fix(MAT-166): robust option matching + language-aware UI for article summary

Replace brittle exact-string matching with keyword/substring classifier
that handles edge cases (punctuation, partial matches, German variants).
Detect article language and present all prompts in the users language.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

article_summary/__init__.py

@@ -23,6 +23,14 @@ URL_PATTERN = re.compile(r'https?://[^\s\)>\]"]+')
 
 CANCEL_WORDS = {"cancel", "stop", "abbrechen", "abbruch", "nevermind"}
 
+# Keyword sets for robust option matching (substring search, not exact match)
+_DISCUSS_KW = {"discuss", "diskutieren", "besprechen", "reden", "talk", "chat"}
+_TEXT_KW = {"text", "zusammenfassung", "summary", "lesen", "read", "schriftlich", "written"}
+_AUDIO_KW = {"audio", "mp3", "anhören", "vorlesen", "hören", "listen", "blinkist", "abspielen", "podcast"}
+
+# Simple German detection: common words that appear frequently in German text
+_DE_INDICATORS = {"der", "die", "das", "und", "ist", "ein", "eine", "für", "mit", "auf", "den", "dem", "sich", "nicht", "von", "wird", "auch", "nach", "wie", "aber"}
+
 LANGUAGE_OPTIONS = {
     "1": ("en", "English"),
     "2": ("de", "German"),
@@ -43,6 +51,44 @@ DURATION_OPTIONS = {
 }
 
 
+def _detect_content_lang(text: str) -> str:
+    """Detect language from text content. Returns 'de' or 'en'."""
+    words = set(re.findall(r'\b\w+\b', text.lower()))
+    de_hits = len(words & _DE_INDICATORS)
+    return "de" if de_hits >= 4 else "en"
+
+
+def _classify_choice(body: str) -> str | None:
+    """Classify user's action choice from free-form text.
+
+    Returns 'discuss', 'text', 'audio', or None (unrecognized).
+    """
+    # Normalize: lowercase, strip punctuation around digits
+    raw = body.strip().lower()
+    # Extract bare number if message is just "3." or "3!" or "nummer 3" etc.
+    num_match = re.search(r'\b([123])\b', raw)
+    bare_num = num_match.group(1) if num_match else None
+
+    # Number-only messages (highest priority — unambiguous)
+    stripped = re.sub(r'[^\w\s]', '', raw).strip()
+    if stripped in ("1", "2", "3"):
+        return {"1": "discuss", "2": "text", "3": "audio"}[stripped]
+
+    # Keyword search (substring matching)
+    if any(kw in raw for kw in _AUDIO_KW):
+        return "audio"
+    if any(kw in raw for kw in _TEXT_KW):
+        return "text"
+    if any(kw in raw for kw in _DISCUSS_KW):
+        return "discuss"
+
+    # "nummer 3" / "option 3" / "3. bitte" — number in context
+    if bare_num:
+        return {"1": "discuss", "2": "text", "3": "audio"}[bare_num]
+
+    return None
+
+
 class ArticleSummaryHandler:
     """Handles the interactive article summary conversation flow."""
 
@@ -76,8 +122,9 @@ class ArticleSummaryHandler:
 
         # Cancel from any active state
         if session.state != ArticleState.IDLE and body_lower in CANCEL_WORDS:
+            ui_de = session.ui_language == "de"
             self.sessions.reset(sender, room_id)
-            return "Summary cancelled."
+            return "Zusammenfassung abgebrochen." if ui_de else "Summary cancelled."
 
         # Route based on current state
         if session.state == ArticleState.IDLE:
@@ -100,6 +147,8 @@ class ArticleSummaryHandler:
             return self._on_topics(room_id, sender, body)
 
         elif session.state == ArticleState.GENERATING:
+            if session.ui_language == "de":
+                return "Zusammenfassung wird noch erstellt, bitte warten..."
             return "Still generating your summary, please wait..."
 
         elif session.state == ArticleState.COMPLETE:
@@ -145,6 +194,19 @@ class ArticleSummaryHandler:
         if session.detected_topics:
             topics_hint = f"\nTopics: {', '.join(session.detected_topics)}"
 
