DevOps & Quality Study Portal

How to use it

The day-to-day developer loop is change → verify → deploy → watch.

shared Tilt resources.

1. Change. Edit code locally. Don't start your own dev server or watcher — use the

1. Verify (inner loop, via Tilt). Trigger the relevant checks and fix until green:

``bash tilt trigger <app>-lint tilt trigger <app>-typecheck tilt trigger <app>-unit-tests tilt trigger <app>-prod-build ``

For UI/flow changes, run the relevant Playwright suite in isolation (not the whole battery) to avoid rate-limit flakes:

``bash npx playwright test --project=<area> ``

audit on the PR.

1. Let CI gate it. Push; GitHub Actions re-runs the same gates plus the security

roll it out with a digest-pinned manifest:

1. Deploy. Build the image, push it to the private registry over WireGuard, and

``bash manage.sh deploy <app> ``

1. Watch the deployed result. Confirm reality matches intent:

OpenTelemetry traces if latency moved.

critical-alert email immediately.

SEO, Lighthouse ≥ 80, axe-core a11y).

• Smoke the surface: manage.sh smoke-test production
• Watch Sentry for new errors, Prometheus/Loki for metric/log anomalies,
• Read the Daily Environment Report email; respond to any Alertmanager
• For public web apps, confirm the static-smoke audit stays green (analytics,

is just re-pinning that digest — no ambiguity about "which build."

1. Roll back if needed. Because the previous deploy was a specific digest, rollback