Review Requests and the Notification Center
A step in an automation asks a person to look: a flag, a soft review with a timeout, or a hard gate. Review is never a story state, and the Notification Center is safe to ignore. Built.
Review Requests and the Notification Center
Most of what an automated publication does should not wait for a person. Some of it should be seen. A Review Request is how a step in an automation, a procedure or an AI step asks a person to look without turning review into a stage every story has to wait in.
This is built and tested. Review requests arrive in the Notification Center, the first page an editor sees in a site's console.
Three modes
| Mode | Does anything wait? | At the deadline |
|---|---|---|
| Flag | Nothing waits. | None. It stays until someone acknowledges it. |
| Soft review | The process waits until a deadline, seven days by default. | If nobody answered, it records "proceeded without review" and the process continues. |
| Hard gate | The process never continues without an explicit answer. | Never. |
Safe to ignore
The principle is that an editor can interact with a stream of review and steering opportunities selectively, ignore some of them, and let the system run. If you only look, it keeps doing what it was doing. If you dig in and steer, say "adjust" or "not this kind", the next run reads your note and takes it into account, and the console shows "Applied in run". Only a hard gate waits for a person.
Marks and words accompany every colour, and the home page works from the keyboard.
Where you see them
Cited Trends raises one soft review per weekly run. Stage Bots raise a flag when a bot holds or fails and a hard gate when a chain hits its hop limit. The details are in the Papyrus repository, in docs/review-requests.md and docs/notification-center.md. See the editorial flows recipe.