Skip to main content

Docker Setup & Containerization

Running GoingMerry inside a Docker container allows you to execute local model workloads in a sandbox environment, ensuring isolation and clean resource boundaries.

1. Running on CPU​

If your container boundary does not require GPU resources, run the base image with volume mappings:

docker run -d -v merry:/root/.merry -p 11434:11434 --name merry dopove/merry

2. Running on NVIDIA GPUs​

To offload calculations to NVIDIA graphics hardware, you must install the NVIDIA Container Toolkit on the host machine.

Execute the container with GPU access:

docker run -d --gpus=all -v merry:/root/.merry -p 11434:11434 --name merry dopove/merry

3. Running on AMD GPUs (ROCm)​

For AMD card setups, map the appropriate device interfaces into the container container:

docker run -d --device /dev/kfd --device /dev/dri -v merry:/root/.merry -p 11434:11434 --name merry dopove/merry:rocm

4. Docker Compose Setup​

For unified multi-service deployments, create a docker-compose.yml file:

version: '3.8'

services:
  merry:
    image: dopove/merry:latest
    container_name: merry_daemon
    ports:
      - "11434:11434"
    volumes:
      - merry_storage:/root/.merry
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

volumes:
  merry_storage:

Launch the stack:

docker compose up -d