Skip to content

v0.1.0

Choose a tag to compare

View all tags
@HelgDemidov HelgDemidov released this 19 Aug 17:50
· 146 commits to main since this release
v0.1.0
968c178

DOCX/XLSX → Markdown conversion that treats embedded charts and composite figures as first-class content instead of silently dropping or fragmenting them.

Highlights

  • Native OOXML chart-data extraction — reads numCache/strCache straight from the chart XML, no rasterize/OCR/VLM. Validated against 407 real charts across 27 real-world documents (400 render as mermaid diagrams).
  • Zero-loss composite-figure markers — grouped shapes/infographics get a positioned marker with caption preserved, instead of the silent fragmentation most DOCX converters produce (see Docling #1287, open >1 year, unfixed).
  • Optional VLM interpretation (implemented, not yet promoted as a headline v1 feature) — provider-agnostic cloud description + mermaid diagram on top of the zero-loss floor.
  • Rich, typed API — ConversionResult (markdown + warnings + counts), typed exceptions, strict mode, path/bytes/file-like input.
  • CLI (refigure command) — single-file and batch conversion.
  • Lightweight — refigure[docx,xlsx] installs at ~5.6MB, no PyTorch/CUDA stack.
  • 100% combined test coverage (unit + integration), gated in CI.

Scope: DOCX + XLSX only — PDF is deliberately out of scope (see README).

pip install "refigure[docx,xlsx]"

Full details: README.

Assets 2
Loading