The requirement this solves: someone who is not comfortable with a terminal plugs a drive in, sees it in the app, starts using it, and later unplugs it safely, entirely from the UI.

Using a drive

Settings, Storage lists this server's own storage, every other registered library, and a Detected drives section.

A drive with a filesystem Local Drive can use shows Use this drive. One tap mounts it and registers it as a library. No further steps.
A drive with no usable filesystem shows Format, which is the one destructive action in the whole flow and is guarded accordingly: a sheet states plainly that everything on the drive will be permanently deleted, and the confirm button stays inert until you type ERASE THIS DRIVE exactly.
A drive already in use shows Eject, which flushes and unmounts before telling you it is safe to unplug.

More than one library can be in use at once. Each shows its own used, free, and total space, and any one can be marked Default for new top level uploads.

Combining drives

Selecting two or more mounted drives offers Combine into one drive. This uses mergerfs, a mature Linux union filesystem, rather than a custom allocator, because hand rolled disk spanning is exactly the kind of thing that quietly causes data loss.

Before anything happens, the sheet states what this is:

This adds capacity, not redundancy. If one drive fails, the files that happened to land on it are lost while the rest keep working.

Backups matter more with a combined drive, not less. See Backups.

The pool mounts with the mfs placement policy, which spreads new files toward whichever member has the most room. Everything else in the system treats the pool exactly like a single drive.

When a drive is unplugged

Removable media going away is a normal condition here, not an error state. The server checks periodically that each external library's mount point still resolves to the same device. If it does not:

The library is marked offline.
Reads and writes against it return one clear error rather than failing in unrelated ways.
The app shows a persistent banner instead of an error per file.
Every other library keeps working.

It returns to available on its own once the same device is back.

Without the privileged helper

An admin who would rather not grant SYS_ADMIN to any container can run scripts/setup-automount.sh once on the host instead. It installs a udev rule that mounts USB storage under the same directory, and the backend picks those up identically. Both paths land in the same Detected drives screen.

Drive management is Linux only either way, because mount and mergerfs are Linux tools. On a Windows host the helper says so plainly and everything else works.