feat(yolo-tracker): add detection_count and summary metadata

plugins/forgesyte-yolo-tracker/src/forgesyte_yolo_tracker/inference/radar.py

@@ -206,6 +206,7 @@ def generate_radar_json(
     return {
         "radar_points": radar_points,
         "radar_size": [radar_w, radar_h],
+        "detection_count": len(radar_points),
     }
 
 
@@ -287,6 +288,7 @@ def radar_json_with_annotated_frame(
     return {
         "radar_points": radar_points,
         "radar_size": [radar_w, radar_h],
+        "detection_count": len(radar_points),
         "radar_base64": _encode_radar_to_base64(radar_image),
     }
 

plugins/forgesyte-yolo-tracker/src/forgesyte_yolo_tracker/manifest.json

@@ -96,7 +96,8 @@
       },
       "outputs": {
         "frames": "array",
-        "total_frames": "integer"
+        "total_frames": "integer",
+        "summary": "object"
       }
     },
     {

plugins/forgesyte-yolo-tracker/src/forgesyte_yolo_tracker/plugin.py

@@ -59,6 +59,7 @@ def on_unload(self) -> None:  # pragma: no cover  # noqa: B027
 from forgesyte_yolo_tracker.inference.radar import radar_json_with_annotated_frame
 from forgesyte_yolo_tracker.configs import load_model_config
 from forgesyte_yolo_tracker.utils.json_sanitize import sanitize_json
+from forgesyte_yolo_tracker.utils.summary import compute_video_summary
 
 logger = logging.getLogger(__name__)
 
@@ -310,12 +311,14 @@ def _run_video_tool(
     logger.info(f"Completed: {frame_index} frames")
 
     # v0.10.0: Sanitize output for JSON serialization
+    sanitized = sanitize_json({
+        "total_frames": frame_index,
+        "frames": frame_results,
+    })
+    sanitized["summary"] = compute_video_summary(sanitized["frames"])
     return {
         "success": True,
-        "result": sanitize_json({
-            "total_frames": frame_index,
-            "frames": frame_results,
-        })
+        "result": sanitized,
     }
 
 
@@ -486,10 +489,12 @@ def _tool_video_player_tracking(
         frame_index += 1
 
     # v0.10.0: Sanitize output for JSON serialization
-    return sanitize_json({
+    sanitized = sanitize_json({
         "total_frames": frame_index,
         "frames": frame_results,
     })
+    sanitized["summary"] = compute_video_summary(sanitized["frames"])
+    return sanitized
 
 
 class Plugin(BasePlugin):  # type: ignore[misc]

plugins/forgesyte-yolo-tracker/src/forgesyte_yolo_tracker/utils/__init__.py

@@ -10,9 +10,11 @@
 Custom forgeSYTE modules:
 - ball.py - Ball tracking (not annotating)
 - soccer_pitch.py - Soccer pitch drawing utilities
+- summary.py - Video summary computation utilities
 """
 
 from . import ball, soccer_pitch
+from .summary import compute_video_summary
 
 # Lazy import to avoid torch/transformers at module load time
 def __getattr__(name: str):
@@ -33,6 +35,8 @@ def __getattr__(name: str):
     "create_batches",
     "TeamClassifier",
     "ViewTransformer",
+    # Video summary
+    "compute_video_summary",
     # Custom forgeSYTE modules
     "ball",
     "soccer_pitch",

plugins/forgesyte-yolo-tracker/src/forgesyte_yolo_tracker/utils/summary.py

@@ -0,0 +1,59 @@
+"""Video summary computation utilities.
+
+Provides functions to compute aggregated metadata from video tool results.
+"""
+
+import logging
+from typing import Any, Dict, List
+
+logger = logging.getLogger(__name__)
+
+
+# Registry pattern for detection extraction (avoid elif chains)
+_DETECTION_FIELD_MAP: Dict[str, str] = {
+    "detections": "detections",
+    "tracked_objects": "tracked_objects",
+    "radar_points": "radar_points",
+}
+
+
+def _get_detections_from_frame(frame: Dict[str, Any]) -> List[Dict[str, Any]]:
+    """Extract detections list from frame using registry lookup.
+
+    Args:
+        frame: Frame result dictionary
+
+    Returns:
+        List of detection dicts, or empty list if not found
+    """
+    for field_name in _DETECTION_FIELD_MAP.values():
+        if field_name in frame:
+            return frame[field_name]
+    return []
+
+
+def compute_video_summary(frames: List[Dict[str, Any]]) -> Dict[str, int]:
+    """Compute aggregated summary metadata from video tool frames.
+
+    Args:
+        frames: List of frame results from video processing
+
+    Returns:
+        Dictionary with:
+        - detection_count: Total detections across all frames
+        - frame_count: Number of frames processed
+    """
+    total_detection_count = 0
+
+    for frame in frames:
+        detections = _get_detections_from_frame(frame)
+        total_detection_count += len(detections)
+
+    logger.info(
+        f"Video summary: {total_detection_count} detections across {len(frames)} frames"
+    )
+
+    return {
+        "detection_count": total_detection_count,
+        "frame_count": len(frames),
+    }