+        # Detect article language for localized UI
+        lang = _detect_content_lang(session.content[:2000])
+        session.ui_language = lang
+
+        if lang == "de":
+            return (
+                f"**Gefunden:** {session.title} (~{read_time} min Lesezeit){topics_hint}\n\n"
+                f"Was möchtest du damit machen?\n"
+                f"1\ufe0f\u20e3 **Diskutieren** \u2014 Ich lese den Artikel und wir reden darüber\n"
+                f"2\ufe0f\u20e3 **Textzusammenfassung** \u2014 Kurze schriftliche Zusammenfassung\n"
+                f"3\ufe0f\u20e3 **Audiozusammenfassung** \u2014 Blinkist-Style MP3\n\n"
+                f"_(oder schreib einfach weiter \u2014 ich unterbreche nicht)_"
+            )
         return (
             f"**Found:** {session.title} (~{read_time} min read){topics_hint}\n\n"
             f"What would you like to do?\n"
@@ -159,14 +221,24 @@ class ArticleSummaryHandler:
     ) -> str | None:
         """Handle language selection."""
         lang = LANGUAGE_OPTIONS.get(choice)
+        session = self.sessions.get(sender, room_id)
+        ui_de = session.ui_language == "de"
         if not lang:
+            if ui_de:
+                return "Bitte wähle eine Sprache: **1** für Englisch, **2** für Deutsch."
             return "Please pick a language: **1** for English, **2** for German."
 
-        session = self.sessions.get(sender, room_id)
         session.language = lang[0]
         session.state = ArticleState.LANGUAGE
         self.sessions.touch(sender, room_id)
 
+        if ui_de:
+            return (
+                f"Sprache: **{lang[1]}**. Wie lang soll die Zusammenfassung sein?\n"
+                f"1️⃣ 5 Min (kurz)\n"
+                f"2️⃣ 10 Min (standard)\n"
+                f"3️⃣ 15 Min (ausführlich)"
+            )
         return (
             f"Language: **{lang[1]}**. How long should the summary be?\n"
             f"1️⃣ 5 min (short)\n"
@@ -179,10 +251,13 @@ class ArticleSummaryHandler:
     ) -> str | None:
         """Handle duration selection."""
         duration = DURATION_OPTIONS.get(choice)
+        session = self.sessions.get(sender, room_id)
+        ui_de = session.ui_language == "de"
         if not duration:
+            if ui_de:
+                return "Bitte wähle: **1** (5 Min), **2** (10 Min) oder **3** (15 Min)."
             return "Please pick: **1** (5 min), **2** (10 min), or **3** (15 min)."
 
-        session = self.sessions.get(sender, room_id)
         session.duration_minutes = duration
         session.state = ArticleState.DURATION
         self.sessions.touch(sender, room_id)
@@ -191,12 +266,23 @@ class ArticleSummaryHandler:
             topic_list = "\n".join(
                 f"  • {t}" for t in session.detected_topics
             )
+            if ui_de:
+                return (
+                    f"Dauer: **{duration} Min**. Auf welche Themen fokussieren?\n"
+                    f"{topic_list}\n\n"
+                    f"Antworte mit Themennummern (kommagetrennt), bestimmten Themen oder **alle**."
+                )
             return (
                 f"Duration: **{duration} min**. Focus on which topics?\n"
                 f"{topic_list}\n\n"
                 f"Reply with topic numbers (comma-separated), specific topics, or **all**."
             )
         else:
+            if ui_de:
+                return (
+                    f"Dauer: **{duration} Min**. Bestimmte Themen im Fokus?\n"
+                    f"Antworte mit Themen (kommagetrennt) oder **alle** für eine allgemeine Zusammenfassung."
+                )
             return (
                 f"Duration: **{duration} min**. Any specific topics to focus on?\n"
                 f"Reply with topics (comma-separated) or **all** for a general summary."
@@ -234,23 +320,21 @@ class ArticleSummaryHandler:
     ) -> str | None:
         """Handle user's choice after URL detection: discuss, text summary, or audio."""
         session = self.sessions.get(sender, room_id)
+        choice = _classify_choice(body)
 
-        # Option 1: Discuss — reset FSM, return article context for AI handler
+        if choice == "discuss":
-        if body_lower in ("1", "discuss", "diskutieren", "besprechen"):
             article_context = session.content[:8000]
             title = session.title
             self.sessions.reset(sender, room_id)
             return f"__DISCUSS__{title}\n{article_context}"
 
-        # Option 2: Text summary — generate and return text, no TTS
+        if choice == "text":
-        if body_lower in ("2", "text", "text summary", "zusammenfassung", "textzusammenfassung"):
             return await self._generate_text_summary(room_id, sender)
 
-        # Option 3: Audio summary — enter language selection (existing flow)
+        if choice == "audio":
-        if body_lower in ("3", "audio", "audio summary", "audiozusammenfassung", "audio zusammenfassung"):
             return self._prompt_language(room_id, sender)
 
