d6776f9f57
The details backfill queue is keyed on tmdb_id, but the gate (hasattr match) let the
TVDB worker run it too — feeding TMDB ids to TVDB's /series/{id}/extended (→ 404 on
every show) and double-processing each show (TMDB backfilled it, TVDB then 404'd but
still logged 'Backfilled'). Gate on self.service=='tmdb' so only the TMDB worker runs
it. Regression test: the TVDB worker no-ops (never calls its client, leaves the item
pending). 95 enrichment tests pass.
340 lines
17 KiB
Python
340 lines
17 KiB
Python
"""Video enrichment worker — one per source (TMDB, TVDB).
|
|
|
|
Mirrors the music worker: a daemon loop that pulls the next item needing
|
|
enrichment from video.db, asks its CLIENT to match it, and records the result.
|
|
The client is injected (a thin TMDB/TVDB adapter), so the worker's loop/queue/
|
|
status logic is fully testable with a fake client. Isolated: imports only
|
|
video.db helpers; no music code.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import threading
|
|
import time
|
|
|
|
from utils.logging_config import get_logger
|
|
|
|
logger = get_logger("video_enrichment.worker")
|
|
|
|
|
|
class VideoEnrichmentWorker:
|
|
def __init__(self, db, service, client, display_name=None, interval=2.0, retry_days=30):
|
|
self.db = db
|
|
self.service = service
|
|
self.client = client
|
|
self.display_name = display_name or service.upper()
|
|
self.interval = interval
|
|
self.retry_days = retry_days
|
|
|
|
# OMDb is a ratings filler, not a matcher — it fetches scores by imdb_id
|
|
# instead of running a match queue.
|
|
self.is_ratings = hasattr(client, "ratings") and not hasattr(client, "match")
|
|
|
|
self.running = False
|
|
self.paused = False
|
|
self.should_stop = False
|
|
self._thread = None
|
|
self._stop = threading.Event()
|
|
self.current_item = None
|
|
self.stats = {"matched": 0, "not_found": 0, "errors": 0}
|
|
self.note = None # a human reason when auto-paused (e.g. bad key)
|
|
self._rating_errors = 0 # consecutive ratings failures (transient backoff)
|
|
self._cooldown_until = 0.0 # monotonic time to idle until (daily-limit backoff)
|
|
|
|
# ── lifecycle ─────────────────────────────────────────────────────────────
|
|
def start(self):
|
|
if self.running:
|
|
return
|
|
self.should_stop = False
|
|
self._stop.clear()
|
|
self.running = True
|
|
self._thread = threading.Thread(target=self._run, daemon=True)
|
|
self._thread.start()
|
|
|
|
def stop(self):
|
|
self.should_stop = True
|
|
self._stop.set()
|
|
if self._thread:
|
|
self._thread.join(timeout=1.0)
|
|
self.running = False
|
|
|
|
def pause(self, persist=True):
|
|
self.paused = True
|
|
if persist:
|
|
self._persist_paused()
|
|
|
|
def resume(self, persist=True):
|
|
self.paused = False
|
|
if persist:
|
|
self._persist_paused()
|
|
|
|
def _persist_paused(self):
|
|
# Survives restart, like music's <service>_enrichment_paused config flag.
|
|
try:
|
|
self.db.set_setting(self.service + "_paused", "1" if self.paused else "0")
|
|
except Exception:
|
|
logger.exception("video enrichment: could not persist pause for %s", self.service)
|
|
|
|
def restore_paused(self):
|
|
try:
|
|
self.paused = str(self.db.get_setting(self.service + "_paused") or "") == "1"
|
|
except Exception:
|
|
logger.exception("video enrichment: could not restore pause for %s", self.service)
|
|
|
|
@property
|
|
def enabled(self):
|
|
return bool(getattr(self.client, "enabled", False))
|
|
|
|
# ── loop ──────────────────────────────────────────────────────────────────
|
|
def _run(self):
|
|
while not self.should_stop:
|
|
if self.paused or not self.enabled:
|
|
self._stop.wait(1.0)
|
|
continue
|
|
# Daily-limit cooldown: idle until the quota resets, then auto-resume
|
|
# (re-checks periodically so we pick back up shortly after midnight UTC).
|
|
if self._cooldown_until > time.monotonic():
|
|
self.current_item = None
|
|
self._stop.wait(15.0)
|
|
continue
|
|
try:
|
|
did = self.process_one()
|
|
except Exception:
|
|
logger.exception("video enrichment %s loop error", self.service)
|
|
self.stats["errors"] += 1
|
|
self._stop.wait(5.0)
|
|
continue
|
|
if did:
|
|
self._stop.wait(self.interval) # rate-limit between items
|
|
else:
|
|
self.current_item = None
|
|
self._stop.wait(10.0) # nothing to do — back off
|
|
|
|
def process_one(self) -> bool:
|
|
"""Process a single item. Returns True if one was processed."""
