Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

CI/CD

Deployment validation runs locally against the published images. The Forgejo workflow publishes this template’s documentation; it does not deploy a running stack.

Local validation

  • validate Compose rendering
  • run a local Compose smoke test against the published image
  • validate Kustomize rendering
  • run a local Kubernetes smoke test against the dev overlay
  • run the validation-helper tests with Python’s standard library

Repository entry points:

  • Makefile
  • scripts/validate-local-compose.sh
  • scripts/validate-local-kubernetes.sh
  • scripts/validate_see_service.py

Useful local targets:

  • make validate
  • make validate-helpers
  • make validate-compose
  • make validate-compose-re-indicators
  • make validate-compose-see-impacts
  • make validate-compose-calculations
  • make validate-kubernetes
  • make validate-kubernetes-re-indicators
  • make validate-kubernetes-see-impacts
  • make validate-kubernetes-calculations

Deployment checks cover configuration rendering, service readiness, initialization, and access to required core schemas. Operations documents the checks for each service. The helper tests also check failure reporting and that validation cannot overwrite the operator’s Compose environment or remove an existing Kubernetes deployment.

For production-oriented overlays, local validation currently proves deployment and health. End-to-end authenticated functional checks still require real auth infrastructure and credentials.

The local validation scripts assume:

  • Docker or a compatible Compose provider is installed
  • Python 3 and curl are installed
  • kubectl is installed
  • a local Kubernetes cluster is available for the Kubernetes smoke test
  • Podman/Docker networking allows outbound DNS and HTTPS from the runtime containers

Versioned dependencies

Service image references are pinned to the published releases listed in Configuration, consistently across Compose and Kubernetes. Model entries use versioned Codeberg artifact URLs in both catalogs. Keep both deployment formats and model catalogs aligned when updating these dependencies.

Documentation publishing

The repository uses mdBook for documentation publishing. The existing pages workflow can publish the generated site from docs/.

Release archival

The GitHub mirror creates an archival release when a version tag is pushed. Its release notes are extracted automatically from the matching version heading in CHANGELOG.md, with the leading v removed from the tag for matching. Missing, duplicate, or empty entries stop release creation. The workflow also links to the canonical Codeberg repository and documentation; it does not build container images, SDKs, or API artifacts.