⚡ Bolt: [performance improvement] Add verbose=False to YOLO predictions to reduce stdout overhead

.jules/bolt.md

@@ -0,0 +1,3 @@
+## 2026-04-18 - Ultralytics YOLO verbose Output Overhead
+**Learning:** By default, Ultralytics YOLO models log predictions to stdout on every call. In a loop processing multiple pages, synchronous logging blocks the main thread, introducing significant unneeded performance overhead.
+**Action:** Always pass `verbose=False` to `model.predict()` when running inference in loops or batch processes unless active debugging is needed.

commonforms/inference.py

@@ -124,6 +124,7 @@ def extract_widgets(
                     conf=confidence,
                     augment=False,
                     imgsz=ONNX_IMAGE_SIZE,
+                    verbose=False,
                 )
                 for p in pages
             ]
@@ -135,6 +136,7 @@ def extract_widgets(
                 augment=True,
                 imgsz=image_size,
                 device=self.device,
+                verbose=False,
             )
 
         widgets = {}