Skip to main content

Docker

hunt's image, ghcr.io/arabesquelabs/hunt, is small: no shell, and it runs as a non-root user. It listens on :4848, keeps its SQLite database in /data (a volume), and answers /healthz once its database does.

Requirements​

  • Docker
  • Docker Compose, to run it from a compose file (optional)

1. Run it​

docker run -d --name hunt --restart unless-stopped \
  -p 127.0.0.1:4848:4848 -v hunt-data:/data \
  ghcr.io/arabesquelabs/hunt:latest

Or with Docker Compose: save this as compose.yaml, and run docker compose up -d beside it.

services:
  hunt:
    image: ghcr.io/arabesquelabs/hunt:latest
    ports: ["127.0.0.1:4848:4848"]
    volumes: ["hunt-data:/data"]
    restart: unless-stopped
volumes:
  hunt-data:

Your data is in the hunt-data volume, and stays when the container is replaced.

2. Open it​

Visit http://localhost:4848. 127.0.0.1:4848 keeps hunt to this computer. Before opening it to a network (-p 4848:4848), give it a password (-e HUNT_PASSWORD=…), or put a sign-in in front of it: see "What's public, and what needs signing in".

3. Choose a version (optional)​

latest is main's newest build. Each release is also tagged with its version (v0.15.0), and each build with its commit (sha-…): use one of those to choose when you upgrade.

4. Use PostgreSQL or MySQL (optional)​

Give the container the database's address as HUNT_DB (-e HUNT_DB=postgres://hunt:secret@db/hunt): see "The database".

Updating​

docker pull ghcr.io/arabesquelabs/hunt:latest
docker rm -f hunt        # the data stays, in its volume

Then run it again, as in step 1. With Compose: docker compose pull && docker compose up -d. A newer hunt upgrades a SQLite database in place.

Building the image yourself​

docker build --platform linux/amd64 -t hunt .
docker run -p 127.0.0.1:4848:4848 -v hunt-data:/data hunt

The image workflow (.github/workflows/image.yml) runs make ci, then publishes ghcr.io/arabesquelabs/hunt (latest from main, v<version>, and each commit's sha), so you needn't build it yourself.

Builds reuse what they can. The Dockerfile builds hunt's dependencies in a layer of their own, from cargo-chef's recipe (what hunt depends on, without its own version), and the workflow keeps layers in GitHub's Actions cache. A code change or a release (a new version) only compiles hunt itself; changing the dependencies compiles them again. The cache holds 10 GB per repository, so an occasional full build after old entries are evicted is expected.