Computer export & migration

The Computer keeps its whole desktop — logins, bookmarks, downloads, and files — on one Docker named volume. Export packs that volume into a single archive; import restores it on this or another machine. One file out, one file in.

#Where files live

• Everything durable sits on the named volume mounted at /config (the desktop user's home). Removing the container is safe; the volume is the Computer.
• Workspace convention: keep Human- and bot-produced files in ~/workspace (that is /config/workspace). Every start creates the directory if it is missing, it sits on the volume, and it travels through export → import. Browser state stays separate under ~/.config/chromium.
• The export directory is a path on the host: it defaults to ~/Desktop/BotHarness Exports (or ~/BotHarness Exports when the home has no Desktop) and can be changed in Settings → BotHarness → Computer when a directory picker is available. The archive file appears there; import only lists and accepts files from that directory.

#Export

1. Open Settings → BotHarness → Computer. Nothing has to be configured — the export directory already shows its current (default) path — but you can pick another destination with Choose… when a picker is available.
2. Click Export to… (or Export), choose the destination, and confirm. The Computer closes its browser gracefully, stops, packs the volume, and restarts if it was running; the rows show the live stage and elapsed time.
3. When it finishes, the note shows the archive path — one timestamped .tar file. On deployments without a working picker, where the directory was fixed, the folder also opens automatically in the Host's file manager.

#Import on another machine

1. Install a container runtime (Docker or Colima) and DeepSeek Harness with BotHarness on the target machine; start the profile once so the Computer plugin loads.
2. Copy the .tar archive over (USB, scp, network share).
3. In Settings → BotHarness → Computer, note the target's export directory (the default shown works with no setup) and put the archive in it.
4. Click Import…, pick the archive, confirm. The Computer recreates the volume, unpacks the archive, and starts — logins, bookmarks, and ~/workspace files are back.

#Browser download and upload

On web deployments (no directory picker, no Host folder to open), transfer runs through the browser instead of Host paths:

• After an export finishes, Download fetches the archive through the browser's save dialog — streamed, so a ~1 GB file never sits in memory. The export directory stays as the server-side listing location.
• To import, Choose archive file… picks a local .tar, then Authorize and import streams it up and runs the standard import. Long archive names wrap with an ellipsis; the full name is one hover away. An upload only replaces an existing archive once its bytes have fully arrived.

#Size and time expectations

• The archive is an uncompressed tar of the volume. After normal use the volume is often around 1 GB (browser profile, caches, and workspace files); check the produced file, or docker system df -v for the live volume.
• Local export and import of ~1 GB typically take under a minute each way on a developer machine (disk-bound; importing also verifies the pinned image is present).

#Treat archives as secrets

The archive contains the Computer's cookies and saved logins — the same sensitivity as a browser-profile backup. Store and transfer it accordingly. It is a profile-level facet and never part of a PersonaBot export.

#Reading developer diagnostics

GET /api/computer/diagnostics returns the bounded lifecycle, container, and viewer event log (newest last). The viewer card narrates its own mounts, stream phase changes, reloads, and retries there — start here when the picture misbehaves instead of guessing.

#See also

• ADR-0052 — storage stays on a named volume; portability is an explicit export.
• ADR-0062 — quiesced exports and the ~/workspace convention.