app/api/scenes.py

@@ -1851,7 +1851,7 @@ def enrich_tags_from_tube(
     from app.models.tag import Tag
     from app.normalize.scenes import NormalizedTag
     from app.normalize.text import slugify
-    from app.resolve.tag_resolver import resolve_tag, resolve_tags_by_altkey
+    from app.resolve.tag_resolver import resolve_tag
 
     scene = session.get(Scene, scene_id)
     if scene is None:
@@ -1923,17 +1923,10 @@ def enrich_tags_from_tube(
         {t.slug: t for t in session.execute(select(Tag).where(Tag.slug.in_(direct))).scalars()}
         if direct else {}
     )
-    # Nazwy spoza slugow: wariant pisowni istniejacego tagu (alnum-klucz, liczba mnoga)
-    # tez jednym zapytaniem. resolve_tag zostaje tylko dla naprawde nowych nazw.
-    direct_set = set(direct)
-    by_alt = resolve_tags_by_altkey(
-        session,
-        [n for n in tag_names if slugify(n) in direct_set and slugify(n) not in by_slug],
-    )
     tag_ids: list = []
     for name in tag_names:
         slug = slugify(name)
-        tag = by_slug.get(slug) or by_alt.get(name)
+        tag = by_slug.get(slug)
         if tag is None:
             tag = resolve_tag(session, norm=NormalizedTag(name=name, slug=slug, external_id=None))
         if tag is not None and tag.id not in tag_ids:

app/api/sources.py

@@ -110,7 +110,7 @@ class SourceListOut(BaseModel):
 # tubey mają nietypowe TLDs / brakujące kropki w sitetag.
 _DISPLAY_OVERRIDES: dict[str, str] = {
     "fpoxxx": "fpo.xxx",
-    "siskavideo": "siska.tv",
+    "siskavideo": "siska.video",
     "porn4dayspw": "porn4days.pw",
     "porn00org": "porn00.org",
     "freshpornoorg": "freshporno.org",

app/connectors/direct_scrapers/siska.py

@@ -1,8 +1,8 @@
-"""siska.tv (do 2026-09-22 siska.video), latest-vids browse scraper.
+"""siska.video, latest-vids browse scraper.
 
 Historia: dawniej performer-driven search scraper (`?s=<q>`), ale siska zepsuła
 wyszukiwarkę (zwraca latest niezależnie od query). Przerobione na BROWSE (latest
-chronologicznie z listingu, patrz `_listing_url`), re-enabled 2026-06-20 (user fa4083a2).
+chronologicznie z `/page/<n>/`), re-enabled 2026-06-20 (user fa4083a2).
 
 Cały blok kafelka listingu ma komplet metadanych (zero detail-fetchy):
   <a title='<Tytuł>' href='https://siska.video/video.php?videoID=<n>'>
@@ -36,18 +36,11 @@ from app.extractors import browser_get
 
 log = logging.getLogger(__name__)
 
-# Domena siska.video wygasla 2026-09-22 (NXDOMAIN), ten sam serwis dziala dalej pod
+_BASE = "https://siska.video"
-# siska.tv: te same numery videoID i sciezki miniatur. Adresy w bazie przepisane na
-# siska.tv tego samego dnia. Kafelki przyjmujemy z obu domen i z www, ale zapisujemy
-# zawsze jeden kanoniczny adres (_canonical), bo external_id sceny i unikalny klucz
-# zrodla to sam adres; www.siska.video i siska.video dawaly 103 zdublowane zrodla.
-_BASE = "https://siska.tv"
-_HOST_RE = re.compile(r"^https://(?:www\.)?siska\.(?:tv|video)/", re.IGNORECASE)
 
