LightRAG based on HepAI and Open Embedding Model

Quick Start

1.Install Requirements

pip install -r requirements.txt

2.Setup environment variables

# .env file in lightrag_hepai/
Embedding_model_path="/path/to/embedding_model"
embedding_dimension=your_embedding_dimension
max_token_size=your_max_token_size
gpu_id=your_gpu_id
llm_api_key="your_api_key"
llm_base_url="your_base_url"
llm_model_name="openai/gpt-4o-mini"
working_dir="/path/to/working_dir"

3.Run LightRAG

python run_worker.py

4.Run LightRAG with Docker

4.1.Build the Docker Image

Run the following command in the root directory to build the Docker image:

docker build -t lightrag-backend:latest .

4.2.Run the Docker Container

Start a new container from the image:

docker run -d -p 4260:4260 --name lightrag-backend-container lightrag-backend:latest

4.3.Verify the Deployment

Check if the container is running:

docker ps

You should see an active container named lightrag-backend-container.

4.4.Access the Application

Open your browser and navigate to http://localhost:4260 to access the FastAPI backend.

• Swagger UI: http://localhost:4260/docs

5.Managing the Docker Container

5.1.Stop the Container

docker stop lightrag-backend-container

5.2.Start the Container Again

docker start lightrag-backend-container

5.3.Remove the Container

First, stop the container if it's running:

docker stop lightrag-backend-container

Then, remove it:

docker rm lightrag-backend-container

5.4.Remove the Docker Image

docker rmi lightrag-backend:latest

5.5.Additional Tips

• Logs: To view logs from the running container, use:

  docker logs -f lightrag-backend-container

• Interactive Shell: To access the container's shell:

  docker exec -it lightrag-backend-container /bin/bash

• Rebuilding the Image: If you make changes to the code or dependencies, rebuild the image:

  docker build -t lightrag-backend:latest .
  docker stop lightrag-backend-container
  docker rm lightrag-backend-container
  docker run -d -p 4260:4260 --name lightrag-backend-container lightrag-backend:latest

6.Testing LightRAG

python request_lightrag.py