Changelog

🚧 In development — changelog will be published on release.

Every scheduled backup was creating a brand new Google Drive folder and re-uploading the entire dataset from scratch, instead of incrementally updating the existing one.

• scheduler.ts — Scheduled cron callback now queries the most recent job and passes incrementalFromJobId
• scheduler.ts — Health Watcher auto-recovery now passes incrementalFromJobId from the stalled job
• Eliminated redundant full backups consuming 824 GB across 12 clients

The "Backup Now" button returned a 500 error. The API route ordered jobs by createdAt — a field that doesn't exist in the schema.

• Updated orderBy from createdAt to startedAt
• Wrapped auto-inference query in try-catch for graceful degradation
• Manual backups now correctly auto-infer incrementalFromJobId

When a backup hit Google's daily API quota, workers would park but the job kept running silently for ~5 minutes until the Health Watcher killed it.

• job-runner.ts — Added [Parked] detection in both completion and fatal catch blocks
• Jobs now exit cleanly within seconds with a distinct "parked" status
• The parked status preserves runFolderId and counters for incremental continuation

Cancelling a running backup left zombie threads alive — long-running Google API requests continued consuming RAM, eventually crashing the server.

• Replaced boolean cancellation flag with AbortController registry in cancel-store.ts
• Threaded AbortSignal throughout the entire pipeline
• All HTTP sockets now tear down immediately upon cancellation

Permanently deleting a client appeared to succeed but the client remained. Multiple missing columns and methods in the DB wrapper caused silent crashes.

• Added telemetry columns to AuditLog schema migration
• Added count() and aggregate() methods to backupFileIndex
• Fixed getGlobalAuth import and TypeScript type casting

The custom node:sqlite wrapper didn't support the not operator and silently coerced the filter object to "[object Object]".

• Added not operator support to buildWhere() in lib/db.ts
• Archived clients now correctly hidden from the main list

• New distinct amber "Parked (Quota)" badge in the StatusBadge component
• "Parked" option added to the Runs tab status filter
• Client card displays parked status with warning colour

The optimization tool was only running SQLite VACUUM, which reported "already optimised" even with 400 MB of bloat.

• Phase 1: Purges soft-deleted BackupJob rows (500 KB–5 MB each)
• Phase 2: Sweeps orphaned BackupFileIndex rows
• Phase 3: Clears stale AuditLog entries older than 90 days
• Phase 4: Runs VACUUM to rebuild the SQLite file
• Phase 5: Runs PRAGMA wal_checkpoint(TRUNCATE) to shrink WAL to zero

• Creates a timestamped .pre-vacuum.bak safety backup before destructive operations
• Includes WAL and SHM files in the backup
• If backup creation fails, the vacuum is aborted entirely

• Added PRAGMA wal_checkpoint(TRUNCATE) after VACUUM
• WAL file now shrinks to zero bytes after optimization

• Added ALTER TABLE bootstrap migrations for lastVacuumAt, vacuumTargetHour, and masterScheduleEnabled

Clients can now be archived (soft-deleted with configurable retention) before permanent destruction.

• New "Destruction Modal" — choose between Archive or Permanent Delete
• Archive Hub (/clients/archive) with countdown timers and restore capability
• Custom glassmorphic modals for Restore and Force Deletion

• "Settings" button renamed to "Delete Account" and repositioned in the page header
• Delete button directly opens the Destruction Modal

• Checks against both business name and domain when creating clients
• Clean, human-readable error messages replacing JSON parse failures

• Timeline labels permanently visible (no hover required)
• Compact dropdown replacing manual number input
• Busy time slots hard-disabled and unclickable

Replaced the legacy 3-step restore modal with a comprehensive 5-step Recovery Wizard.

• Three dedicated modes: Disaster Recovery, Point-in-Time Rollback, Migration/Clone
• "Type RESTORE" safety confirmation gate
• Pre-flight validation via /preflight API

• update.php now searches a cascade of DB paths before extraction

• npm run pack generates update.zip (Enterprise) and update-family.zip (Family)
• Resolved TypeScript strictness errors in production builds