-        # Anything else — user is just chatting, reset and pass through with article context
+        # Unrecognized — user is just chatting, pass through with article context
         article_context = session.content[:8000]
         title = session.title
         self.sessions.reset(sender, room_id)
@@ -261,6 +345,12 @@ class ArticleSummaryHandler:
         session = self.sessions.get(sender, room_id)
         session.state = ArticleState.AWAITING_LANGUAGE
         self.sessions.touch(sender, room_id)
+        if session.ui_language == "de":
+            return (
+                "In welcher Sprache soll die Audiozusammenfassung sein?\n"
+                "1\ufe0f\u20e3 Englisch\n"
+                "2\ufe0f\u20e3 Deutsch"
+            )
         return (
             "What language for the audio summary?\n"
             "1\ufe0f\u20e3 English\n"
@@ -293,6 +383,11 @@ class ArticleSummaryHandler:
             session.summary_text = summary
             session.state = ArticleState.COMPLETE
             self.sessions.touch(sender, room_id)
+            if session.ui_language == "de":
+                return (
+                    f"**Zusammenfassung: {session.title}**\n\n{summary}\n\n"
+                    f"_Stelle Folgefragen oder teile einen neuen Link._"
+                )
             return (
                 f"**Summary: {session.title}**\n\n{summary}\n\n"
                 f"_Ask follow-up questions or share a new link._"
@@ -306,12 +401,20 @@ class ArticleSummaryHandler:
         """Run the full pipeline: summarize → TTS → upload MP3."""
         session = self.sessions.get(sender, room_id)
 
-        topics_str = ", ".join(session.topics) if session.topics else "all topics"
+        ui_de = session.ui_language == "de"
-        await bot._send_text(
+        topics_str = ", ".join(session.topics) if session.topics else ("alle Themen" if ui_de else "all topics")
-            room_id,
+        if ui_de:
-            f"Generating {session.duration_minutes}-min {session.language.upper()} "
+            await bot._send_text(
-            f"summary of **{session.title}** (focus: {topics_str})...",
+                room_id,
-        )
+                f"Erstelle {session.duration_minutes}-Min {session.language.upper()} "
+                f"Zusammenfassung von **{session.title}** (Fokus: {topics_str})...",
+            )
+        else:
+            await bot._send_text(
+                room_id,
+                f"Generating {session.duration_minutes}-min {session.language.upper()} "
+                f"summary of **{session.title}** (focus: {topics_str})...",
+            )
 
         try:
             # Step 1: Summarize
@@ -343,20 +446,32 @@ class ArticleSummaryHandler:
             transcript_preview = summary[:500]
             if len(summary) > 500:
                 transcript_preview += "..."
-            await bot._send_text(
+            if ui_de:
-                room_id,
+                await bot._send_text(
-                f"**Summary of:** {session.title}\n\n{transcript_preview}\n\n"
+                    room_id,
-                f"_You can ask follow-up questions about this article._",
+                    f"**Zusammenfassung von:** {session.title}\n\n{transcript_preview}\n\n"
-            )
+                    f"_Du kannst Folgefragen zu diesem Artikel stellen._",
+                )
+            else:
+                await bot._send_text(
+                    room_id,
+                    f"**Summary of:** {session.title}\n\n{transcript_preview}\n\n"
+                    f"_You can ask follow-up questions about this article._",
+                )
 
             session.state = ArticleState.COMPLETE
             self.sessions.touch(sender, room_id)
 
         except Exception:
             logger.exception("Article summary pipeline failed for %s", session.url)
-            await bot._send_text(
+            if ui_de:
-                room_id, "Sorry, I couldn't generate the audio summary. Please try again."
+                await bot._send_text(
-            )
+                    room_id, "Entschuldigung, die Audiozusammenfassung konnte nicht erstellt werden. Bitte versuche es erneut."
+                )
+            else:
+                await bot._send_text(
+                    room_id, "Sorry, I couldn't generate the audio summary. Please try again."
+                )
             self.sessions.reset(sender, room_id)
 
     async def _on_followup(

article_summary/state.py

@@ -24,7 +24,8 @@ class ArticleSession:
     url: str = ""
     title: str = ""
     content: str = ""
-    language: str = ""
+    language: str = ""       # Audio output language (set by user choice)
+    ui_language: str = "en"   # Detected from article content, used for UI strings
     duration_minutes: int = 10
     topics: list[str] = field(default_factory=list)
     detected_topics: list[str] = field(default_factory=list)