Add lazy comment fetching to watch page

- VideoComment model (video_id, author, text, likes, is_pinned, published_at)
- fetch_video_comments() in ytdlp.py: top 20 comments, no reply threads,
  sorted pinned-first then by likes
- GET /videos/by-yt/{id}/comments — returns cached comments instantly
- POST /videos/by-yt/{id}/comments/refresh — fetches from YouTube, stores, returns
- Watch page: CommentsSection shows "Load comments" button when uncached,
  renders comments with author/likes once loaded; Refresh link to re-fetch

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

backend/models.py

@@ -165,6 +165,20 @@ class UserTagAffinity(Base):
     updated_at = Column(DateTime, default=datetime.utcnow)
 
 
+class VideoComment(Base):
+    __tablename__ = "video_comments"
+
+    id = Column(Integer, primary_key=True, index=True)
+    video_id = Column(Integer, ForeignKey("videos.id", ondelete="CASCADE"), nullable=False, index=True)
+    youtube_comment_id = Column(String)
+    author = Column(String)
+    text = Column(Text)
+    likes = Column(Integer, default=0)
+    is_pinned = Column(Boolean, default=False)
+    published_at = Column(DateTime)
+    fetched_at = Column(DateTime, default=datetime.utcnow)
+
+
 class Collection(Base):
     __tablename__ = "collections"
 

backend/routers/videos.py

@@ -638,6 +638,72 @@ def delete_bookmark(
         db.commit()
 
 
+@router.get("/by-yt/{youtube_video_id}/comments")
+def get_comments(
+    youtube_video_id: str,
+    db: Session = Depends(get_db),
+    current_user: User = Depends(get_current_user),
+):
+    from ..models import VideoComment
+    video = db.query(Video).filter_by(youtube_video_id=youtube_video_id).first()
+    if not video:
+        return []
+    comments = (
+        db.query(VideoComment)
+        .filter_by(video_id=video.id)
+        .order_by(VideoComment.is_pinned.desc(), VideoComment.likes.desc())
+        .all()
+    )
+    return [
+        {
+            "author": c.author,
+            "text": c.text,
+            "likes": c.likes,
+            "is_pinned": c.is_pinned,
+            "published_at": c.published_at,
+        }
+        for c in comments
+    ]
+
+
+@router.post("/by-yt/{youtube_video_id}/comments/refresh")
+def refresh_comments(
+    youtube_video_id: str,
+    db: Session = Depends(get_db),
+    current_user: User = Depends(get_current_user),
+):
+    from ..models import VideoComment
+    video = db.query(Video).filter_by(youtube_video_id=youtube_video_id).first()
+    if not video:
+        raise HTTPException(status_code=404, detail="Video not found")
+
+    # Clear existing and re-fetch
+    db.query(VideoComment).filter_by(video_id=video.id).delete()
+    db.commit()
+
+    fetched = ytdlp.fetch_video_comments(youtube_video_id)
+    for c in fetched:
+        db.add(VideoComment(video_id=video.id, **c))
+    db.commit()
+
+    comments = (
+        db.query(VideoComment)
+        .filter_by(video_id=video.id)
+        .order_by(VideoComment.is_pinned.desc(), VideoComment.likes.desc())
+        .all()
+    )
+    return [
+        {
+            "author": c.author,
+            "text": c.text,
+            "likes": c.likes,
+            "is_pinned": c.is_pinned,
+            "published_at": c.published_at,
+        }
+        for c in comments
+    ]
+
+
 @router.get("/queue", response_model=list[VideoDetail])
 def queued_videos(
     db: Session = Depends(get_db),

backend/services/ytdlp.py

@@ -382,6 +382,46 @@ def fetch_channel_links(channel_id: str) -> list[str]:
     return list(channel_ids)
 
 
+def fetch_video_comments(youtube_video_id: str, max_comments: int = 20) -> list[dict]:
+    """Fetch top comments for a single video. Returns empty list on failure."""
+    url = f"https://www.youtube.com/watch?v={youtube_video_id}"
+    args = [
+        "yt-dlp", url,
+        "--dump-json",
+        "--write-comments",
+        "--extractor-args", f"youtube:max_comments={max_comments},max_comment_depth=1",
+        "--no-download",
+        "--no-playlist",
+        "--quiet",
+        *_cookie_args(),
+    ]
+    stdout, _, code = _run(args, timeout=60)
+    if not stdout.strip():
+        return []
+    try:
+        info = json.loads(stdout.strip())
+    except json.JSONDecodeError:
+        return []
+
+    result = []
+    for c in (info.get("comments") or []):
+        if c.get("parent") not in (None, "root"):
+            continue  # skip replies
+        ts = c.get("timestamp")
+        published_at = datetime.utcfromtimestamp(ts) if ts else None
+        result.append({
+            "youtube_comment_id": c.get("id"),
+            "author": c.get("author"),
+            "text": c.get("text"),
+            "likes": c.get("like_count") or 0,
+            "is_pinned": bool(c.get("is_pinned")),
+            "published_at": published_at,
+        })
+    # Sort pinned first, then by likes
+    result.sort(key=lambda c: (not c["is_pinned"], -(c["likes"] or 0)))
+    return result[:max_comments]
+
+
 QUALITY_FORMATS = {
     "best":  "bestvideo[ext=mp4][vcodec^=avc1]+bestaudio[ext=m4a]/bestvideo[ext=mp4]+bestaudio[ext=m4a]/22/18/bestvideo+bestaudio/best",
     "2160p": "bestvideo[ext=mp4][height<=2160]+bestaudio[ext=m4a]/bestvideo[height<=2160]+bestaudio/best[height<=2160]",