Skip to content

Data Lifecycle Backup Restore and Exports

lukeskytorep-bot edited this page Sep 30, 2026 · 3 revisions

Data Lifecycle, Backup, Restore and Exports

Current database boundary

v0.7.13 uses SQLite schema 27. The controlled exact-green chain is:

  • 23→24 Viewer Learning;
  • 24→25 provider continuation persistence;
  • 25→26 typed Workspaces;
  • 26→27 factory Training classification.

Legacy databases are classified before normal load and are not silently migrated through the controlled v0.7.13 epoch.

Active, archived and permanently deleted

Supported primary objects use:

Active → Archived → Restore or Delete permanently

Permanent deletion is available only through the controlled purge path after a read-only preview. Sealed, frozen and locked guards remain in force outside that transaction.

Backup and restore

Backup validation includes database integrity and expected schema. Restore creates a safety backup before replacement. WAL/SHM sidecars and database lifecycle are handled by native code rather than ordinary WebView file access.

Provider continuation state is part of database backup/restore when present. It is stored outside visible message content, bound to the exact message/event and removed with its owner during controlled purge.

Target and learning history

  • schema 022 preserves immutable Viewer Notes source snapshots;
  • schema 024 stores Field Guide history;
  • schema 027 updates only the known factory classification metadata needed by the current curriculum;
  • stable target IDs and historical session/Research ownership remain intact.

Exports

Ordinary user exports favor readable formats. Research exports retain technical reproducibility material where needed. Provider-native continuation payloads, API credentials and hidden reasoning/signature state are excluded from ordinary transcripts and exports.

Exports may contain prompts, blind evidence, Reveals, targets or personal research material. Review files before sharing them.

Clone this wiki locally