 # Kafelek: <a title='..' href='..videoID=N'>. Reszta pól w oknie po tym matchu.
 _A_RE = re.compile(
-    r"<a\s+title='(?P<title>[^']*)'\s+"
+    r"<a\s+title='(?P<title>[^']*)'\s+href='(?P<url>https://siska\.video/video\.php\?videoID=\d+)'",
-    r"href='(?P<url>https://(?:www\.)?siska\.(?:tv|video)/video\.php\?videoID=\d+)'",
     re.IGNORECASE,
 )
 _DUR_RE = re.compile(r"th_video_duration'>\s*([\d]{1,2}(?:\s*:\s*[\d]{2}){1,2})\s*<")
@@ -56,10 +49,6 @@ _ALT_RE = re.compile(r"alt='([^']*)'")
 _CAT_RE = re.compile(r"/category/([^/]+)/", re.IGNORECASE)
 
 
-def _canonical(url: str) -> str:
-    return _HOST_RE.sub(_BASE + "/", url)
-
-
 def _parse_duration(text: str | None) -> int | None:
     """`40 : 27` → 2427 (MM:SS); `1 : 05 : 30` → H:MM:SS. None gdy brak."""
     if not text:
@@ -84,13 +73,11 @@ class SiskaScraper(BaseBrowseScraper):
     sitetag = "siskavideo"
 
     def _listing_url(self, page: int) -> str:
-        # Na siska.tv `/page/<n>/` oddaje zawsze strone 1. Najnowsze po kolei (18 na
+        return f"{_BASE}/page/{page}/"
-        # strone, videoID malejaco) sa pod `best_xvideos.php`, mimo nazwy.
-        return f"{_BASE}/best_xvideos.php?page={page}"
 
     # crawl_page nadpisany → poniższe abstrakcje nieużywane, ale wymagane do instancji.
     def _extract_scene_urls(self, listing_html: str) -> list[str]:
-        return [_canonical(m.group("url")) for m in _A_RE.finditer(listing_html)]
+        return [m.group("url") for m in _A_RE.finditer(listing_html)]
 
     def _parse_detail(self, scene_url: str, detail_html: str) -> RawScene | None:
         return None
@@ -107,7 +94,7 @@ class SiskaScraper(BaseBrowseScraper):
         out: list[RawScene] = []
         seen: set[str] = set()
         for m in _A_RE.finditer(html):
-            scene_url = _canonical(m.group("url"))
+            scene_url = m.group("url")
             if scene_url in seen:
                 continue
             seen.add(scene_url)
@@ -119,7 +106,7 @@ class SiskaScraper(BaseBrowseScraper):
             dm = _DUR_RE.search(window)
             duration_sec = _parse_duration(dm.group(1) if dm else None)
             tm = _THUMB_RE.search(window)
-            thumbnail_url = _canonical(tm.group(1)) if tm else None
+            thumbnail_url = tm.group(1) if tm else None
 
             # alt='Performer - Tytuł - Studio' → performer (pierwszy), studio (ostatni).
             performers: list[RawPerformer] = []

app/connectors/yourdailypornmovies.py

@@ -51,19 +51,14 @@ log = logging.getLogger(__name__)
 
 _BASE = "https://yourdailypornmovies.ws"
 
-# Karta listingu: okładka + URL + tytuł + rok. Okładka jest opcjonalna i w dowolnym
+# Karta listingu: id produktu Adult Empire w miniaturce + URL + tytuł + rok.
-# formacie: od ~2026-09-15 nowe filmy mają `imgs4cdn.adultempire.com/movie/<id>?w=...`
-# zamiast `.../products/NN/<id>h.jpg`, część ma okładkę z i.ibb.co. Regex wymagający
-# starego formatu po cichu gubił 33 z 40 filmów strony 1 (ingest: 0 nowych przez tydzień).
 _CARD_RE = re.compile(
     r'<div class="movie"><div class="imagen">\s*'
-    r'(?:<img src="(?P<poster>[^"]*)"[^>]*>\s*)?'
+    r'<img src="(?P<poster>https://imgs\d*cdn\.adultempire\.com/products/\d+/(?P<ae>\d+)h\.jpg)"'
-    r'<a href="(?P<url>https://yourdailypornmovies\.ws/[^"]+)">.*?'
+    r'[^>]*>\s*<a href="(?P<url>https://yourdailypornmovies\.ws/[^"]+)">.*?'
     r'</div><h2>(?P<title>[^<]+)</h2>\s*<span class="year">(?P<year>\d{4})</span>',
     re.DOTALL,
 )
-# Id produktu Adult Empire z adresu okładki, w obu formatach (to ta sama numeracja).
-_AE_ID_RE = re.compile(r"adultempire\.com/(?:products/\d+/(\d+)h\.jpg|movie/(\d+))")
 
