Notifications
Cinatra runs long. Agent runs take minutes, hours, sometimes days. The platform tells you when work finishes (or fails) instead of expecting you to keep a tab open.
What you see
Section titled “What you see”- A bell with a badge in the top bar — opens the notification feed.
- Toasts for things you just did (saved a draft, started a run) — transient, not persisted.
- The notification feed for things that finished while you were elsewhere — durable, persisted, scoped to you, marked read / unread.
What gets notified
Section titled “What gets notified”- The run you started finished — successfully, with a failure, or stopped.
- A run paused for your review at a human-in-the-loop (HITL) gate.
- A background job an admin operates failed and needs attention.
- A workflow you own moved a stage forward, an approval landed, or a milestone slipped.
How it works (briefly)
Section titled “How it works (briefly)”Notifications live in Postgres so they survive sessions, deploys, and reloads. The feed updates in real time over a Postgres LISTEN/NOTIFY channel — you don’t have to refresh.
- The implementation reference: Notifications.
- The operational angle (Sentry-compatible error reporting for the operator): Error reporting.
Where to go next
Section titled “Where to go next”- Durable workflows — why runs survive reloads in the first place
- Human-in-the-loop by design — what a HITL pause looks like before the bell rings