# Changelog

All notable changes to Swish Migrate and Backup will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [1.5.0] - 2026-11-05

### Fixed
- **Backups from the Backups page could never finish on real hosts and could produce unreadable archives.** The browser-driven pipeline had several independent defects, all fixed:
  - A file that did not fit in one request's time budget was appended again from the start on the next request, forever: the archive grew without bound and the header of the first attempt pointed into garbage. Large files are now resumed from the exact byte where the previous request stopped; the archive position is recorded with every step and anything a crashed request appended without recording is cut off before writing continues.
  - The whole database was dumped inside a single request, and that dump silently dropped tables once it had used `max_execution_time - 10` seconds (on hosts with a 20-second limit it wrote zero tables and still reported success). The dump now runs in time slices across requests, resumed from a checkpoint, and is streamed into the archive instead of being read into memory (a 200 MB dump fataled under a 128M memory limit).
  - The archive header recorded a file's size at index time but the content was whatever the file contained at archive time, so a log, cache or upload that changed in between desynchronised every entry after it. Exactly the recorded size is now written (long files are cut, short files padded).
  - Two requests could write the same archive at once (a proxy timed out one request while PHP kept running it, and the browser retried). A per-job database lock now serialises steps; a request that meets the lock gets a `busy` answer and the browser simply waits.
  - File indexing ran inside the "start" request. On large sites that request outlived proxy timeouts and the browser never learned the job id. Start now returns immediately and indexing runs in the first continue steps.
  - The browser gave up on a backup after 30 polls total, so one hiccup on a long backup silently abandoned it. It now retries consecutive errors with backoff and shows the server's failure message when a job really fails.
  - The backup type was ignored: a "Database" backup archived every file and a "Files" backup included the database. Types are honoured.
  - Every archive is now walked header by header before it is registered; a structurally broken archive is discarded with a clear error instead of being listed as a good backup. Verification and the MD5 checksum run in time-bounded, resumable steps (hashing a 5 GB archive used to be one 86-second request).
  - With "WordPress core files" enabled, `wp-content` was indexed twice (once via the site root, once on its own) and archives were twice their real size. Directories nested inside another scanned directory are skipped.
  - On hosts with `max_execution_time = 0` (unlimited) the pipeline never archived a single file: `ServerLimits::is_approaching_time_limit()` compared the float "unlimited" marker with a strict integer check that never matched, so every request looked out of time before it started. The same stall was triggered by the new `set_time_limit(0)` in continue requests until this was fixed.
  - The job state option is no longer autoloaded on every page load.
  - Hard-kill windows (PHP-FPM `request_terminate_timeout`, out-of-memory, worker restart) are handled: a request that dies after finishing files but before saving its cursor no longer causes the next request to cut those files out of the archive (the archive tail is walked header by header and only the incomplete entry is removed); a partial entry with zero content bytes is restarted instead of being given a second header; a database dump slice that died after writing rows is truncated back to the recorded length before resuming; job state is re-read under the per-job lock so a retry never acts on a stale cursor.
  - An archive that cannot be written (disk full, quota) now fails the job with the real reason instead of being retried forever as a "timeout".
  - If the database cannot provide named locks (`GET_LOCK` returns NULL), steps run unlocked instead of waiting forever.
  - Failed jobs remove their `temp-<job>/database.sql` directory.
- **Imports (Migration page) failed or stalled on anything but small sites.**
  - `.swish` archives were extracted in one request (300-second internal limit) and files were copied into place in one request. Both now run in bounded, resumable chunks; a request that dies mid-file is cut back to the last recorded byte on resume.
  - The browser poll and the server's own loopback request both executed the same import step concurrently (duplicate SQL, racing offsets). A per-session database lock now allows one executor at a time.
  - Choosing "clear stale session and start fresh" fataled (`$this->logger` was never set in the REST controller).
  - Moving files from the temp directory to `wp-content` across filesystems no longer turns into one multi-gigabyte `rename()` copy (PHP silently copies on cross-device renames; the move now checks the device first and copies in resumable chunks).
  - `/*!40000 ALTER TABLE ... DISABLE KEYS */;` lines in dumps lost only their opening comment marker, so every table produced two SQL syntax errors during restore. The closing marker is stripped as well.
  - The per-request import budget now adapts to the host's `max_execution_time` like the backup pipeline (10-25 seconds) instead of a fixed 25.
  - An extraction that makes no progress (truncated archive, unwritable destination) or a file copy that fails mid-way (disk full) now fails the import with a message instead of looping or silently installing a truncated file; the extracted-file list is written atomically; the extraction directory is kept alive for the whole import so a concurrent backup's orphan cleanup cannot remove it.
- **Scheduled backups**: a request killed during the file scan no longer leaves the job stuck in "processing" (the scan is checkpointed and restarted); an archive that was already finalized when the request died is kept instead of being rebuilt; temp directories and on-disk file lists are cleaned up when a job fails; overlapping scan directories are de-nested instead of keeping every path in memory.
  - The plugin's own directory is skipped on extraction so an import can no longer overwrite the code that is running it; backups also exclude it (as scheduled backups already did).
- **Scheduled backups silently stopped at 50,000 files** and stored the entire file list inside the checkpoint row rewritten every 10 seconds (exceeding `max_allowed_packet` and memory on large sites, after which the job failed with "Backup checkpoint expired or not found"). The file list now lives on disk and is streamed; files-only scheduled backups now include a manifest so they can be imported.

