sailfork / t3code-github / .repos/ alchemy-effect/ website/ src/ content/ docs/ docker/ index.mdx

The Docker provider manages images, containers, networks, and volumes by shelling out to the docker CLI's active context — Docker Desktop, a remote or SSH context, a CI daemon. There is no daemon API client and no credentials of its own, and the resources live in the same Stack as your cloud resources. It is separate from Cloudflare.Container; registry image references are the boundary between Docker-managed images and cloud container platforms.

New here? Set up Docker first.

Resources

RemoteImage, Network, Volume, and Container compose into a local service — here, Postgres with a named volume, a network alias, a published port, and a healthcheck:

export default Alchemy.Stack(
  "DockerPostgresExample",
  {
    providers: Layer.merge(Docker.providers(), Alchemy.RandomProvider()),
    state: Alchemy.localState(),
  },
  Effect.gen(function* () {
    const password = yield* Alchemy.makeRandom("PostgresPassword", {
      bytes: 16,
    });

    const image = yield* Docker.RemoteImage("postgres-image", {
      name: "postgres",
      tag: "18-alpine",
      alwaysPull: false,
    });
    const network = yield* Docker.Network("app-network");
    const data = yield* Docker.Volume("postgres-data");

    const postgres = yield* Docker.Container("postgres", {
      name: "alchemy-example-postgres",
      image,
      environment: {
        POSTGRES_DB: "app",
        POSTGRES_USER: "alchemy",
        POSTGRES_PASSWORD: password,
      },
      ports: [{ external: 15432, internal: 5432 }],
      volumes: [
        { hostPath: data.name, containerPath: "/var/lib/postgresql/data" },
      ],
      networks: [{ name: network.name, aliases: ["postgres"] }],
      healthcheck: {
        cmd: ["CMD-SHELL", "pg_isready -U alchemy -d app"],
        interval: "5 seconds",
        timeout: "5 seconds",
        retries: 10,
      },
      start: true,
    });
    return { container: postgres.name, hostPort: postgres.ports };
  }),
);

Container accepts the image resource directly, alwaysPull: false pins the pulled tag so re-deploys are no-ops, and environment values accept Redacted secrets — they are passed via process env, never on the CLI. Run local services walks through this composition.

Image builds from a Dockerfile:

const image = yield* Docker.Image("app", {
  name: "my-app",
  tag: "latest",
  build: {
    context: "./app",
    dockerfile: "Dockerfile",
    args: { NODE_ENV: "production" },
  },
});

The build is the diff: diff runs docker build and compares the resulting imageId against the last deploy, and the result is memoized so plan + deploy build once — see Build and push an image for tagging and pushing to a registry.

Docker.inspectContainer(name) reads live runtime details for any container by name, including the bound host ports.

:::caution Most prop changes on Network, Volume, and Container are replacements (delete-first) — the old resource is removed before the new one is created. And a same-name pre-existing container, network, or volume without alchemy labels surfaces as Unowned: the deploy fails with OwnedBySomeoneElse until you opt in with .pipe(adopt(true)) or --adopt. :::

Compose with your cloud

ECS Tasks build their image with your local Docker — the AWS provider layers the same Docker service — and Cloudflare Containers bring your own image by consuming the registry reference you push from Build and push an image. By contrast, MicroVMs build server-side on AWS and need no local Docker at all.

Reference

Container · Image · Network · RemoteImage · Volume