 # Blok metadanych filmu. WSZYSTKO czytamy TYLKO stąd, patrz pułapka w docstringu.
 _XMLL_RE = re.compile(r'<div class="xmll">.*?</div>', re.DOTALL)
@@ -213,21 +208,18 @@ class YourDailyPornMoviesConnector(BaseMovieConnector):
             desc = html_mod.unescape(om.group(1)).strip() or None
 
         slug = urlparse(url).path.strip("/").split("/")[-1]
-        poster = html_mod.unescape(card.group("poster") or "").strip() or None
-        aem = _AE_ID_RE.search(poster or "")
-        ae_id = (aem.group(1) or aem.group(2)) if aem else None
         return RawMovie(
             external_id=slug,
             title=title,
             description=desc,
             release_year=year,
             url=url,
-            poster_url=poster,
+            poster_url=card.group("poster"),
             studio=studio,
             performers=performers,
             tags=tags,
             playback_sources=playback,
             # id produktu Adult Empire, mocniejsza kotwica do dedupu niż tytuł
             # (przy ocenie źródła złapało 9 filmów, których tytuł się nie zgadzał).
-            raw={"source": self.name, "url": url, "adultempire_id": ae_id},
+            raw={"source": self.name, "url": url, "adultempire_id": card.group("ae")},
         )

app/resolve/tag_resolver.py

@@ -29,43 +29,35 @@ def _resolve_by_altkey(session: Session, name: str) -> Tag | None:
     scene_count wygrywa). Zapobiega regeneracji duplikatów: nowy wariant "Deep Throat"
     trafia do istniejącego "deepthroat" zamiast tworzyć osobny tag. Używa indeksu
     funkcyjnego ix_tags_name_altkey."""
-    return resolve_tags_by_altkey(session, [name]).get(name)
+    key = _altkey(name)
-
+    if not key:
-
+        return None
-def resolve_tags_by_altkey(session: Session, names) -> dict[str, Tag]:
+    expr = func.regexp_replace(func.lower(func.btrim(Tag.name)), "[^a-z0-9]", "", "g")
-    """`_resolve_by_altkey` dla wielu nazw jednym zapytaniem: {nazwa: tag} dla trafień.
+    hit = (
-
+        session.execute(select(Tag).where(expr == key).order_by(Tag.scene_count.desc()))
-    Po zapytaniu na nazwę enrich-tags robił N+1 przy kilku nowych nazwach ze strony
+        .scalars()
-    tuba (Sentry GOON-5F).
+        .first()
-    """
+    )
+    if hit is not None:
+        return hit
     # Liczba mnoga. Alnum-klucz z natury jej NIE łapie ("blowjob" != "blowjobs"), więc
     # duplikaty odrastały mimo tamtej prewencji (audyt 2026-08-02: 1944 pary, m.in.
     # Blowjob/Blowjobs, Teen/Teens, Cumshot/Cumshots). Próbujemy obu kierunków: nowe
     # "Blowjobs" trafia w istniejące "Blowjob", a nowe "Blowjob" w istniejące "Blowjobs".
     # Tylko proste `+s`, form nieregularnych (panty/panties) świadomie nie ruszamy, bo
-    # zbyt łatwo o fałszywy zlew (glass/glasses, bra/bras). Dokładny klucz ma pierwszeństwo.
+    # zbyt łatwo o fałszywy zlew (glass/glasses, bra/bras).
-    keys: dict[str, tuple[str, str | None]] = {}
+    alts = [key + "s"] if not key.endswith("s") else [key[:-1]]
-    for name in names:
+    for alt in alts:
-        key = _altkey(name)
+        if len(alt) < 3:
-        if not key:
             continue
-        alt = key + "s" if not key.endswith("s") else key[:-1]
+        hit = (
-        keys[name] = (key, alt if len(alt) >= 3 else None)
+            session.execute(select(Tag).where(expr == alt).order_by(Tag.scene_count.desc()))
-    wanted = {k for k, _ in keys.values()} | {a for _, a in keys.values() if a}
+            .scalars()
-    if not wanted:
+            .first()
-        return {}
+        )
-    expr = func.regexp_replace(func.lower(func.btrim(Tag.name)), "[^a-z0-9]", "", "g")
+        if hit is not None:
-    best: dict[str, Tag] = {}
+            return hit
-    for key, tag in session.execute(
+    return None
-        select(expr, Tag).where(expr.in_(wanted)).order_by(Tag.scene_count.desc())
-    ).all():
-        best.setdefault(key, tag)
-    out: dict[str, Tag] = {}
-    for name, (key, alt) in keys.items():
-        tag = best.get(key) or (best.get(alt) if alt else None)
-        if tag is not None:
-            out[name] = tag
-    return out
 
