Feature: YouTube heatmap extraction for viral segment detection

server.py

@@ -405,7 +405,6 @@ def download_video(youtube_url: str, max_retries: int = 3) -> Path:
     raise RuntimeError(f"Download failed after {max_retries} attempts:\n{last_err}")
 
 
-# ── Step 2: Video info ────────────────────────────────────────────────────────
 def get_video_info(video_path: Path):
     """Return (duration, width, height) via ffprobe JSON."""
     r = subprocess.run(
@@ -424,6 +423,83 @@ def get_video_info(video_path: Path):
     return duration, width, height
 
 
+def get_youtube_heatmap(video_path: Path, url: str):
+    """
+    Fetch YouTube's 'Most Replayed' heatmap data using yt-dlp.
+    Returns: [{start_time: float, end_time: float, score: float}] or []
+    """
+    print("[>>] Fetching YouTube heatmap...", flush=True)
+    
+    # We use the same hardened bypass settings as download_video
+    info_json_path = video_path.with_suffix(".info.json")
+    
+    cmd = [
+        "yt-dlp",
+        "--force-ipv4",
+        "--ignore-config",
+        "--no-cache-dir",
+        "--user-agent", _USER_AGENT,
+        "--extractor-args", "youtube:player_client=web,tv,ios;player_skip=web_embedded_check",
+        "--remote-components", "ejs:github",
+        "--no-check-certificates",
+        "--geo-bypass",
+        "--add-header", "Accept-Language:en-US,en;q=0.9",
+        "--add-header", "Accept:text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
+        "--write-info-json",
+        "--skip-download",
+        "-o", str(video_path.with_suffix("")),  # This will result in filename.info.json
+        url,
+    ]
+    
+    # Add cookies if available
+    cookies_path = _find_cookies_file()
+    if cookies_path:
+        cmd[1:1] = ["--cookies", cookies_path]
+        
+    # Add Proxy if defined
+    if YTDLP_PROXY:
+        cmd[1:1] = ["--proxy", YTDLP_PROXY]
+
+    try:
+        # Run yt-dlp to get JSON
+        subprocess.run(cmd, capture_output=True, text=True, timeout=30)
+        
+        if not info_json_path.exists():
+            print("[!] Heatmap info JSON not found.", flush=True)
+            return []
+            
+        with open(info_json_path, "r", encoding="utf-8") as f:
+            data = json.load(f)
+            
+        heatmap = data.get("heatmap")
+        if not heatmap:
+            print("[!] No heatmap data found in YouTube metadata.", flush=True)
+            return []
+            
+        # Normalize scores to 0.0 - 1.0
+        max_val = max((item.get("value", 0) for item in heatmap), default=1.0)
+        if max_val == 0: max_val = 1.0
+        
+        normalized = []
+        for item in heatmap:
+            normalized.append({
+                "start_time": float(item["start_time"]),
+                "end_time": float(item["end_time"]),
+                "score": round(float(item["value"]) / max_val, 4)
+            })
+            
+        print(f"[OK] Extracted {len(normalized)} heatmap segments.", flush=True)
+        
+        # Cleanup
+        try: info_json_path.unlink()
+        except: pass
+        
+        return normalized
+    except Exception as e:
+        print(f"[!] Heatmap extraction error: {e}", flush=True)
+        return []
+
+
 # ── Step 3: Audio energy analysis ────────────────────────────────────────────
 def extract_audio_energy(video_path: Path, duration: float):
     """
@@ -921,6 +997,9 @@ def process():
         try:
             # 1. Download
             video_path = download_video(youtube_url)
+            
+            # 1.1 Heatmap (New Viral Signalling)
+            heatmap_data = get_youtube_heatmap(video_path, youtube_url)
 
             # 2. Info
             duration, width, height = get_video_info(video_path)