v0.1.1

Fixes

• Fix pdftotext extraction for vertical Japanese text: Switched from vertical CID font (HeiseiMin-W3, WMode=1) to horizontal font (HeiseiKakuGo-W5) for the invisible text layer, so pdftotext can extract vertical text correctly
• Normalize vertical punctuation Convert vertical presentation forms (︑︒﹁﹂ etc.) to standard CJK punctuation (、。「」) in OCR output
• Fix vertical text overlay positioning: Rotate canvas -90° for vertical text so the horizontal font renders top-to-bottom within bounding boxes
• Suppress ndlocr-lite stdout noise: Redirect stray prints from ndlocr-lite so they don't break tqdm progress bar

v0.1.0

Initial release of ndlpdf, a CLI for converting scanned PDFs to searchable PDFs using ndlocr-lite OCR.