5ojg,SrSSKJrJrJrJrJr /SQrg)uReading text off page images, with provenance. One module lives here today: `well`, the OCR grounding well. It exists for content the parse cannot ground — scanned pages and image-bound figures, where no native text layer sits under the region. The well reads the rendered page with RapidOCR and keeps each fragment with its box and confidence, in PDF points, so consumers can trace an extracted value back to the pixels that support it. The split of responsibilities: this package only reads and locates. Deciding which regions need a well (the zero-native-text discriminator), attributing values to fragments, and correcting readings all belong to consumers. ) DocumentWell OcrFragmentPageWell build_well read_imageN)__doc__quber.core.ocr.wellrrrrr__all__5/home/mande/repo/quber/src/quber/core/ocr/__init__.pyrs \[ r