feat: scheduled reminders + less aggressive article summary

Add scheduled messages/reminders system:
- New scheduled_messages table in memory-service with CRUD endpoints
- schedule_message, list_reminders, cancel_reminder tools for the bot
- Background scheduler loop (30s) sends due reminders automatically
- Supports one-time, daily, weekly, weekdays, monthly repeat patterns

Make article URL handling non-blocking:
- Show 3 options (discuss, text summary, audio) instead of forcing audio wizard
- Default to passing article context to AI if user just keeps chatting
- New AWAITING_LANGUAGE state for cleaner audio flow FSM

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

article_summary/__init__.py

@@ -84,7 +84,11 @@ class ArticleSummaryHandler:
             return await self._check_for_url(room_id, sender, body)
 
         elif session.state == ArticleState.URL_DETECTED:
-            # Waiting for language selection
+            # Waiting for user to pick action (discuss, text summary, audio)
+            return await self._on_action_choice(room_id, sender, body, body_lower)
+
+        elif session.state == ArticleState.AWAITING_LANGUAGE:
+            # Audio flow: waiting for language selection
             return self._on_language(room_id, sender, body_lower)
 
         elif session.state == ArticleState.LANGUAGE:
@@ -143,10 +147,11 @@ class ArticleSummaryHandler:
 
         return (
             f"**Found:** {session.title} (~{read_time} min read){topics_hint}\n\n"
-            f"Want an audio summary? What language?\n"
-            f"1️⃣ English\n"
-            f"2️⃣ German\n\n"
-            f"_(or say \"cancel\" to skip)_"
+            f"What would you like to do?\n"
+            f"1\ufe0f\u20e3 **Discuss** \u2014 I'll read the article and we can talk about it\n"
+            f"2\ufe0f\u20e3 **Text summary** \u2014 Quick written summary\n"
+            f"3\ufe0f\u20e3 **Audio summary** \u2014 Blinkist-style MP3\n\n"
+            f"_(or just keep chatting \u2014 I won't interrupt)_"
         )
 
     def _on_language(
@@ -224,6 +229,79 @@ class ArticleSummaryHandler:
         self.sessions.touch(sender, room_id)
         return "__GENERATE__"
 
+    async def _on_action_choice(
+        self, room_id: str, sender: str, body: str, body_lower: str
+    ) -> str | None:
+        """Handle user's choice after URL detection: discuss, text summary, or audio."""
+        session = self.sessions.get(sender, room_id)
+
+        # Option 1: Discuss — reset FSM, return article context for AI handler
+        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 body_lower in ("2", "text", "text summary", "zusammenfassung"):
+            return await self._generate_text_summary(room_id, sender)
+
+        # Option 3: Audio summary — enter language selection (existing flow)
+        if body_lower in ("3", "audio", "audio summary"):
+            return self._prompt_language(room_id, sender)
+
+        # Anything else — user is just chatting, reset and pass through with article context
+        article_context = session.content[:8000]
+        title = session.title
+        self.sessions.reset(sender, room_id)
+        return f"__DISCUSS__{title}\n{article_context}"
+
+    def _prompt_language(self, room_id: str, sender: str) -> str:
+        """Present language selection for audio summary."""
+        session = self.sessions.get(sender, room_id)
+        session.state = ArticleState.AWAITING_LANGUAGE
+        self.sessions.touch(sender, room_id)
+        return (
+            "What language for the audio summary?\n"
+            "1\ufe0f\u20e3 English\n"
+            "2\ufe0f\u20e3 German"
+        )
+
+    async def _generate_text_summary(self, room_id: str, sender: str) -> str | None:
+        """Generate a text-only summary of the article."""
+        session = self.sessions.get(sender, room_id)
+        try:
+            resp = await self.llm.chat.completions.create(
+                model=self.model,
+                messages=[
+                    {
+                        "role": "system",
+                        "content": (
+                            "Summarize this article concisely in 3-5 paragraphs. "
+                            "Respond in the same language as the article."
+                        ),
+                    },
+                    {
+                        "role": "user",
+                        "content": f"Article: {session.title}\n\n{session.content[:12000]}",
+                    },
+                ],
+                max_tokens=1000,
+                temperature=0.3,
+            )
+            summary = resp.choices[0].message.content.strip()
+            session.summary_text = summary
+            session.state = ArticleState.COMPLETE
+            self.sessions.touch(sender, room_id)
+            return (
+                f"**Summary: {session.title}**\n\n{summary}\n\n"
+                f"_Ask follow-up questions or share a new link._"
+            )
+        except Exception:
+            logger.warning("Text summary failed", exc_info=True)
+            self.sessions.reset(sender, room_id)
+            return None
+
     async def generate_and_post(self, bot, room_id: str, sender: str) -> None:
         """Run the full pipeline: summarize → TTS → upload MP3."""
         session = self.sessions.get(sender, room_id)

article_summary/state.py

@@ -10,6 +10,7 @@ from enum import Enum, auto
 class ArticleState(Enum):
     IDLE = auto()
     URL_DETECTED = auto()
+    AWAITING_LANGUAGE = auto()  # Audio flow: waiting for language selection
     LANGUAGE = auto()
     DURATION = auto()
     TOPICS = auto()