Trust
Reliability and backups
What keeps NEMUQU running and recoverable today, what has been tested and when, and what is not promised.
On this page
How the service runs
- The website is served from Vercel's edge network. The API, analysis worker, report renderer, scheduled jobs and job cache run on Render in Oregon. The database and file storage are a Supabase project in us-west-2.
- Each of those is a single deployment in one region. There is no automatic failover to a second region.
- Uploads are processed by a background worker with job leases and bounded retries, so a worker restart resumes work instead of losing it.
Monitoring
A scheduled check runs every ten minutes against the API, database, analysis pipeline, job cache and report renderer and alerts the operator on failure. Service status shows what your browser can observe of the API right now, beside incident updates written by the operator.
Backups and what has been tested
| Mechanism | What it covers | Evidence |
|---|---|---|
| Daily file backup | A copy of each original file in a separate storage bucket | Scheduled job, running daily; a synthetic file was restored with a matching hash on September 27, 2026 |
| Provider database backups | The whole database, on the provider's daily schedule | The provider states seven days of daily backups for the plan in use. A full restore into an isolated project matched production record counts on September 27, 2026 |
| Deletion markers | Records of deleted cases, so a restore cannot bring deleted files back | A synthetic backup, delete, restore and replay test passed on October 4, 2026. Replay after a real provider restore has not been exercised |
The October 2026 outage
From October 11 to October 13, 2026 the investigation service was unavailable because the hosting provider had suspended its services. The scheduled check was suspended with them, so it could not raise an alert. The full timeline is on Service status. It is the reason this page does not describe the monitoring as independent of the hosting provider.
During an outage
- Stored cases and files stay where they are; an outage of the application does not delete data.
- Public pages, the synthetic demo cases and sample reports are served separately and usually stay available.
- Updates are posted on the status page as they are confirmed.
Last updated .