### Added
- `tests/ddev/` integration tests runnable against the DDEV site (`bash tests/ddev/run.sh`), covering archive resume, chunked database dump, structural verification, the REST pipeline, locks, chunked import and the scheduled engine.
- Filter `swish_import_time_budget` (default 25 seconds per import request).

## [1.4.0] - 2026-08-03

### Added
- **Full translations for Spanish (es_ES), French (fr_FR), Simplified Chinese (zh_CN), and Arabic (ar).** All 908 strings translated; `.po`/`.mo` files are bundled in `languages/`. Japanese (ja) remains available as a translate.wordpress.org language pack.
- Bundled translations are loaded via `load_plugin_textdomain()` on `init`; wordpress.org language packs still take priority when installed.

## [1.3.1] - 2026-07-14

### Changed
- **The plugin now activates without the mysqli PHP extension.** SQLite-based environments (WordPress Playground, WordPress Studio) lack mysqli, and the hard activation block prevented translators from previewing the UI via the translate.wordpress.org Playground link. mysqli was removed from the hard-required extension list (`zip` and `json` remain required); when it is missing, a warning notice on the plugin's admin pages explains that backup, restore, and migration features require a MySQL/MariaDB database.

### Fixed
- **Literal `%` characters in database content leaked into backups as `{64-hex}` placeholder tokens.** All four SQL dump writers (backup engine, export, multisite shared/site tables) escaped values with `$wpdb->_real_escape()`, whose output contains WordPress's placeholder-escape token for every literal `%` — core only strips it when a query executes, never in dump text. Restoring such a backup persisted the token into post content and corrupted serialized options (broken `s:N:` length prefixes → `unserialize()` failures, e.g. a `%s %v` currency format fataling a destination plugin on PHP 8). Values are now passed through `$wpdb->remove_placeholder_escape()` before being written.

## [1.3.0] - 2026-07-13

### Fixed
- **Multisite migration imports failed with repeated 500/403 errors** and never reported completion, even when the import had actually finished. Three root causes, all fixed:
  - The database restore overwrites `wp_usermeta`, invalidating the admin session's nonce so every progress poll returned 403. The progress handler now verifies the nonce softly and falls back to a read-only response gated by the unguessable job UUID, so the browser still receives the final status after its session is gone (also registered the `nopriv` AJAX variant).
  - Import progress was stored only in a transient inside `wp_options` — the very table the restore replaces — so progress was lost. Progress is now mirrored to `wp-content/swish-backups/temp/import-job-{id}.json` and read from there first.
  - Half-restored tables can fatal other destination plugins (500s) during the restore window. The progress dialog now tolerates 5xx/network errors for up to ~6 minutes once the restore has started, showing a "site may be briefly unreachable" message instead of failing after 3 attempts.

