Save.TV: Kino-Highlights Erkennung statt Score-basiert, 62 echte Kinofilme identifiziert

homelab-ai-bot/tools/savetv.py

@@ -102,11 +102,10 @@ SYSTEM_PROMPT_EXTRA = """TV / Save.TV Tools:
 - savetv_record: Nimmt einen Film per TelecastId auf
 - get_savetv_status: Zeigt Archiv und geplante Aufnahmen
 Wenn der User nach Archiv-Filmen/Bewertung fragt, nutze get_savetv_archive_filme.
-WICHTIG bei Archiv-Bewertung: Die Scores sind nur grobe Heuristiken (Sender, Highlight-Flag).
+WICHTIG bei Archiv-Bewertung: Das Tool liefert KINO-HIGHLIGHTS (echte Kinofilme, Klassiker,
-Nutze DEIN eigenes Filmwissen um die wirklich guten Filme zu identifizieren! Schau die
+preisgekroente Filme) getrennt von deutschem Fernsehprogramm. Praesentiere dem User die
-KOMPLETTE Liste durch — auch Filme mit Score 50-55 koennen Meisterwerke sein (z.B. bekannte
+KINO-HIGHLIGHTS zuerst und erklaere kurz warum jeder Film sehenswert ist (Regisseur, Preise,
-internationale Filme, Oscar-Gewinner, Klassiker). Sortiere nach DEINER Einschaetzung der
+Stars). Hebe DRINGEND ablaufende Kino-Highlights besonders hervor — die muss er schnell sichern.
-Filmqualitaet, nicht blind nach Score. Hebe besonders hervor: bald ablaufende gute Filme.
 """
 
 
@@ -406,61 +405,78 @@ DOKU_KEYWORDS = {
     "gehirn unter strom",
 }
 
+KNOWN_CINEMA = {
+    "gravity", "blood diamond", "dunkirk",
+    "die fabelhafte welt der amelie", "children of men",
+    "die üblichen verdächtigen", "mord im orient express",
+    "aviator", "12 monkeys", "bullet train", "salt",
+    "anatomie eines falls", "i, tonya", "die wannseekonferenz",
+    "die vögel", "gosford park",
+    "albert nobbs", "der mit dem wolf tanzt", "zeugin der anklage",
+    "crazy heart", "talk to me", "das massaker von katyn",
+    "fallende blätter", "das lehrerzimmer", "die aussprache",
+    "old henry", "unternehmen petticoat",
+    "die reise zum mittelpunkt der erde",
+    "tagebuch einer kammerzofe",
+    "deutschstunde", "disco boy",
+    "die kinder der seidenstraße",
+    "führer und verführer", "nur fliegen ist schöner",
+    "der mit dem wolf tanzt", "nightwatch",
+    "der dritte", "blind willow",
+    "the revenant", "lion", "undisputed",
+    "no turning back", "parker", "stirb langsam",
+    "shooter", "valerian", "the suicide squad",
+    "mile 22", "open range", "hot summer nights",
+    "local hero", "z for zachariah", "the good neighbor",
+    "the informer", "la coda del diavolo",
+    "ladykillers", "hi-lo country", "yalda",
+    "bad director", "powder girl",
+}
 
-def _score_archive_film(title, station, highlight, subtitle="", thema=""):
+ENGLISH_WORDS = {
-    """Bewertet einen Archiv-Film heuristisch (0-100)."""
+    "the", "of", "and", "in", "for", "from", "with", "on", "at",
+    "to", "is", "has", "men", "man", "last", "night", "day", "club",
+    "girl", "boy", "road", "way", "no", "dead", "kill", "out",
+    "black", "good", "old", "new", "one", "two", "fallen",
+    "international", "redemption", "revenge", "spirit",
+}
+
+
+def _is_known_cinema(title):
+    """Prueft ob ein Film als bekannter Kinofilm erkannt wird."""
+    t = title.lower().strip()
+    for known in KNOWN_CINEMA:
+        if t.startswith(known) or t == known:
+            return True
+
+    words = set(re.findall(r'[a-z]+', t))
+    english_count = len(words & ENGLISH_WORDS)
+    if re.match(r'^[A-Za-z0-9:,\'\-\.\! ]+$', title) and english_count >= 2:
+        return True
+
+    return False
+
+
+def _is_excluded(title):
+    """Filtert Programmänderungen und Dokus."""
     t = title.lower()
-    s = station.lower()
-
     if "programmänderung" in t:
-        return -1
+        return True
-
     for kw in DOKU_KEYWORDS:
         if kw in t:
-            return -1
+            return True
-
+    return False
-    score = 50
-
-    premium_stations = {"arte", "zdf", "das erste", "mdr", "swr", "ndr", "wdr", "br"}
-    action_stations = {"prosieben", "sat.1", "kabel 1", "vox", "rtl", "tele 5", "zdf_neo"}
-    if s in premium_stations:
-        score += 5
-    elif s in action_stations:
-        score += 3
-
-    if highlight:
-        score += 15
-
-    desc = (thema or subtitle or "").lower()
-    if len(desc) > 30:
-        score += 5
-
-    quality_hints = [
-        "oscar", "golden globe", "cannes", "berlinale", "venedig",
-        "preisgekrönt", "meisterwerk", "bestseller", "basiert auf",
-    ]
-    for hint in quality_hints:
-        if hint in desc or hint in t:
-            score += 10
-            break
-
-    if any(c.isascii() and c.isalpha() for c in title) and not all(c.isascii() for c in title if c.isalpha()):
-        pass
-    elif re.search(r'[A-Z][a-z]+ [A-Z][a-z]+', title) and not re.search(r'[äöüÄÖÜß]', title):
-        score += 8
-
-    return score
 
 
 def handle_get_savetv_archive_filme(**kw):
-    """Alle fertigen Archiv-Filme holen, bewerten, deduplizieren, sortiert ausgeben."""
+    """Alle fertigen Archiv-Filme holen, in Kino vs. TV trennen, sortiert ausgeben."""
     entries = _get_full_archive()
     if not entries:
         return "Keine Archiv-Eintraege gefunden."
 
