Verify all recommendations against YouTube Music - no more hallucinated songs, direct YT Music links

backend/app/services/recommender.py

@@ -186,7 +186,7 @@ Already in their library (do NOT recommend these):
 {disliked_instruction}
 {exclude_instruction}
 
-Respond with exactly {count} music recommendations as a JSON array. Each item should have:
+Respond with exactly {count * 3} music recommendations as a JSON array. Give more than requested so we can verify they exist. Each item should have:
 - "title": song title
 - "artist": artist name
 - "album": album name (if known)
@@ -216,7 +216,36 @@ Return ONLY the JSON array, no other text."""
     except json.JSONDecodeError:
         return [], remaining
 
-    # Save to DB with YouTube links
+    # Verify each recommendation exists on YouTube Music before saving
+    import asyncio
+    from app.services.youtube_music import search_track
+    from difflib import SequenceMatcher
+
+    def _normalize(s: str) -> str:
+        import re
+        return re.sub(r'[^a-z0-9\s]', '', s.lower()).strip()
+
+    def _sim(a: str, b: str) -> float:
+        return SequenceMatcher(None, _normalize(a), _normalize(b)).ratio()
+
+    def verify_track(artist: str, title: str) -> dict | None:
+        """Search YouTube Music to verify a track exists and get the real data."""
+        try:
+            results = search_track(f"{artist} {title}")
+            for r in results[:3]:
+                r_artist = r.get("artist", "")
+                r_title = r.get("title", "")
+                # Check artist similarity (>0.5) and title similarity (>0.4)
+                if _sim(r_artist, artist) >= 0.5 and _sim(r_title, title) >= 0.4:
+                    return r
+                # Or very strong artist match with any title
+                if _sim(r_artist, artist) >= 0.8:
+                    return r
+        except Exception:
+            pass
+        return None
+
+    # Save to DB — only keep verified tracks
     recommendations = []
     for rec in recs_data:
         if len(recommendations) >= count:
@@ -224,17 +253,34 @@ Return ONLY the JSON array, no other text."""
 
         artist = rec.get("artist", "Unknown")
         title = rec.get("title", "Unknown")
+        reason = rec.get("reason", "")
 
-        # YouTube search link for every recommendation
+        # Verify on YouTube Music (run sync in thread)
-        youtube_url = f"https://www.youtube.com/results?search_query={quote_plus(f'{artist} {title} official music video')}"
+        verified = await asyncio.to_thread(verify_track, artist, title)
+
+        if not verified:
+            continue  # Skip hallucinated songs
+
+        # Use verified data (correct artist/title from YouTube Music)
+        real_artist = verified.get("artist", artist)
+        real_title = verified.get("title", title)
+        youtube_id = verified.get("youtube_id")
+        image_url = verified.get("image_url")
+
+        # Direct YouTube link if we have a video ID, otherwise search
+        if youtube_id:
+            youtube_url = f"https://music.youtube.com/watch?v={youtube_id}"
+        else:
+            youtube_url = f"https://www.youtube.com/results?search_query={quote_plus(f'{real_artist} {real_title} official music video')}"
 
         r = Recommendation(
             user_id=user.id,
             playlist_id=playlist_id,
-            title=title,
+            title=real_title,
-            artist=artist,
+            artist=real_artist,
             album=rec.get("album"),
-            reason=rec.get("reason", ""),
+            image_url=image_url,
+            reason=reason,
             score=rec.get("score"),
             query=query,
             youtube_url=youtube_url,

frontend/src/pages/BandcampDiscover.tsx

@@ -1,6 +1,6 @@
 import { useState, useEffect } from 'react'
 import { Disc3, Music, ExternalLink, Loader2 } from 'lucide-react'
-import { discoverBandcamp, getBandcampTags, BandcampRelease } from '../lib/api'
+import { discoverBandcamp, getBandcampTags, type BandcampRelease } from '../lib/api'
 
 const SORT_OPTIONS = [
   { value: 'new', label: 'New Releases' },