|
|
if self.is_ratings:
|
|
return self._process_ratings_one()
|
|
priority = None
|
|
try:
|
|
priority = self.db.get_setting("enrichment_priority") or None
|
|
except Exception:
|
|
logger.debug("enrichment priority lookup failed", exc_info=True)
|
|
item = self.db.enrichment_next(self.service, self.retry_days, priority=priority)
|
|
if not item:
|
|
# No pending matches → use idle time for the full episode-list sync
|
|
# (which also fills details for shows it touches), then the details
|
|
# backfill for everything already episode-synced but missing `status`.
|
|
return self._sync_episodes_once() or self._detail_backfill_one()
|
|
self.current_item = {"type": item["kind"], "name": item["title"]}
|
|
try:
|
|
# Prefer the provider id the server already gave us (enrich BY ID, no
|
|
# re-search); the client falls back to a title/year search if it's None.
|
|
result = self.client.match(item["kind"], item["title"], item.get("year"),
|
|
known_id=item.get("known_id"))
|
|
except Exception:
|
|
logger.exception("video enrichment %s match failed for %s", self.service, item["title"])
|
|
self.stats["errors"] += 1
|
|
# The CALL failed (network/rate-limit/timeout) — record 'error', NOT
|
|
# 'not_found', so a transient blip isn't permanently logged as "no
|
|
# match". enrichment_next retries 'error' items after retry_days.
|
|
self.db.enrichment_apply(self.service, item["kind"], item["id"], matched=False, error=True)
|
|
return True
|
|
if result and result.get("id"):
|
|
self.db.enrichment_apply(self.service, item["kind"], item["id"], matched=True,
|
|
external_id=result["id"], metadata=result.get("metadata"))
|
|
self.stats["matched"] += 1
|
|
# Visible progress in app.log, mirroring the music workers' style.
|
|
logger.info("Matched %s '%s' -> %s ID: %s%s", item["kind"], item["title"],
|
|
self.display_name, result["id"],
|
|
" (by server id)" if item.get("known_id") else "")
|
|
# Cascade: a matched show backfills its episodes' art/overview/rating
|
|
# from the same provider (one call per season), so episodes ride along
|
|
# with their show instead of being a separate (huge) queue.
|
|
if item["kind"] == "show" and hasattr(self.client, "season_episodes"):
|
|
nums = [s["season_number"] for s in (result.get("metadata") or {}).get("seasons") or []]
|
|
self._cascade_episodes(item["id"], result["id"], nums)
|
|
else:
|
|
self.db.enrichment_apply(self.service, item["kind"], item["id"], matched=False)
|
|
self.stats["not_found"] += 1
|
|
logger.info("No %s match for %s '%s'", self.display_name, item["kind"], item["title"])
|
|
return True
|
|
|
|
def _process_ratings_one(self) -> bool:
|
|
"""OMDb worker: fetch IMDb/RT/Metacritic for the next library item that has
|
|
an imdb_id but no ratings yet."""
|
|
from .clients import OMDbAuthError
|
|
item = self.db.ratings_next()
|
|
if not item:
|
|
self._rating_errors = 0
|
|
return False
|
|
self.current_item = {"type": item["kind"], "name": item["title"]}
|
|
try:
|
|
r = self.client.ratings(item["imdb_id"])
|
|
except OMDbAuthError as e:
|
|
msg = str(e)
|
|
if "limit" in msg.lower():