## [1.1.11] - 2026-05-27

### Changed
- **SWISH is now the only archive format** for new backups - removed Archive Format selector from Settings page
- Settings save handler (admin form + REST API) always stores `archive_format = 'swish'`

### Fixed
- **Windows copy/paste of curl and aria2c download commands** - commands were generated with backslash line continuations (POSIX shell syntax), causing CMD/PowerShell to parse each line as a separate command. Commands are now generated as a single line that copies as one command on bash, zsh, PowerShell and CMD

### Notes
- Restore still supports existing `.zip` and `.tar.gz` backups - underlying archiver/extractor classes remain in place

## [1.1.10] - 2026-05-18

### Fixed
- **PHP 7.4 compatibility** - converted `match` expressions to `switch` statements for broader PHP version support
- **Cancel job button** - now always cleans up localStorage even when server returns "job not found"
- **Stale jobs in UI** - polling error handler now properly removes job cards from UI when jobs not found on server

### Added
- Debug logging for backup progress troubleshooting (console messages with `[SwishBackup]` prefix)

## [1.1.9] - 2026-05-18

### Added
- **SWISH archive format option** in Settings dropdown - our custom streaming format with true append support and resume capability

## [1.1.8] - 2026-05-13

### Fixed
- **Missing files in backups** (e.g., uploads/2026 folder) - removed timeout checks during file indexing phase
- File enumeration now completes fully before archiving begins (AI1WM approach)
- Only memory pressure triggers yielding during indexing, not time limits
- **Job steps log not updating** during pipeline backups - now correctly shows phases
- Steps log displays pipeline phases: Scanning files → Archiving files → Finalizing
- Initial steps now displayed immediately when backup starts
- **Path resolution for symlinks and Docker** - improved `get_relative_path()` with realpath matching for plugins, themes, uploads, and wp-content directories
- **Plugin vendor directories not backed up** - removed `vendor` from default exclusions (was breaking plugins like Yoast Duplicate Post, WooCommerce that use Composer autoloaders)

### Changed
- **Backup size limit increased** from 2GB to 4GB for free version

### Added
- **Settings button on Backups page** for quick access before creating backups
- New settings modal with performance controls and hosting presets
- Quick presets for Shared Hosting, VPS/Managed, and Dedicated servers
- Backup contents toggles (Database, Plugins, Themes, Uploads, Core Files)
- CLI mode for BackupPipeline that disables time-based yielding
- `is_cli()` detection in ServerLimits class

### Changed
- **CLI backup performance** - now uses pipeline approach with optimized settings
  - Batch size increased to 500 files per iteration (no HTTP timeout concerns)
  - Disabled time-based yielding - only yields on memory pressure
  - Files write completely without internal chunking
  - Per-file logging reduced from info to debug level
- **Browser backup optimization** - dynamic settings based on server capabilities
  - Dynamic time budget using 60% of max_execution_time (10-25s range)
  - Adaptive batch size via ServerLimits detection for different hosts
  - Default batch size increased from 50 to 100 files per request
  - Automatic tuning for WP Engine, Kinsta, Flywheel, and managed hosts
- **Simplified settings UI** - consolidated to key settings: Files per Request, Database Rows per Batch
- Import pipeline reliability improvements
  - Fixed race condition in file queue processing
  - Increased stale session timeout from 2 to 10 minutes
  - Added database index on `updated_at` for faster queries
  - Fixed table prefix detection from SQL header comment
  - Added table prefix verification after database restore
- Performance optimizations
  - Chunked URL replacement prevents timeouts on large databases
  - Transient deletion loop limited to 100k iterations
  - URL replacement processes tables incrementally with state persistence
- Detailed migration progress indicators
  - Human-readable phase labels (e.g., "Restoring Database")
  - Progress shows actual data: bytes processed, tables imported, files copied
  - Finalize phase lists actions taken

### Added (Shared Components)
- `CacheManager` class for centralized cache operations with safeguards
- `count_tables_in_sql()` helper for accurate table counting

### Added (WP-CLI)
- `wp swish backup` command for creating backups from command line
- `wp swish import` command for importing/migrating backups via CLI
- `wp swish status` and `wp swish cleanup` utility commands
- CLI commands support all backup types: full, database, files

## [1.1.7] - 2026-05-12

### Fixed
- **Large file downloads now work with curl/wget** - moved download handler outside wp-admin
- Download URLs now use token-based authentication without requiring WordPress login
- Downloads support resumable transfers (`curl -C -` flag works correctly)
- **LiteSpeed X-LiteSpeed-Location disabled** - falls back to PHP streaming for reliable Range request support
- **Resumable downloads on GoDaddy/LiteSpeed + Cloudflare** - HTTP/2 stream resets no longer break downloads
- **Download tokens no longer deleted prematurely** - allows proper resume after connection failure

