Deployment & Release
The service deploys to Cloud Run through GitHub Actions. Staging deploys on every merge; production deploys when a release is tagged.
| Environment | Trigger | Workflow |
|---|---|---|
| Staging | Merge to main | Build and Deploy Staging |
| Production | A vX.Y.Z tag is pushed | Build and Deploy Production |
Changes to markdown files alone don't trigger a staging deploy.
Every pull request and every deploy runs Biome, the type check and the tests first. A failure stops the deploy.
Releasing to productionβ
- In GitHub Actions, run π·οΈ Tag new release on
mainand pick Patch, Minor or Major. - The workflow creates the next version tag and a GitHub release.
- The tag starts Build and Deploy Production, which also writes release notes and posts them to
#birdy-bot.
To roll back, run Re-Deploy Production (by tag) with an earlier tag.
Configurationβ
The deploy workflows don't set any variables. Configuration lives on the Cloud Run service, so a change is made with gcloud run services update in that environment's project. Use --update-secrets for secrets, never --set-secrets, which removes every other secret.
Routingβ
Birdy Bot is reached through the platform load balancer, not API Gateway, so Slack and Linear can call it directly. Each route checks its own credentials.
Monitoringβ
There are no Cloud Monitoring alert policies for this service, and Sentry isn't installed. Logs are in Cloud Logging. When a Linear triage fails, the bot comments on the issue.