v0.0.21 — guided migration wizard, simpler exposure setup, panel fixes

Released 2026-07-25.

Highlights

• Moving a Raspberry Pi’s system from its memory card to an SSD is now a full, guided wizard. It opens with a clear plan (what will happen, that the node reboots once on its own partway through, and that the job ends with you physically removing the memory card), then walks through each phase live — copying the system, setting the boot order, the reboot, verifying it worked, and finally retiring the memory card — so you always know what stage you’re at and what’s left. If you close the panel partway through and come back, it picks up exactly where it left off.
• Publishing a service to the Internet now needs only one Cloudflare API token. Buddy creates and manages the tunnel for you from that single token, instead of asking you to hand-create one and paste in two separate credentials. The guided setup was also rebuilt around five clear steps — what’s needed, how your connection reaches the Internet, your Cloudflare details, Buddy setting everything up, and confirmation you’re ready — with no step assuming you already know what a tunnel or a certificate is.
• A data disk from a node that was rebuilt or re-added now offers to be recovered, not just wiped. If a disk still holds a storage pool from before but the panel had lost track of it, it now offers an Adopt action that recovers the pool and what’s stored on it, instead of the only option being to prepare (and erase) it as if it were empty.
• Signing in takes one fewer step. Once your six-digit authenticator code is complete, it signs you in automatically. There’s nothing left to click.
• A batch of smaller panel fixes: an error message could get stuck on screen with no way to dismiss it; a few storage cards showed a value so long it spilled past the edge of the card; and the required-field marker on a few forms could drop onto its own line instead of sitting next to the label.

Known limitations

• The migration wizard above has not been run start-to-finish on real hardware yet. It’s been tested thoroughly in isolation, but the live, on-a-real-node run-through — including a real reboot onto the new drive — hasn’t happened for this release. Treat it with extra care until that’s done: keep the memory card as your fallback until you’ve confirmed the node comes back up correctly.
• Recovering a pool from a disk that was already mounted, and restoring an adopted pool’s stored-app records after a more unusual failure, are both covered by automated tests only. Neither has been exercised on a real cluster yet.