Skip to content

Deployment

Docker

The included Dockerfile is multi-stage:

  1. Build stage — PDM exports production dependencies to requirements.txt from pyproject.toml / pdm.lock.
  2. Runtime stage — pip install -r requirements.txt, copy src/ and LICENSE, run as non-root user appuser.

Default command:

CMD ["python", "src"]

The image does not bake in config.yaml. Provide configuration at runtime:

  • Mount config.yml / config.yaml, or
  • Inject BOTKIT__… environment variables (see Configuration)

Generate env lines from a local YAML file before deploy:

pdm run convert-config -i config.yaml --terminal

Process layout (use, cache, REST mode) is covered in Configuration.

Environment-first deploys

Platforms without persistent volumes (Docker Swarm, some PaaS) often prefer env-only config. Merge order means env overrides file — you can ship a minimal file plus secrets as env vars.

Required minimum:

BOTKIT__bot__token=…
BOTKIT__use__bot=true

Health and observability

  • Use extension HTTP routes (e.g. bundled /ping when ping's backend is enabled) or your own setup_webserver routes for load balancer checks.
  • Enable logging.file: true and mount logging.directory if you need log persistence on disk.
  • Sentry can be integrated in extensions (bundled nice-errors patches error paths when enabled).

What ships in the image

Production export includes runtime Python dependencies only — not dev tools (pytest, ruff, zensical). Build and test in CI or locally before pushing the image.