v1.9.0

What's Changed

• docs(changelog): add v1.8.1 entry by @thomas-villani in #59
• fix(search): invalidate persisted index when corpus or options change by @thomas-villani in #60
• feat(cache): opt-in on-disk conversion cache for grep/search/chunk/view by @thomas-villani in #61
• fix: conversion-fidelity quick wins (dehyphenation, DOCX blank line, .md detection) by @thomas-villani in #62
• feat(docx): round-trip run-level named character styles by @thomas-villani in #63
• feat(confidence): conversion confidence report ("quality card") by @thomas-villani in #64
• fix(docx): forward-port bullet/template fix from v1.8.2 by @thomas-villani in #66
• chore(release): let bump-my-version manage uv.lock by @thomas-villani in #67
• feat(roundtrip): structural fidelity scoring (all2md roundtrip) by @thomas-villani in #68
• fix(docs): make the options reference regenerate reproducibly by @thomas-villani in #69
• docs(roadmap): mark the Fidelity & Trust batch shipped, unblock the optimizer by @thomas-villani in #73
• fix(pdf): dehyphenate natively-extracted text, not just OCR output by @thomas-villani in #74
• fix(pdf): stop inventing tables from prose, and stop deleting the prose by @thomas-villani in #77
• fix(pdf): make auto_trim_headers_footers actually remove running furniture by @thomas-villani in #78
• feat(optimize): auto-tune converter options against a reference-free objective by @thomas-villani in #76
• fix(optimize): stop rewarding a parse that breaks words in half by @thomas-villani in #80
• fix(pages): stop double-converting string page ranges to 0-based (#75) by @thomas-villani in #81
• fix(html): stop dropping tables inside and inline wrappers (#79) by @thomas-villani in #82
• fix(html): stop double-wrapping loose list items in a nested paragraph (#72) by @thomas-villani in #88
• fix(docx): map Word 'Title' style back to a title heading on parse (#70) by @thomas-villani in #89
• fix(docx): round-trip inline code and block quotes via named styles (#71) by @thomas-villani in #90
• fix: avoid leaking OSError when long content is mistaken for a file path by @santhreal in #83
• fix: preserve markdown footnotes with mistune 3.x tokens by @santhreal in #84
• Fix loose multi-paragraph list roundtrip collapse by @santhreal in #85
• fix(markdown): indent list-item continuation lines and un-double-indent nested first child by @thomas-villani in #91
• Fix definition-list roundtrip collapse by @santhreal in #86
• fix(html): preserve
  instead of resetting list numbering to 1 by @santhreal in #87
  1. perf(cli): short-circuit --version/--about, cache type hints, + cold-start benchmark by @thomas-villani in #92
  2. perf(conversion): cache detection/introspection hot paths and skip DOCX blobs by @thomas-villani in #93
  3. fix(markdown): indent tables nested in list items (#96) by @thomas-villani in #99
  4. test(bench): markdown roundtrip fidelity benchmark (Phase 0/1) by @thomas-villani in #98
  5. fix(markdown): roundtrip footnotes as Markdown under default flavor (#94) by @thomas-villani in #100
  6. fix(markdown): roundtrip inline marks as Markdown by default (#95) by @thomas-villani in #102
  7. fix(markdown): keep inline $$…$$ display math instead of dropping it (#97) by @thomas-villani in #103
  8. fix(markdown): keep list intact when admonition degrades to a labelled quote (#104) by @thomas-villani in #105
  9. fix(optimize): search only valid figures/details parsing values by @thomas-villani in #106
  10. fix(markdown): roundtrip underline and non-GFM strikethrough by default by @thomas-villani in #107
  11. fix(pdf): preserve prose from every rejected table, not just degenerate grids by @thomas-villani in #108
  12. fix(roundtrip): score code/math/HTML/alt payloads in the text dimension by @thomas-villani in #109
  13. fix(docx): clamp title-promotion inversion at heading level 6 by @thomas-villani in #110
  14. fix(confidence): band uninstrumented conversions "not_assessed", not "high" by @thomas-villani in #111
  15. docs: post-1.8.0 refresh (D1-D6) by @thomas-villani in #112
  16. release: prepare v1.9.0 by @thomas-villani in #114

New Contributors

• @santhreal made their first contribution in #83

Full Changelog: v1.8.1...v1.9.0