soulsync/database
BoulderBadgeDad 402a1fec50 video side: video.db schema + isolated VideoDatabase
Separate SQLite file (database/video_library.db, env VIDEO_DATABASE_PATH),
fully disconnected from music — never imported by music, imports nothing from
music, no shared write lock.

Schema (database/video_schema.sql), designed to dodge the music DB's known
pain points:
- movies; shows->seasons->episodes; channels->channel_videos (YouTube as a
  first-class peer); media_files (the library); downloads (queue+history);
  activity feed; root_folders / quality_profiles / video_settings config.
- No polymorphic ids: media_files/downloads use separate nullable FKs + a CHECK
  that exactly one owner is set; real cascades.
- Explicit external-id columns (tmdb/tvdb/imdb/youtube), no source-id blob.
- Watchlist/Wishlist/Calendar are DERIVED VIEWS over monitored + file state
  (single source of truth, can't drift like music's wishlist table did).

VideoDatabase mirrors music's conventions (WAL, foreign_keys ON, 30s busy
timeout, Row factory, once-per-process init, user_version backstop) but is an
independent implementation. 13 seam tests: schema builds, CHECK constraints
reject bad rows, cascades fire, views return correct membership, KV roundtrips,
and a guard that the module imports nothing from music.
2026-06-13 22:30:55 -07:00
..
__init__.py basic db structure 2025-08-05 22:43:04 -07:00
music_database.py M3U export: resolve paths via one bulk read instead of a per-artist search loop (fixes 'Export M3U hangs forever' under active enrichment/scan DB writes) 2026-06-13 08:55:46 -07:00
personalized_schema.py Personalized pipeline: auto-refresh stale snapshots after watchlist scan 2026-05-15 20:53:03 -07:00
video_database.py video side: video.db schema + isolated VideoDatabase 2026-06-13 22:30:55 -07:00
video_schema.sql video side: video.db schema + isolated VideoDatabase 2026-06-13 22:30:55 -07:00