|
|
# Free-tier daily quota (1,000/req) — resets at midnight UTC. Cool
|
|
# down and AUTO-RESUME (so a big library just spreads across days)
|
|
# rather than pausing for good. Item not marked synced → retried.
|
|
if self._cooldown_until <= time.monotonic():
|
|
logger.warning("OMDb daily request limit reached — idling ratings ~30 min; "
|
|
"auto-resumes after the daily reset")
|
|
self._cooldown_until = time.monotonic() + 1800
|
|
self.note = "OMDb daily limit reached — resumes after reset"
|
|
return False
|
|
# Bad/expired/unactivated key affects EVERY item — pause instead of
|
|
# churning the whole library + flooding the log. Transient pause (not
|
|
# persisted) so fixing the key (which rebuilds the engine) resumes
|
|
# automatically. The item is NOT marked synced, so it retries once
|
|
# the key works.
|
|
if not self.paused:
|
|
logger.warning("OMDb rejected the API key — pausing ratings until it's fixed (%s)", msg)
|
|
self.note = "OMDb API key rejected"
|
|
self.pause(persist=False)
|
|
return False
|
|
except Exception as e:
|
|
# Transient (network / rate-limit / 5xx) — don't burn the item to
|
|
# 'synced'; back off, and pause after a few in a row so we don't spin.
|
|
self.stats["errors"] += 1
|
|
self._rating_errors = getattr(self, "_rating_errors", 0) + 1
|
|
logger.warning("OMDb ratings fetch failed for '%s': %s", item["title"], e)
|
|
if self._rating_errors >= 3:
|
|
logger.warning("OMDb: pausing ratings after repeated errors")
|
|
self.note = "OMDb temporarily unavailable"
|
|
self.pause(persist=False)
|
|
self._rating_errors = 0
|
|
return False
|
|
self._rating_errors = 0
|
|
self._cooldown_until = 0.0
|
|
self.note = None
|
|
if r:
|
|
self.db.apply_ratings(item["kind"], item["id"], r) # marks synced
|
|
self.stats["matched"] += 1
|
|
logger.info("Rated %s '%s' -> IMDb %s", item["kind"], item["title"], item["imdb_id"])
|
|
else:
|
|
self.db.mark_ratings_synced(item["kind"], item["id"]) # genuine no-data
|
|
self.stats["not_found"] += 1
|
|
return True
|
|
|
|
def _sync_episodes_once(self) -> bool:
|
|
"""Background episode-sync: pull the FULL season/episode list for one
|
|
already-matched show that hasn't been synced, so library cards show real
|
|
owned/total. TMDB-only (it owns season_episodes). Returns True if it did
|
|
work (so the loop rate-limits between shows)."""
|
|
if not hasattr(self.client, "season_episodes"):
|
|
return False
|
|
show = self.db.episode_sync_next()
|
|
if not show:
|
|
return False
|
|
self.current_item = {"type": "episodes", "name": show["title"]}
|
|
try:
|
|
result = self.client.match("show", show["title"], show.get("year"),
|
|
known_id=show.get("tmdb_id"))
|
|
if result and result.get("id"):
|
|
self.db.enrichment_apply("tmdb", "show", show["id"], matched=True,
|
|
external_id=result["id"], metadata=result.get("metadata"))
|
|
nums = [s["season_number"] for s in (result.get("metadata") or {}).get("seasons") or []]
|
|
self._cascade_episodes(show["id"], result["id"], nums) # marks synced
|
|
logger.info("Synced full episode list for show '%s'", show["title"])
|
|
else:
|
|
self.db.mark_episodes_synced(show["id"]) # no match → don't re-pick
|
|
except Exception:
|
|
logger.exception("episode sync failed for show '%s'", show["title"])
|
|
self.db.mark_episodes_synced(show["id"]) # move on (never loop on one show)
|
|
return True
|
|
|
|
def _detail_backfill_one(self) -> bool:
|
|
"""Background TMDB details backfill: re-fetch ONE already-matched show/movie
|
|
that's missing details-only fields (status, network, tagline, rating…) and
|
|
gap-fill them. The matcher skips server-pre-matched items, so without this
|
|
their `status` stays blank — and the watchlist's airing-default can't see
|
|
them. Returns True if it did work (so the loop rate-limits).
|
|
|
|
TMDB-ONLY: `status` (+ network/tagline/…) come from TMDB and the queue is
|
|
keyed on tmdb_id. Running it on the TVDB worker would feed a TMDB id to
|
|
TVDB (→ 404 on every show) and double-process the queue."""
