# A deployment that survives a reboot and can be upgraded without ceremony:
#
#     docker compose -f docker/compose.yaml up -d
#     docker compose -f docker/compose.yaml exec altero altero user add <name>
#     docker compose -f docker/compose.yaml exec altero altero user password <name>
#     docker compose -f docker/compose.yaml exec altero altero login list
#
# Run those from the repository root. `-f` is what the file living in docker/
# costs; `export COMPOSE_FILE=docker/compose.yaml` buys it back for a shell.
#
# A checkout is not needed to run altero: the image is published, so this file
# on its own is a deployment. Downloaded next to a `.env`, it is
# `docker compose up -d` and nothing else. Building it from a checkout instead
# is compose.build.yaml beside this file.
#
# Point the Zotero client at http://<host>:8000/ and turn its streaming off --
# see "Point Zotero at altero" in README.md.
#
# Named explicitly: the project name otherwise follows the directory the file
# sits in, and renaming it to "docker" would leave an existing deployment's
# volumes behind under the old prefix with its library in them.
name: altero

services:
  db:
    image: postgres:18-alpine
    environment:
      POSTGRES_USER: altero
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-altero}
      POSTGRES_DB: altero
    # One mount at /var/lib/postgresql, not at .../data: 18 keeps its files in
    # a major-version subdirectory, and mounting the parent is what the image
    # asks for, so that a later pg_upgrade --link is not working across a mount
    # boundary. A volume written by 17 stops the container here with an error
    # naming the version that wrote it -- it is never adopted silently -- but
    # that also means moving from 17 is a dump and restore, not a pull.
    volumes:
      - db:/var/lib/postgresql
    # altero waits for this rather than crash-looping until the database is up.
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U altero"]
      interval: 5s
      timeout: 5s
      retries: 10
    restart: unless-stopped

  altero:
    # Pulled rather than built. `latest` is the newest release, prereleases
    # included, and moves when one is tagged; ALTERO_IMAGE_TAG pins a release
    # (`1.0.0-beta.3`) or follows master (`dev`) instead. An upgrade is
    # `docker compose pull && docker compose up -d`.
    image: ghcr.io/eseifert/altero:${ALTERO_IMAGE_TAG:-latest}
    # Migrations run on start, so this is also how an upgrade applies them.
    depends_on:
      db:
        condition: service_healthy
    # Every ALTERO_ setting in docker/.env, not only the ones this file happens
    # to name: a setting that has to be added here before it does anything is a
    # setting people set and then report as broken. The whole file is handed to
    # the container, so keep in it only what the application should hold.
    # Optional, because an instance on the defaults never writes one.
    # docs/configuration.md lists what can go in it.
    env_file:
      - path: .env
        required: false
    # These win over the file above, which is what makes a value exported in
    # the shell -- a CI job, a systemd unit -- reach the container as well.
    environment:
      ALTERO_DATABASE_URL: postgresql+asyncpg://altero:${POSTGRES_PASSWORD:-altero}@db:5432/altero
      # Outgoing mail. Empty is the working default: the messages go to the
      # log, which is where the first account reads its own confirmation link.
      # See docs/email.md.
      ALTERO_SMTP_URL: ${ALTERO_SMTP_URL:-}
      ALTERO_MAIL_FROM: ${ALTERO_MAIL_FROM:-altero@localhost}
      ALTERO_PUBLIC_URL: ${ALTERO_PUBLIC_URL:-}
    # So that a relay running on the host is reachable as
    # host.docker.internal: inside the container `localhost` is the container,
    # and a submission port on the host is not on it.
    extra_hosts:
      - "host.docker.internal:host-gateway"
    volumes:
      - storage:/data
    # Bound to the loopback interface: put a TLS terminator in front rather
    # than exposing the API to the network directly.
    ports:
      - "127.0.0.1:${ALTERO_PUBLISH_PORT:-8000}:8000"
    restart: unless-stopped

volumes:
  db:
  storage: