Skip to content

Data Storage

Apple Juicer persists its state across three layers: the filesystem, PostgreSQL, and Redis. Understanding how each layer is used helps with operations, scaling, and recovery.

Filesystem Mounts

  • Backup data – Finder/iTunes backups remain on disk and are mounted read-only into backend + worker containers (default ./data:/data/ios_backups). No payloads are copied unless a user downloads a file.
  • Temporary sandbox – Decrypted files extracted for download or parsing are staged under backup_paths.temp_path (default /tmp/apple_juicer). A Starlette BackgroundTask cleans up these folders after streaming. (@api/routes/backups.py#122-141)

PostgreSQL Schema

Core Tables

  • backups – canonical list of discovered backups, including device metadata, size, encryption state, and status lifecycle from DISCOVERED → INDEXING → INDEXED. (@core/db/models.py#14-37)
  • Artifact tables (photo_assets, whatsapp_*, messages, notes, calendar_*, contacts) – normalized views of SQLite artifacts parsed by the worker. Each row links back to backups.id with ON DELETE CASCADE. (@core/db/artifacts.py#13-197)
  • artifact_search_index – denormalized text search surface combining display text + structured payload per artifact. Useful for global queries. (@core/db/artifacts.py#186-197)

Sessions & Unlock Metadata

Session state is maintained by UnlockManager in memory, but the BackupRegistry tracks last unlock/index operations in Postgres. When you rescan (POST /backups/refresh), new entries are upserted via SQLAlchemy. (@worker/tasks.py#40-90)

Engine Configuration

The async SQLAlchemy engine connects via APPLE_JUICER_POSTGRES__DSN, defaulting to sqlite+aiosqlite:///./temp_data/apple_juicer.db out of the box. Development builds can call core.db.session.init_models() to bootstrap tables without Alembic. (@core/db/session.py#1-24)

Redis

Redis serves two roles:

  1. RQ Queue – core.queue.get_queue() creates a cached Redis connection and defines the default queue with a 10-minute timeout. (@core/queue.py#11-24)
  2. Worker coordination – RQ stores job metadata, heartbeats, and result payloads here. Set APPLE_JUICER_REDIS__URL to point at your Redis deployment.

Redis does not store session tokens or decrypted data; those remain in-memory within the backend/worker processes.

Backup & Restore Strategy

  1. Filesystem – treat the mounted backup directory as authoritative; snapshot or rsync from there.
  2. Postgres – use native tools (pg_dump, pg_basebackup) to preserve registry + artifact indexes.
  3. Redis – optional to persist; you can rebuild queues by re-triggering indexing runs if Redis data is lost.

Because decrypted payloads are never persisted in the database, a restore requires only re-pointing to the backup directory and optionally re-indexing artifacts.