|
|
if self.service != "tmdb" or not hasattr(self.client, "match"):
|
|
return False
|
|
item = self.db.detail_backfill_next("show") or self.db.detail_backfill_next("movie")
|
|
if not item:
|
|
return False
|
|
self.current_item = {"type": item["kind"], "name": item["title"]}
|
|
try:
|
|
result = self.client.match(item["kind"], item["title"], item.get("year"),
|
|
known_id=item.get("tmdb_id"))
|
|
if result and result.get("id"):
|
|
# Gap-fill only (never clobbers server data); fills status et al.
|
|
self.db.enrichment_apply("tmdb", item["kind"], item["id"], matched=True,
|
|
external_id=result["id"], metadata=result.get("metadata"))
|
|
logger.info("Backfilled details for %s '%s'", item["kind"], item["title"])
|
|
self.db.mark_details_synced(item["kind"], item["id"]) # attempted once → don't re-pick
|
|
except Exception:
|
|
# Transient call failure — leave details_synced=0 so it retries later.
|
|
logger.exception("detail backfill failed for %s '%s'", item["kind"], item["title"])
|
|
self.stats["errors"] += 1
|
|
return True
|
|
|
|
def _cascade_episodes(self, show_id, tv_id, season_numbers=None) -> None:
|
|
"""Backfill a show's FULL episode list from the provider (one call per
|
|
season) — owned + missing. Best-effort: a season failure never aborts the
|
|
show's enrichment. Falls back to the known seasons if none are passed."""
|
|
seasons = season_numbers
|
|
if not seasons:
|
|
try:
|
|
seasons = self.db.show_season_numbers(show_id)
|
|
except Exception:
|
|
logger.exception("episode backfill: season list failed for show %s", show_id)
|
|
return
|
|
for snum in seasons:
|
|
try:
|
|
data = self.client.season_episodes(tv_id, snum)
|
|
if data and data.get("episodes"):
|
|
self.db.backfill_episodes(show_id, snum, data["episodes"],
|
|
data.get("overview"), data.get("poster_url"))
|
|
except Exception:
|
|
logger.exception("episode backfill failed: show %s season %s", show_id, snum)
|
|
try:
|
|
self.db.mark_episodes_synced(show_id)
|
|
except Exception:
|
|
logger.exception("episode backfill: could not mark synced for show %s", show_id)
|
|
|
|
# ── status (same shape the music enrichment API returns) ──────────────────
|
|
def get_stats(self) -> dict:
|
|
breakdown = self.db.enrichment_breakdown(self.service)
|
|
# Errored items are outstanding (retried later), so they count as pending
|
|
# work — the worker isn't "Complete" while any remain. Episode art is a
|
|
# coverage-only cascade (no queue), so it's excluded from idle/pending.
|
|
pending = sum(b["pending"] + b.get("errors", 0)
|
|
for b in breakdown.values() if not b.get("coverage_only"))
|
|
# Shows still needing their full episode list pulled count as outstanding
|
|
# work for the TMDB worker (so it isn't "Complete" while syncing).
|
|
if hasattr(self.client, "season_episodes"):
|
|
try:
|
|
pending += self.db.episode_sync_pending_count()
|
|
except Exception:
|
|
logger.debug("episode_sync_pending_count failed", exc_info=True)
|
|
# NB: the TMDB details backfill (status/network/…) is a background gap-fill on
|
|
# ALREADY-matched items — like episode coverage, it doesn't block "Complete".
|
|
# It still runs in the idle loop; it's just not counted as blocking pending.
|
|
cooling = self._cooldown_until > time.monotonic()
|
|
running = self.running and not self.paused and self.enabled and not cooling
|
|
idle = running and pending == 0 and self.current_item is None
|
|
progress = {}
|
|
for kind, b in breakdown.items():
|
|
total = b["matched"] + b["not_found"] + b.get("errors", 0) + b["pending"]
|
|
done = b["matched"] + b["not_found"]
|
|
progress[kind] = {"matched": b["matched"], "total": total,
|
|
"percent": round(done / total * 100) if total else 0}
|
|
return {
|
|
"enabled": self.enabled,
|
|
"needs_key": True, # matchers (TMDB/TVDB/OMDb) always require an API key
|
|
"running": running,
|
|
"paused": self.paused or cooling, # cooldown reads as paused in the UI
|
|
"idle": idle,
|
|
"current_item": self.current_item,
|
|
"note": self.note,
|
|
"cooldown": cooling,
|
|
"stats": {**self.stats, "pending": pending},
|
|
"progress": progress,
|
|
"breakdown": breakdown,
|
|
}
|