Reef speaks a small HTTP object API — PUT / GET / DELETE /o/<bucket>/<key> — on port 9000. Any machine on your network can push data in with plain curl. The admin interface isn't a web dashboard you configure: it's the SeaOS desktop itself. Open the Reef app and watch objects arrive.
Reef answers 201 Created only after the object's bytes are fsynced to disk and atomically named — each object is one real file on the data partition you can ls. Pull the plug mid-afternoon; on the next boot the catalog is rebuilt from the disk itself and the object is there. A backup box that says "saved" must never mean "probably".
write → fsync → atomic rename, before the HTTP response. Power cuts lose nothing acknowledged, and never leave half an object.
Objects are plain files; the catalog is rebuilt by reading the disk at start. Nothing to corrupt, nothing to drift.
Plug in a monitor: usage tiles, the live object list, start/stop with one key. No SSH, no config files, no web login.
A default boot makes zero outbound connections except DHCP/DNS. Your data — and the fact that you have data — stays home.
What Reef is not: a NAS clone. No SMB shares, no RAID manager, no plugin store. It stores objects over HTTP and never loses them — one job, done properly.