Backup and restore

Exporting is a job, not a download

A .zip with the whole volume does not fit in the time of one request. So the route schedules the generation and returns the state; the file is downloaded later, once the status reaches READY.

Here or in the bucket

The .zip that stays in the container has a fixed name, one copy and an expiry. The one that goes to the bucket has a timestamped name, accumulates and never expires: it is the history, and it is where a remote restore reads from.

Automatic backup

Daily, weekly or by cron expression, in the container's timezone. The schedule is stored in Settings and travels as a whole object: “weekly with no day” and “cron with no expression” have no next occurrence.

Restoring: inspect first

1. 1Inspect reads the .zip and answers what a restore *would* do: one block per workspace, with the count of each scope. It writes no byte.
2. 2Select what comes back, per workspace: only the cards, only the projects, everything.
3. 3Import applies it. With no selection, it restores everything, except accounts, which take an explicit ask.

• The file can come from disk (file) or from the bucket (remote), never both: that is 400.
• A remote source loosens no validation: from there down there are only bytes.
• Scopes depend on scopes: epics requires projects, comments requires cards.