Fields
| Field | Type | Required | Description |
|---|
autosave | bool | optional | Enable autosave snapshots. Default: false |
autosaveDebounceSeconds | int | optional | Minimum seconds between autosaves. Default: 30 |
retention | object | optional | Snapshot retention policy |
retention.maxSnapshots | int | optional | Maximum snapshots to retain. Default: 100 |
volumes | volume[] | optional | Volume definitions with consistency policies |
volumes[]
| Field | Type | Required | Description |
|---|
name | string | required | Volume name, referenced by services |
consistency | enum | required | sqlite, postgres, or generic |
Consistency values
| Value | Behavior |
|---|
sqlite | WAL checkpoint (PRAGMA wal_checkpoint(TRUNCATE)) + fsync before snapshot |
postgres | Logical dump via pg_dump for portable, transaction-consistent snapshots |
generic | Best-effort copy — pause or stop the container, tar the directory, resume |
Example
state:
autosave: true
autosaveDebounceSeconds: 30
retention:
maxSnapshots: 100
volumes:
- name: dbdata
consistency: postgres
- name: uploads
consistency: generic
- name: app-db
consistency: sqlite
How autosave works
When autosave: true, the host app schedules a snapshot every autosaveDebounceSeconds seconds. Snapshots older than maxSnapshots are pruned automatically (oldest first). The current state is never deleted — only older snapshots are pruned.
Snapshots are content-addressed and deduplicated — unchanged data between snapshots is stored only once. Disk usage grows proportional to actual data changes, not linearly with snapshot count.
