Data and backups

Everything Computer needs to survive a reinstall lives in one data directory: ~/.cptr by default, /data in Docker, or wherever CPTR_DATA_DIR points. The one exception is chats: they live inside each workspace folder and travel with the project.

What lives where

Inside the data directory:

Path	What it holds
app.db	SQLite database in WAL mode (app.db-wal and app.db-shm sit next to it). Users and auth, instance config, automations, chat/message records, upload metadata.
config.toml	The server secret plus a mirror of app config. On startup the file is re-seeded into the database (the file wins), so it's safe to hand-edit while stopped, and config survives a lost database. See config.toml.
uploads/	Uploaded file blobs.
logs/	Audit and upstream-request logs, when enabled.
memory/	Per-user AI memory.
skills/	Managed global skills.

Inside each workspace, Computer keeps a .cptr/ folder:

Path	What it holds
<workspace>/.cptr/chats/<chat_id>.json	Every chat for that workspace; move the folder and the chats come with it.
<workspace>/.cptr/artifacts/	Artifacts produced in that workspace.
<workspace>/.cptr/task_logs/	Logs from task runs.

A data-directory backup does not include your workspaces. Project folders need their own backup (git remote, Time Machine, whatever you already use), and because chats live in <workspace>/.cptr/, that backup covers them too.

Back up the data directory

Stop the server first

app.db uses SQLite WAL mode. Copying it while the server is writing can produce a corrupt or inconsistent backup. Stop Computer, copy, then start it again.

For a Python install, archive ~/.cptr:

tar -C "$HOME" -czf cptr-data-backup.tgz .cptr

For Docker, archive the stopped cptr-data volume into the current directory:

docker run --rm \
  -v cptr-data:/data:ro \
  -v "$PWD:/backup" \
  alpine tar -C /data -czf /backup/cptr-data-backup.tgz .

Confirm the archive contains the two files that matter most:

tar -tzf cptr-data-backup.tgz | grep -E '(^|/)app\.db$|(^|/)config\.toml$'

Test a restore

Don't wait for a disaster to find out the backup works. Unpack into a fresh directory and start a second instance on another port with CPTR_DATA_DIR pointing at it:

mkdir -p /tmp/cptr-restore
tar -C /tmp/cptr-restore -xzf cptr-data-backup.tgz
CPTR_DATA_DIR=/tmp/cptr-restore/.cptr cptr run --port 8001

Sign in at http://localhost:8001 and check that your account and settings are there.

For Docker, restore into a new volume rather than overwriting production:

docker volume create cptr-data-restore
docker run --rm \
  -v cptr-data-restore:/data \
  -v "$PWD:/backup:ro" \
  alpine tar -C /data -xzf /backup/cptr-data-backup.tgz

Then start a test container with -v cptr-data-restore:/data and sign in. Remember that workspaces are restored separately: the data backup gets you accounts, config, and settings, not project files.