drawer upgrades (the TMDB call we already make returns all this — just render it):
- FIX cast photos — used c.profile_url; the field is c.photo. now shows headshots.
- title LOGO header (falls back to text) + meta line (year · ⭐rating · runtime · network).
- tagline, director/creator, ▶ trailer link, and a 'Watch on' provider-logo row.
- meta endpoint widened to return logo/cast-photo/rating/runtime/crew/trailer/providers.
OMDb spam fix (the 'Request limit reached!' tracebacks):
- the new bulk schedule-refresh called refresh_show_art per show, which did an OMDb
ratings backfill each → blew the daily quota. refresh_show_art gains with_ratings;
the automation passes False (it only needs episode schedules).
- _backfill_ratings now latches _omdb_blocked on OMDbAuthError → one quiet warning +
stops calling OMDb for the rest of the process, instead of a traceback per show.
tested (cast/logo/trailer contract, no-ratings plumbing, the latch-off behavior).
the airing automation reads the LOCAL episodes table, which only refreshes on initial
enrichment / manual re-match / lazy on-view — so a newly-announced or rescheduled
episode could be missed indefinitely. (the deep TV scan is a SERVER scan, unrelated.)
new daily automation re-pulls TMDB episode schedules (air dates/stills) for still-airing
watchlist shows so the calendar is fresh when the airing run reads it:
- db.watchlist_continuing_shows(): effective-watchlist library shows that are still airing
(skips tmdb-only follows — no episodes — and ended/canceled shows; keeps unknown status).
- handler mirrors the standard: injected fetch_shows + refresh_show seams, live per-show
progress, _manages_own_progress; reuses engine.refresh_show_art (match + episode cascade).
- seeded daily at 23:00 (2h before the 01:00 airing run), registered, UI block, sorted in
the page right before the airing automation.
seam-level tested + a regression that it's seeded before the airing run.
