Deployment

Deployment

We practice continuous deployment. Every merge to main triggers an automated pipeline that builds, tests, and ships to production.

Pipeline stages

1. Build — compile TypeScript, bundle assets
2. Test — unit tests, integration tests, type checking
3. Preview — deploy to a preview environment for visual verification
4. Production — automatic promotion after checks pass

Deployment schedule

• Continuous — code merged to main deploys automatically
• Feature flags — use flags for large features that aren't ready for all users
• Freeze periods — no deploys on Fridays after 4 PM or during incidents

Rollback procedure

If something goes wrong:

1. Click Rollback in the deployment dashboard
2. Post in #incidents with a brief description
3. Investigate the root cause
4. Write a postmortem if the issue affected users

Monitoring

After deploying, keep an eye on:

• Error rates in Sentry
• Latency metrics in Grafana
• User-facing alerts in PagerDuty

// Example: Feature flag check
if (await featureFlag("new-dashboard")) {
  return <NewDashboard />;
}
return <LegacyDashboard />;

Best practices

• Deploy small changes frequently
• Monitor your deploys for at least 15 minutes
• Never deploy and disappear — stay available to respond