Skip to content

/deploy

  • Type: /deploy
  • What it does: Deploy with a safety net — PRE/DEPLOY/POST/ROLLBACK checklist with evidence and explicit approval
  • In Spanish: /desplegar (when Senzu is installed with --lang es)
  • It is a shortcut: asking for it in plain words triggers the same thing through the router.

Usage: /deploy [environment or target, e.g. "production" or "staging"] — the argument is optional unless stated otherwise; if none is given, the default behavior below applies.

Apply deploy-ops §references/deploy-checklist.md to whatever the user wrote after the command, with evidence for each point:

  1. PRE: suites green (verify-build), /launch at READY if applicable, fresh verified backup, rollback plan written in one line, migrations reviewed (expansive?), and — for production — ask for the user’s explicit approval NOW (without it, the guard blocks and nothing is deployed).
  2. DEPLOY: through the stack’s reproducible path (references/deploy-by-stack.md): pipeline/hook/ script, never improvised commands. SENZU_ALLOW_DEPLOY=1 only after the approval.
  3. POST (10 min): smoke test (home + /health + 1 critical flow), clean logs, Sentry with no new errors, devlog with SHA/time/result.
  4. If POST fails → ROLLBACK first (the plan from PRE), verify, investigate locally afterwards.

First deploy of a project: before that, review references/envs-secrets.md (secrets on the platform, APP_DEBUG=false) and references/backups-monitoring.md (backups + uptime + Sentry set up).

After the deploy and the 10 minutes of monitoring: /handoff if it is the delivery to the client; if not, version and time to the devlog and /next if there are cards left.