### Changed
- Download hook changed from `admin_init` to `init` (runs before WordPress admin auth)
- Download URLs changed from `/wp-admin/admin.php?...` to `/?swish_download=...`
- Removed `current_user_can()` check from download handler (token IS the authentication)
- LiteSpeed server detection returns false in `get_sendfile_method()` forcing PHP streaming fallback
- Download transient preserved until natural expiration (1 hour) instead of deleted after headers sent

## [1.1.6] - 2026-05-12

### Added
- REST API endpoints for schedule management (`/schedule/{id}/toggle`, `/schedule/{id}/run`, `/schedule/{id}/delete`)
- "Schedule Backup" button to empty state on Schedules page
- Scheduler dependency injection into RestController

### Fixed
- **Scheduled backups UI now fully functional** - toggle, run, and delete buttons work
- Backup job step indicators now show progress correctly (phase derived from progress percentage)
- JavaScript handlers now call actual API endpoints instead of stub implementations

## [1.1.5] - 2026-05-11

### Added
- **X-Sendfile/X-Accel-Redirect support** for large file downloads - bypasses PHP completely for file transfer
  - Apache: Uses `mod_xsendfile` when available
  - Nginx: Uses `X-Accel-Redirect` with configurable internal location
  - LiteSpeed: Uses `X-LiteSpeed-Location` (built-in support)
  - Automatic fallback to PHP streaming when not available
- HTTP Range request support for resumable backup downloads (4GB+ files)
- Chunked file streaming for downloads (8MB chunks) to avoid memory exhaustion
- `Accept-Ranges: bytes` header for browser download resume support
- New "Logs" admin page for viewing and managing backup logs
- Logging toggle (default: OFF) to enable/disable log collection
- Log level filter (All, Error, Warning, Info, Debug)
- Delete all logs functionality with file and database cleanup
- `Logger::is_enabled()` static method to check logging status

### Fixed
- Large backup downloads (4GB+) failing due to memory exhaustion or timeout
- Browser unable to resume interrupted backup downloads
- Migration import rejecting .tar.gz and .swish files (now accepts ZIP, TAR.GZ, SWISH)
- Migration analysis failing for .tar.gz backups (was only reading ZIP format)
- Restore process failing for .tar.gz backups (added PharData extraction support)
- **Files not being restored from streaming tar backups** - restore now handles direct file extraction (wp-content, wp-includes, etc. extracted directly instead of nested in files.tar.gz)
- **Backup verification failing for renamed files** - WordPress may rename uploaded files (e.g., `.tar.gz` → `.tar-2.gz`), now uses magic byte detection for reliable format identification
- **CRITICAL: Backups losing folder structure on Docker/symlink setups** - `get_relative_path()` now properly handles WP_CONTENT_DIR, WP_PLUGIN_DIR, theme root, and uploads when they resolve to different paths than ABSPATH (common in Docker, ddev, symlinked setups)

### Changed
- Download handler now streams files in chunks instead of using `readfile()`
- Download token preserved until full download completes (allows resume)
- Logger now checks if logging is enabled before writing entries

## [1.1.3] - 2026-05-11

### Added
- Chunked database backup with checkpoint support for large sites
- `DatabaseBackup::backup_chunked()` method for resumable database backups
- 25-second time slices for database backup (up from 10s, with max_execution_time=300)
- Database backup continuation support via `continue_database_backup()`
- Full progress tracking for database backup (table-by-table progress)
- Keyset pagination for database queries (O(1) instead of O(n) per batch)
- Primary key detection for automatic keyset pagination selection
- Multi-chunk processing: runs ~10 chunks per request before yielding to cron
- Extended INSERT syntax with multiple value tuples for fewer SQL statements

### Fixed
- Database backup timing out on large sites (100k+ rows in tables like `postmeta`)
- 77+ tables being skipped due to timeout on large WooCommerce/blog sites
- Backup appearing to complete but missing most database tables
- Database backup not resuming properly after timeout
- Checkpoint not saving database-specific fields (`db_file`, `db_checkpoint`, `backup_type`)
- Misleading "Async backup completed" log when continuation is scheduled
- WP-Cron not triggering continuation reliably (added fallback via status polling)
- Frontend showing stale "X/Y tables" message during large table processing
- Slow database backup due to OFFSET pagination causing O(n²) row scans
- Excessive cron overhead from yielding after every 25-second chunk

