Run PR-Agent as a pinned webhook server on the Gitea Docker network and review on PR open plus slash commands. The per-PR opencode installer and verdict-posting step are removed. The server-side .env stays outside the repository; deploy/pr-agent/ holds the compose definition and operator notes.
4.0 KiB
okf_version
| okf_version |
|---|
| 0.2 |
Start here
This bundle explains the Nuxt repository: how it is developed, checked, released, and what was learned getting there. It is the "why" behind the config files — the configs say what, these documents say why.
- Open questions - What is still undecided for the Nuxt site.
- Backend endpoints - The smallest API the frontend needs from the Go gateway: health, chapters, metrics, knockdown, events.
- Profile data - The LinkedIn export behind the terminal chapters, what is real vs mock, and how to refresh it.
- Future updates - Pipeline improvement backlog with effort estimates, not yet decided.
- Update log - History of changes to this bundle.
Decisions
CI/CD and releases (2026-10-18)
- PR pipeline layering - Three layers: pre-commit for cheap deterministic checks, Gitea Actions for thorough ones, AI review for judgment.
- PR-Agent for Gitea reviews - A pinned webhook server on server0 replaces the per-PR opencode job; internal-only, scoped bot, OpenCode Go model.
- AI review setup (opencode) - Superseded 2026-10-30; kept for the env-only auth and verdict-gate reasoning.
- Vulnerability audit policy - Blocking audit; reviewed ignore-list for unfixable advisories in pnpm-workspace.yaml.
- Release engineering - release-please in manifest mode; conventional commits enforced at commit time.
- Snyk paused and secrets policy - Quota-paused job kept in place;
write:issuescope required on every workflow token. - Publishing images to the own registry - A release tag builds the image and pushes it to registry.noelsabia.dev; the runner socket it needs is root-equivalent and belongs on a separate label.
Docker and development environment
- Dev container shape - The source is bind-mounted into a container that already holds the dependencies, and the dev server runs as the image's
nodeuser. - Build layer order - Why the manifests are copied before the source, and what invalidates what.
- Build context filter - What
.dockerignoremay exclude, and the one thing it must never exclude. - Production image shape - Multi-stage, digest-pinned, non-root runnable Node server - measured build and serve.
- Repository and compose layout - The Dockerfiles live with the app; the stack is assembled in
Local_Development. - Mockup layout - One
DESIGN.mdper screen, with the screenshot and the source embedded.
UI and icons (2026-10-23)
- Icons via Nuxt Icon - Icons come only from
@nuxt/iconwith a locally installed Material Symbols collection; inline SVG for SSR, scanned client bundle, and no icon webfont. - SEO stack and site metadata -
@nuxtjs/seowith one metadata file, aPersonidentity, a committed 1200x630 OG card, and no dynamic OG renderer or skew protection.
The one-paragraph history
The repository started as the Nuxt app shell. Docker development (dev
container, layer order, context filter) came first, then a single day of
CI/CD work built the PR pipeline layer by layer — lint, typecheck,
secret scan, dependency audit, Dockerfile lint, AI review — fixing one
real failure per layer as it appeared (wrong provider account, wrong
CLI major version, arm64-hostile vendor images, a silently-skipping
hashFiles, a prompt-injectable credential file), then wired releases
via release-please and locked the message format with pre-commit hooks.
The pipeline now fails loudly, posts only real reviews, and needs no
human faith to operate.