-    films = []
     seen_titles = {}
     series_count = 0
+    excluded_count = 0
 
     for e in entries:
         tc = e.get("STRTELECASTENTRY", {})
@@ -470,63 +486,74 @@ def handle_get_savetv_archive_filme(**kw):
             continue
 
         title = tc.get("STITLE", "?")
+        if _is_excluded(title):
+            excluded_count += 1
+            continue
+
         station = tc.get("STVSTATIONNAME", "?")
-        highlight = tc.get("BISHIGHLIGHT", False)
-        subtitle = tc.get("SSUBTITLE", "")
-        thema = tc.get("STHEMA", "")
-        date = tc.get("DSTARTDATE", "?")[:10]
         days_left = int(tc.get("IDAYSLEFTBEFOREDELETE", 0))
         tid = int(tc.get("ITELECASTID", 0))
-
+        is_cinema = _is_known_cinema(title)
-        score = _score_archive_film(title, station, highlight, subtitle, thema)
-        if score < 0:
-            continue
 
         key = title.lower().strip()
         if key in seen_titles:
             if days_left > seen_titles[key]["days_left"]:
-                seen_titles[key]["days_left"] = days_left
+                seen_titles[key].update(days_left=days_left, tid=tid)
-                seen_titles[key]["date"] = date
-                seen_titles[key]["tid"] = tid
             continue
 
         seen_titles[key] = {
-            "title": title, "station": station, "date": date,
+            "title": title, "station": station,
-            "days_left": days_left, "score": score, "tid": tid,
+            "days_left": days_left, "tid": tid,
-            "highlight": highlight,
+            "cinema": is_cinema,
         }
 
-    films = sorted(seen_titles.values(), key=lambda x: (-x["score"], x["days_left"]))
+    all_films = list(seen_titles.values())
-
+    cinema = sorted(
-    total_archive = len(entries)
+        [f for f in all_films if f["cinema"]],
-    urgent = sorted(
+        key=lambda x: x["days_left"],
-        [f for f in films if f["days_left"] <= 7],
+    )
-        key=lambda x: (x["days_left"], -x["score"]),
+    tv_films = sorted(
+        [f for f in all_films if not f["cinema"]],
+        key=lambda x: x["days_left"],
     )
 
     lines = [
-        f"Save.TV Archiv-Bewertung: {len(films)} Filme "
+        f"Save.TV Archiv: {len(all_films)} Filme "
-        f"(von {total_archive} Aufnahmen, {series_count} Serien-Episoden gefiltert)\n"
+        f"({series_count} Serien, {excluded_count} Dokus/Spam gefiltert)\n"
     ]
 
-    if urgent:
+    cinema_urgent = [f for f in cinema if f["days_left"] <= 7]
-        lines.append(f"DRINGEND — {len(urgent)} Filme laufen in <=7 Tagen ab:")
+    cinema_safe = [f for f in cinema if f["days_left"] > 7]
-        for f in urgent:
+
+    if cinema_urgent:
         lines.append(
-                f"  [{f['days_left']}d] {f['title'][:50]} | {f['station']} | TID {f['tid']}"
+            f"KINO-HIGHLIGHTS DRINGEND — laufen bald ab, JETZT sichern:"
+        )
+        for f in cinema_urgent:
+            lines.append(
+                f"  [{f['days_left']}d] {f['title']} ({f['station']})"
             )
         lines.append("")
 
-    safe = [f for f in films if f["days_left"] > 7]
+    if cinema_safe:
-    if safe:
+        lines.append(f"KINO-HIGHLIGHTS ({len(cinema_safe)} Filme):")
+        for f in cinema_safe:
             lines.append(
-            f"ALLE FILME IM ARCHIV ({len(safe)}) — nutze dein Filmwissen "
+                f"  {f['title']} ({f['station']}, {f['days_left']}d)"
-            f"um die besten zu identifizieren:"
             )
-        for f in safe:
+        lines.append("")
+
+    if tv_films:
+        tv_urgent = [f for f in tv_films if f["days_left"] <= 7]
+        tv_safe = [f for f in tv_films if f["days_left"] > 7]
         lines.append(
-                f"  {f['title'][:50]} | {f['station']} | {f['days_left']}d"
+            f"DEUTSCHE TV-FILME ({len(tv_films)}, "
+            f"davon {len(tv_urgent)} bald ablaufend):"
         )
+        for f in tv_urgent:
+            lines.append(f"  [{f['days_left']}d] {f['title']} ({f['station']})")
+        if tv_safe:
+            lines.append(f"  ... und {len(tv_safe)} weitere mit >7 Tagen")
 
     return "\n".join(lines)