### Changed
- `run_full_backup_streaming()` now uses chunked database backup with multi-chunk loop
- `run_database_backup()` now uses chunked database backup with multi-chunk loop
- `continue_database_backup()` runs multiple chunks per request (up to 270s before yield)
- Database backup saves checkpoint after each table for reliable resumption
- Default batch size increased from 200 to 1000 rows per query (5x faster)
- Database queries use `WHERE pk > last_pk` instead of `OFFSET` for large tables
- Time slice increased from 10s to 25s per chunk for better throughput

## [1.1.2] - 2026-05-08

### Added
- New `StreamingTarBackup` class for true streaming file backups
- Pure PHP tar creation - works in all environments (Docker, ddev, managed hosts)
- 10-second time slices with automatic state persistence and continuation
- 1MB file chunk reads - never loads more than 1MB into memory
- USTAR tar format with extended header support for long filenames (>100 chars)
- Streaming gzip compression after tar completion

### Fixed
- Full backups getting stuck at "Backing up files..." on large sites (50,000+ files)
- Memory exhaustion during file backup in Docker/ddev environments
- `should_use_tar()` returning false in containerized environments forcing memory-heavy ZIP mode
- Empty files (0 bytes) causing `fread()` error during streaming backup
- Missing `BackupState::delete_meta()` method causing backup completion to fail

### Changed
- `BackupManager::run_full_backup()` now always uses streaming tar (no more ZIP batch mode)
- File backups process files one at a time instead of loading entire lists into memory
- Backup continuation now handles `streaming_files` phase for seamless resumption

## [1.0.18] - 2026-04-30

### Fixed
- Plugins not activating when importing a single site from a multisite backup
- Network-activated plugins now captured in backup manifest for multisite exports

## [1.0.17] - 2026-04-21

### Fixed
- Duplicate "Migrate Site" card on dashboard
- REST API download route not matching backup IDs with underscores (e.g., `backup_uuid`)

### Changed
- Pipeline Batch Size defaults: Shared hosting (150), VPS (250), Dedicated server (500)

## [1.0.16] - 2026-04-21

### Fixed
- Critical bug: ServerLimits timing check caused immediate yield due to int/float comparison (`0 === 0.0` is false in PHP 8)
- Files stuck in 'processing' status now reset to 'pending' for retry
- Reduced stale processing threshold from 300s to 30s for faster recovery
- Partial file resume corruption - files now restart from scratch instead of attempting resume

### Added
- Detailed logging for file processing: batch info, file status, yield reasons
- Context data now included in log file output for better debugging

### Changed
- Overall progress calculation now includes all phases (indexing 0-10%, processing 10-95%, finalizing 95-100%)

## [1.0.15] - 2026-04-21

### Added
- Archive format setting (Auto/ZIP/TAR.GZ) in settings page
- Streaming file scanning for large sites (writes to disk, not memory)
- Memory-efficient tar creation using `--files-from` for 50k+ files
- Docker/ddev environment detection with automatic ZIP fallback
- CPU/IO throttling with `nice` and `ionice` for tar operations

### Fixed
- Backup failing at 70% on large sites (2GB+) due to memory exhaustion
- Server crashes during archive creation in containerized environments
- Timeout issues during file staging step

### Changed
- Tar backups now skip staging step entirely (direct archive from source)
- File list written to disk instead of held in memory
- Fast GZIP compression (level 1) for better performance
- Auto-fallback to ZIP if tar fails or times out

## [1.0.12] - 2026-04-20

### Changed
- Dashboard now has three action boxes: Backup Now, Search & Replace, and Migrate Site
- Renamed "Migrate Site" panel to "Search & Replace" for clarity
- Added dedicated "Migrate Site" button linking to the migration page

## [1.0.11] - 2026-04-20

### Fixed
- Vendor folders are now included in backups (previously excluded, causing plugin errors after migration)

## [1.0.10] - 2026-04-20

### Added
- Memory-aware adaptive batch sizing for database and file backups
- Memory-aware adaptive batch sizing for URL search/replace during migration
- Automatic batch size reduction when memory pressure is detected
- Stream-based ZIP extraction to prevent memory exhaustion on large backups
- Memory threshold monitoring (32MB) to trigger adaptive processing
- User-friendly error messages for server limit errors (memory, timeout, upload size)
- Pre-import memory check with clear warning if insufficient memory available
- Shutdown handler to catch and report fatal errors during import/migration

