fix: correct center-point calculation in OCR results

ai/compute.py

@@ -45,10 +45,11 @@ def perform_ocr(screenshot: bytes) -> list[dict]:
     for i in range(n):
         text = data['text'][i]
         if text and text.strip():
+            # Fix the center-point calculation (add first, then divide)
             results.append({
                 'text': text,
-                'x': data['width'][i] + data['left'][i] // 2,  # center x position
-                'y': data['height'][i] + data['top'][i] // 2  # center y position
+                'x': data['left'][i] + data['width'][i] // 2,
+                'y': data['top'][i] + data['height'][i] // 2
             })
     
     # check if debug is enabled