Module ocr 

OCR backend selector.

macOS uses VNRecognizeTextRequest (accurate, language correction on). Windows and Linux return Unsupported until P12 (Windows) and later phases. The mock backend short-circuits to an in-memory provider.

Structs

ImageRect

Image-pixel rectangle (top-left origin), used by both the recognizer and the secure-field redaction logic.

OcrBlock

One recognized text line with image-pixel coordinates.

OcrResult

Result of one OCR pass on a captured window.

Functions

recognize

Run OCR on image and return per-line blocks.

redact_blocks

Drop every OCR block whose rect falls inside any secure-field frame.