### Fixed
- Memory exhaustion during import/restore of large backups (now uses streaming extraction)
- Import failing with `wp_handle_upload()` undefined function error
- Large file extraction causing PHP memory limit errors
- Fatal errors during import now show helpful messages instead of silent failures
- Connection reset during "Updating URLs" phase on memory-constrained servers
- Connection reset during "Finalizing" phase caused by pre-migration backup creation

### Changed
- ZIP extraction now reads files in 8KB chunks instead of loading entire files into memory
- Database backup automatically reduces batch size when memory is low
- File backup flushes ZIP archive more frequently under memory pressure
- File backup progress now shows percentage instead of time estimation
- URL search/replace now processes 100 rows per batch (reduced from 500)
- Transient cache clearing now uses batched deletion (1000 at a time)
- Pre-migration backup is now opt-in to avoid memory exhaustion during migration

## [1.0.9] - 2026-04-20

### Added
- Async backup processing to prevent timeouts on managed hosting (WP Engine, Flywheel, etc.)
- Background job processing via WP Cron with automatic fallback for hosts with alternate cron
- Job status polling endpoint (`/job/{id}/process`) for triggering pending backups directly
- Real-time progress updates during backup process
- ETA (estimated time remaining) display during file backup showing files processed and time remaining

### Fixed
- Backup download returning "0" - changed from admin-ajax.php to admin.php endpoint
- Added proper download handler with token validation and security checks

### Changed
- Backup API now uses async processing by default
- Frontend polling improved with automatic fallback when WP Cron doesn't trigger

## [1.0.8] - 2026-04-15

### Fixed
- Replaced inline `<script>` tag with `wp_add_inline_script()` for WordPress.org compliance
- Replaced `move_uploaded_file()` with `wp_handle_upload()` for proper WordPress file handling

## [1.0.7] - 2026-02-25

### Changed
- Compatibility update for Pro plugin import fixes
- Minor stability improvements

## [1.0.6] - 2026-02-25

### Changed
- Compatibility update for Pro plugin URL auto-detection fixes
- Minor stability improvements

## [1.0.5] - 2026-02-25

### Changed
- Compatibility update for Pro plugin size estimation fixes
- Minor stability improvements

## [1.0.4] - 2026-02-25

### Changed
- Compatibility update for Pro plugin import/migration fixes
- Minor stability improvements

## [1.0.3] - 2026-02-25

### Added
- Auto-detection of old site URL during migration (pre-filled from backup manifest)
- Detailed migration progress with stage tracking similar to backup process
- Smooth scrolling and active state highlighting for documentation navigation
- REST endpoint for importing backup files (`/swish-backup/v1/import`)

### Changed
- Moved backup storage location from `wp-content/uploads/swish-backups/` to `wp-content/swish-backups/`
- Improved documentation page with better anchor link navigation

### Fixed
- Import feature Continue button now properly uploads and analyzes backup files
- Backup exclusion now always excludes swish-backups folder from backups
- Prevents backups from including previous backup archives

## [1.0.2] - 2025-02-25

### Added
- Enhanced backup progress modal with detailed stage tracking
- Visual progress log showing each backup stage as it completes
- Green checkmarks for completed stages
- Red indicators for failed stages
- Animated spinner for in-progress stages

### Changed
- Improved progress feedback during backup process
- Better error handling and messaging

### Fixed
- Minor UI improvements for progress display

## [1.0.1] - 2025-02-20

### Added
- Option to include/exclude WordPress core files in backups

### Changed
- Enhanced CSS for admin layout with responsive design
- Refactored JavaScript asset enqueuing for improved performance

### Fixed
- Minor bug fixes and improvements

## [1.0.0] - 2025-02-15

### Added
- Initial release
- Full site backup support (database, plugins, themes, uploads)
- Database-only backup option
- Files-only backup option
- Local storage adapter
- Amazon S3 storage adapter
- Dropbox storage adapter
- Google Drive storage adapter
- Site migration with automatic URL replacement
- Search and replace functionality with serialization support
- Scheduled backups (hourly, daily, weekly, monthly)
- Chunked processing for large sites
- AES-256-CBC encrypted cloud storage credentials
- Full REST API for integration
