feat: thumbnail extraction for Reader — fixes all clients

Server-side (dashboard + iOS + any client):
- Added thumbnail column to reader_entries
- Worker extracts from media:thumbnail, media:content, enclosures, HTML img
- API returns thumbnail in EntryOut with & decoding
- Backfilled 260 existing entries

iOS:
- Prefers API thumbnail, falls back to client-side extraction
- Decodes HTML entities in URLs

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

services/reader/app/api/entries.py

@@ -40,6 +40,7 @@ class EntryOut(BaseModel):
     status: str = "unread"
     starred: bool = False
     reading_time: int = 1
+    thumbnail: str | None = None
     feed: FeedRef | None = None
 
     class Config:
@@ -49,6 +50,10 @@ class EntryOut(BaseModel):
     def from_entry(cls, entry: Entry) -> "EntryOut":
         # Use full_content if available, otherwise RSS content
         best_content = entry.full_content if entry.full_content else entry.content
+        # Extract thumbnail from stored field, or from content
+        thumb = entry.thumbnail
+        if not thumb:
+            thumb = cls._extract_thumbnail(entry.content or entry.full_content or "")
         return cls(
             id=entry.id,
             title=entry.title,
@@ -60,9 +65,25 @@ class EntryOut(BaseModel):
             status=entry.status,
             starred=entry.starred,
             reading_time=entry.reading_time,
+            thumbnail=thumb,
             feed=FeedRef(id=entry.feed.id, title=entry.feed.title) if entry.feed else None,
         )
 
+    @staticmethod
+    def _extract_thumbnail(html: str) -> str | None:
+        """Extract first image URL from HTML content."""
+        if not html:
+            return None
+        import re
+        match = re.search(r'<img[^>]+src=["\']([^"\']+)["\']', html[:3000], re.IGNORECASE)
+        if match:
+            url = match.group(1).replace("&amp;", "&")
+            # Skip tiny tracking pixels and icons
+            if any(skip in url.lower() for skip in ["1x1", "pixel", "tracking", "spacer"]):
+                return None
+            return url
+        return None
+
 
 class EntryListOut(BaseModel):
     total: int

services/reader/app/models.py

@@ -68,6 +68,7 @@ class Entry(Base):
     published_at = Column(DateTime)
     status = Column(String(10), default="unread")
     starred = Column(Boolean, default=False)
+    thumbnail = Column(Text)
     reading_time = Column(Integer, default=1)
     created_at = Column(DateTime, default=datetime.utcnow)
 

services/reader/app/worker/tasks.py

@@ -79,6 +79,31 @@ def _get_entry_content(entry: dict) -> str:
     return ""
 
 
+def _extract_thumbnail(entry: dict, content: str) -> str | None:
+    """Extract thumbnail from feedparser entry or content HTML."""
+    # 1. Check media:thumbnail
+    for mt in entry.get("media_thumbnail", []):
+        if mt.get("url"):
+            return mt["url"]
+    # 2. Check media:content with image type
+    for mc in entry.get("media_content", []):
+        if mc.get("medium") == "image" or (mc.get("type", "").startswith("image")):
+            if mc.get("url"):
+                return mc["url"]
+    # 3. Check enclosures
+    for enc in entry.get("enclosures", []):
+        if enc.get("type", "").startswith("image") and enc.get("href"):
+            return enc["href"]
+    # 4. Extract from content HTML
+    if content:
+        match = re.search(r'<img[^>]+src=["\']([^"\']+)["\']', content[:3000], re.IGNORECASE)
+        if match:
+            url = match.group(1).replace("&amp;", "&")
+            if not any(skip in url.lower() for skip in ["1x1", "pixel", "tracking", "spacer"]):
+                return url
+    return None
+
+
 def _get_entry_author(entry: dict) -> str | None:
     """Extract author from a feedparser entry."""
     if entry.get("author"):
@@ -157,6 +182,7 @@ def fetch_single_feed(feed_id: int):
 
             content = _get_entry_content(fe)
             pub_date = _parse_date(fe)
+            thumb = _extract_thumbnail(fe, content)
 
             stmt = pg_insert(Entry).values(
                 feed_id=feed.id,
@@ -168,6 +194,7 @@ def fetch_single_feed(feed_id: int):
                 published_at=pub_date,
                 status="unread",
                 starred=False,
+                thumbnail=thumb,
                 reading_time=_calc_reading_time(content),
             ).on_conflict_do_nothing(
                 constraint="uq_reader_entries_feed_url"