer images. Consequently, document extraction tools such as Docling, as well as local LLM models like Hugging Face and LMDeploy, can not be used in an off line enviroment. These high-compute-resource-demanding services should not be integrated into LightRAG. Docling will be decoupled and deployed as a standalone service.

📦 Build Multi-Architecture Docker Images

Prerequisites

Before building multi-architecture images, ensure you have:

• Docker 20.10+ with Buildx support
• Sufficient disk space (20GB+ recommended for offline image)
• Registry access credentials (if pushing images)

1. Setup Buildx Builder

Create and configure a multi-architecture builder:

# Create a new buildx builder instance
docker buildx create --name multiarch-builder --use

# Start and verify the builder
docker buildx inspect --bootstrap

# Verify supported platforms
docker buildx inspect multiarch-builder

You should see support for linux/amd64 and linux/arm64 in the output.

2. Registry Authentication

For GitHub Container Registry (ghcr.io)

Option 1: Using Personal Access Token

1. Create a GitHub Personal Access Token:

   • Go to GitHub Settings → Developer settings → Personal access tokens → Tokens (classic)
   • Generate new token with write:packages and read:packages permissions
   • Copy the token

2. Login to registry:

   echo "YOUR_GITHUB_TOKEN" | docker login ghcr.io -u YOUR_GITHUB_USERNAME --password-stdin
   

Option 2: Using GitHub CLI

gh auth token | docker login ghcr.io -u YOUR_GITHUB_USERNAME --password-stdin

For Docker Hub

docker login
# Enter your Docker Hub username and password

For Other Registries

docker login your-registry.example.com
# Enter your credentials

3. Build Commands

A. Local Build (No Push)

Build multi-architecture images locally without pushing to registry:

Normal image:

docker buildx build \
  --platform linux/amd64,linux/arm64 \
  --file Dockerfile \
  --tag ghcr.io/hkuds/lightrag:latest \
  --load \
  .

Lite image:

docker buildx build \
  --platform linux/amd64,linux/arm64 \
  --file Dockerfile.lite \
  --tag ghcr.io/hkuds/lightrag:lite \
  --load \
  .

Note

: --load loads the image to local Docker, but only supports single platform. For multi-platform, use --push instead.

B. Build and Push to Registry

Build and directly push to container registry:

Normal image:

docker buildx build \
  --platform linux/amd64,linux/arm64 \
  --file Dockerfile \
  --tag ghcr.io/hkuds/lightrag:latest \
  --push \
  .

Lite image:

docker buildx build \
  --platform linux/amd64,linux/arm64 \
  --file Dockerfile.lite \
  --tag ghcr.io/hkuds/lightrag:lite \
  --push \
  .

C. Build with Multiple Tags

Add version tags alongside latest:

# Get version from git tag
VERSION=$(git describe --tags --abbrev=0 2>/dev/null || echo "v1.0.0")

# Build with multiple tags
docker buildx build \
  --platform linux/amd64,linux/arm64 \
  --file Dockerfile \
  --tag ghcr.io/hkuds/lightrag:latest \
  --tag ghcr.io/hkuds/lightrag:${VERSION} \
  --push \
  .

4. Verify Built Images

After building, verify the multi-architecture manifest:

# Inspect image manifest
docker buildx imagetools inspect ghcr.io/hkuds/lightrag:latest

# Expected output shows multiple platforms:
# Name:      ghcr.io/hkuds/lightrag:offline
# MediaType: application/vnd.docker.distribution.manifest.list.v2+json
# Platforms: linux/amd64, linux/arm64

5. Troubleshooting

Build Time is Very Slow

Cause: Building ARM64 on AMD64 (or vice versa) requires QEMU emulation, which is slower.

Solutions:

• Use remote cache (--cache-from/--cache-to) for faster subsequent builds
• Build on native architecture when possible
• Be patient - initial multi-arch builds take 30-60 minutes

"No space left on device" Error

Cause: Insufficient disk space for build layers and cache.

Solutions:

# Clean up Docker system
docker system prune -a

# Clean up buildx cache
docker buildx prune

# Check disk space
df -h

"failed to solve: failed to push" Error

Cause: Not logged into the registry or insufficient permissions.

Solutions:

1. Verify you're logged in: docker login ghcr.io
2. Check you have push permissions to the repository
3. Verify the image name matches your repository path

Builder Not Found

Cause: Buildx builder not created or not set as current.

Solutions:

# List builders
docker buildx ls

# Create and use new builder
docker buildx create --name multiarch-builder --use

# Or switch to existing builder
docker buildx use multiarch-builder

6. Cleanup

Remove builder when done:

# Switch back to default builder
docker buildx use default

# Remove multiarch builder
docker buildx rm multiarch-builder

# Prune build cache
docker buildx prune

7. Best Practices

1. Use specific tags: Avoid only using latest, include version tags
2. Verify platforms: Always check the manifest after pushing
3. Monitor resources: Ensure sufficient disk space before building
4. Test both architectures: Pull and test each platform variant
5. Use .dockerignore: Exclude unnecessary files to speed up build context transfer