 
 def _canonical_dup2_slug(session: Session, slug: str) -> str:

tests/test_yourdailypornmovies_cards.py

@@ -1,38 +0,0 @@
-from app.connectors.yourdailypornmovies import _AE_ID_RE, _CARD_RE
-
-_OLD = (
-    '<div class="movie"><div class="imagen"> <img src="https://imgs1cdn.adultempire.com/products/05/5128705h.jpg" '
-    'alt="Ramming Into Her" width="100%" height="100%" /> <a href="https://yourdailypornmovies.ws/ramming-into-her/">'
-    '<span class="player"></span></a></div><h2>Ramming Into Her</h2> <span class="year">2026</span></div>'
-)
-_NEW = (
-    '<div class="movie"><div class="imagen"> <img src="https://imgs4cdn.adultempire.com/movie/5136742?w=500&#038;h=700&#038;fit=cover" '
-    'alt="Big Butts Drive Me Nuts 8" width="100%" height="100%" /> <a href="https://yourdailypornmovies.ws/big-butts-drive-me-nuts-8/">'
-    '<span class="player"></span></a></div><h2>Big Butts Drive Me Nuts 8</h2> <span class="year">2026</span></div>'
-)
-_IBB = (
-    '<div class="movie"><div class="imagen"> <img src="https://i.ibb.co/DHDd9pm9/Private-POV.jpg" alt="Private POV" '
-    'width="100%" height="100%" /> <a href="https://yourdailypornmovies.ws/private-pov/"><span class="player"></span></a>'
-    '</div><h2>Private POV</h2> <span class="year">2026</span></div>'
-)
-_NO_IMG = (
-    '<div class="movie"><div class="imagen"> <a href="https://yourdailypornmovies.ws/no-cover/"><span class="player">'
-    '</span></a></div><h2>No Cover</h2> <span class="year">2025</span></div>'
-)
-
-
-def test_all_card_variants_parse():
-    cards = list(_CARD_RE.finditer(_OLD + _NEW + _IBB + _NO_IMG))
-    assert [c.group("title") for c in cards] == [
-        "Ramming Into Her", "Big Butts Drive Me Nuts 8", "Private POV", "No Cover"
-    ]
-    assert cards[3].group("poster") is None
-    assert cards[3].group("url") == "https://yourdailypornmovies.ws/no-cover/"
-
-
-def test_adultempire_id_from_both_poster_formats():
-    old = _AE_ID_RE.search("https://imgs1cdn.adultempire.com/products/05/5128705h.jpg")
-    new = _AE_ID_RE.search("https://imgs4cdn.adultempire.com/movie/5136742?w=500&h=700&fit=cover")
-    assert (old.group(1) or old.group(2)) == "5128705"
-    assert (new.group(1) or new.group(2)) == "5136742"
-    assert _AE_ID_RE.search("https://i.ibb.co/DHDd9pm9/Private-POV.jpg") is None