Skip to main content

Docker Compose

This deployment method is one step above local development in terms of sophistication. It’s suitable for a development server or simple production environments.

This snippet will give you a quick idea about how to deploy the frontend and backend containers so they play nicely together:

version: "3.8"

volumes:
  singulatron-data:
    name: singulatron-data
    driver: local

services:
  singulatron-frontend:
    image: crufter/singulatron-frontend:latest
    ports:
      - "3901:80"
    environment:
      # The `BACKEND_ADDRESS` must be accessible from the   browser.
      # It is not an internal address, it's the address the   browser will make API requests to.
      - BACKEND_ADDRESS=http://127.0.0.1:58231

  singulatron-backend:
    image: crufter/singulatron-backend:latest
    ports:
      - "58231:58231"
    volumes:
      # We mount the docker socket so the backend can start   containers
      - /var/run/docker.sock:/var/run/docker.sock
      # We mount a volume so data will be persisted
      - singulatron-data:/root/.singulatron
    environment:
      # This volume will be mounted by the LLM containers to access the models downloaded by Singulatron.
      - SINGULATRON_VOLUME_NAME=singulatron-data
      #
      # GPU Acceleration for NVIDIA GPUs
      # Uncomment this envar for NVIDIA GPUs.
      #
      # - SINGULATRON_GPU_PLATFORM=cuda

Put the above into a file called docker-compose.yaml in a folder on your computer and run it with the following command:

docker compose up

Once it's running​

After the containers successfully start, you can go to 127.0.0.1:3901 and log in with the Default Credentials.

Configuring​

See the Backend Environment Variables and Frontend Environment Variables.