github run-llama/liteparse node-v2.19.0
Node.js v2.19.0

55 minutes ago

What's Changed

  • feat(ocr): add MiniCPM-OCR server by @lottshin in #536
  • feat(raw_text): per-glyph loose and strict boxes by @SebasGarcia08 in #542
  • chore(deps): update rust crate tokio to v1.53.3 [security] by @llama-renovate[bot] in #550
  • ci: run binding CI on root Cargo.toml/Cargo.lock changes by @meyerkev in #552
  • Text extraction: re-OCR fonts with misread glyphs, Arabic ligature/mark order, skip OCR on sparse text-layer pages by @logan-markewich in #555
  • Layout/markdown: cut above column-spanning lines, keep prose columns apart, rotated text, figure-split paragraphs, super/subscript, stacked tables by @logan-markewich in #556
  • OCR rasters: mask native text before OCR; PNG pass-through for image inputs by @logan-markewich in #560
  • OCR stages: restore deskewed boxes to page coordinates by @adrianlyjak in #564
  • OCR rasters: cap the long edge per page by @adrianlyjak in #563
  • Markdown: keep a line that holds a stray sideways glyph by @SebasGarcia08 in #561
  • OCR word boxes and rotation: per-word boxes for line-level OCR, caller-owned OCR request fields, server line angles by @SebasGarcia08 in #559
  • fix: decide clipped-text visibility per word, add keep_clipped_text by @logan-markewich in #571
  • OCR: despeckle dotted 1-bit scans by @Patco96 in #567
  • fix: make rotated text group order deterministic by @adrianlyjak in #573
  • docs: clarify source-build prerequisites and setup by @vinsymoyya in #572
  • Preserve isolated centered page numbers by @adrianlyjak in #569
  • Reuse supplied page images for OCR in Rust by @xujustinj in #565
  • fix: preserve native words across sub-point glyph gaps by @adrianlyjak in #570
  • Stage PDFium before building Python CI wheels by @adrianlyjak in #577

New Contributors

Full Changelog: crates-v2.18.0...node-v2.19.0

Don't miss a new liteparse release

NewReleases is sending notifications on new releases.