PgBackWeb is a self-hosted backup UI — Railway's official template around the published Docker image — giving Atlas DB a second, independent backup path to S3. It exists specifically to survive a Railway-side incident, not to replace Railway's own native volume backups.
Atlas DB already has native Railway volume backups. This adds a deliberately separate path to S3 — its own Railway service, its own small Postgres for config, so a Railway-side incident can't take out both copies at once.
Every database gets a dedicated, read-only role created specifically for backups — not the admin credentials used elsewhere. On the S3 side, a single IAM policy scopes access to exactly one bucket, nothing broader.
Every table in the source database's schema was confirmed present in the dump, and a sample table's row count and contents were spot-checked against production — not just a trusted status message. New databases reuse the same bucket and policy; only a new subfolder is configured.
PgBackWeb is an internal disaster-recovery safeguard — not a public product. Documentation-only repo; the service itself runs from a published Docker image on Railway.