libreoffice-rs

name	libreoffice-rs
description	Convert, render, and extract text from office documents (DOCX, PPTX, XLSX, ODT, ODS, ODP, ODG, CSV, Markdown, PDF) using the pure-Rust libreoffice-rs CLI. Use whenever the user wants to turn one office format into another, render slides/documents to PDF or PNG/JPEG, extract Markdown from Word/PowerPoint/Excel files, recalc XLSX formulas, or run a fast headless soffice-compatible conversion without installing LibreOffice.
license	See repository LICENSE
metadata	{"homepage":"https://github.com/stanvanrooy/libreoffice-rs","version":"0.4.1"}
compatibility	Requires a Rust toolchain (cargo) and a checkout of the libreoffice-rs repository. No C/C++ libraries needed; LibreOffice itself is NOT required.

libreoffice-rs

libreoffice-rs is a pure-Rust office suite. The CLI binary is libreoffice-pure and is invoked from inside a checkout with cargo run -p libreoffice-pure -- <subcommand>. Add --release for production-speed conversions on large inputs.

Use this skill when the user asks for any of:

• "convert this docx/pptx/xlsx/odt/ods/odp/odg/csv/md/pdf to X"
• "extract text/Markdown from this Word/PowerPoint/Excel file"
• "render these slides to PNG/JPEG/PDF"
• "headless soffice replacement", --convert-to, batch document conversion
• "recalculate / sanity-check XLSX formulas"
• "generate a Word/Excel/PowerPoint file from Markdown / CSV / a template"

The binary is fast (typically 15–50× faster than soffice --headless) and needs no LibreOffice installation, so prefer it over shelling out to soffice whenever the user already has the repo checked out.

When to use vs not use

Use this skill when:

• The repo libreoffice-rs is checked out (or you can clone it).
• The user wants common office-format conversions, text/Markdown extraction, or page rasterization.

Do NOT use this skill when:

• The user needs full feature-parity LibreOffice behaviour (macros, complex layout edge cases, exotic charts). Fall back to real soffice.
• The user only wants to read/write plain text or JSON — use ordinary file tools.

Quickstart

Always run from the repo root. Two equivalent invocation styles work:

# soffice-compatible style
cargo run -p libreoffice-pure -- --headless --convert-to pdf report.docx
cargo run -p libreoffice-pure -- --headless --convert-to pdf slide.pptx --outdir out

# explicit subcommand style
cargo run -p libreoffice-pure -- convert --to pdf report.docx
cargo run -p libreoffice-pure -- convert --from ods --to xlsx sheet.ods sheet.xlsx

For large or batch jobs add --release:

cargo run --release -p libreoffice-pure -- --headless --convert-to pdf big.docx

Common recipes

DOCX → PDF / TXT / Markdown / PNG / JPEG

cargo run -p libreoffice-pure -- --headless --convert-to pdf report.docx
cargo run -p libreoffice-pure -- docx-to-md   report.docx report.md
cargo run -p libreoffice-pure -- docx-to-pngs report.docx pages_png --dpi 144
cargo run -p libreoffice-pure -- docx-to-jpegs report.docx pages_jpg --dpi 150 --quality 88

PPTX → PDF / Markdown / PNG / JPEG (with native chart rendering)

cargo run -p libreoffice-pure -- --headless --convert-to pdf slides.pptx
cargo run -p libreoffice-pure -- pptx-to-md   slides.pptx slides.md
cargo run -p libreoffice-pure -- pptx-to-pngs slides.pptx slides_png --dpi 96
cargo run -p libreoffice-pure -- pptx-to-jpegs slides.pptx slides_jpg --dpi 150 --quality 85

PPTX charts (barChart, lineChart, areaChart, pieChart, doughnutChart, scatterChart, radarChart, bubbleChart, stockChart, combo charts) are parsed and drawn natively into the PDF/raster output.

XLSX / ODS → CSV / Markdown / round-trip

cargo run -p libreoffice-pure -- --headless --convert-to csv  sheet.xlsx
cargo run -p libreoffice-pure -- xlsx-to-md sheet.xlsx sheet.md
cargo run -p libreoffice-pure -- convert --from ods --to xlsx sheet.ods sheet.xlsx

Sanity-check XLSX formula recalculation against the file's cached values:

cargo run -p libreoffice-pure -- xlsx-recalc-check sheet.xlsx   # JSON report

Generate documents from Markdown / CSV

cargo run -p libreoffice-pure -- writer markdown-to-odt notes.md notes.odt
cargo run -p libreoffice-pure -- writer convert         notes.md notes.docx
cargo run -p libreoffice-pure -- writer convert         notes.md notes.pdf
cargo run -p libreoffice-pure -- writer convert         notes.md notes.html
cargo run -p libreoffice-pure -- calc   csv-to-ods     sheet.csv sheet.ods --sheet Data --has-header
cargo run -p libreoffice-pure -- calc   convert        sheet.csv sheet.xlsx --has-header
cargo run -p libreoffice-pure -- calc   eval "=SUM(B2:B4)" --csv numbers.csv --has-header

Other domains

cargo run -p libreoffice-pure -- impress  demo                  deck.odp
cargo run -p libreoffice-pure -- draw     demo                  diagram.odg
cargo run -p libreoffice-pure -- math     latex-to-odf          formula.txt formula.odf
cargo run -p libreoffice-pure -- base     csv-to-odb            people.csv People out.odb
cargo run -p libreoffice-pure -- package  inspect               out.odt

Filter strings

--convert-to "pdf:writer_pdf_Export" and similar soffice filter strings are accepted and normalized; the part after : is ignored when it maps onto a known target.

Library API (when embedding instead of shelling out)

If the user wants to call the library directly from Rust, point them at these entry points:

• libreoffice_pure::convert_bytes(input: &[u8], from: Format, to: Format)
• libreoffice_pure::convert_path_bytes(path, to: Format)
• libreoffice_pure::sniff_format_from_bytes(input: &[u8])
• libreoffice_pure::convert_bytes_auto(input: &[u8], to: Format) — byte-sniffs the source format
• Per-domain helpers: writer_convert_bytes, calc_convert_bytes, impress_convert_bytes, draw_convert_bytes, math_convert_bytes, base_convert_bytes
• lo_calc::xlsx_recalc_check_json / xlsx_recalc_report
• Raster helpers in lo_core::raster and the high-level docx_to_png_pages, docx_to_jpeg_pages, pptx_to_png_pages, pptx_to_jpeg_pages
• lo_writer::accept_all_tracked_changes_docx_bytes for tracked-changes acceptance (insertions kept, deletions/move-from dropped, comment anchors preserved, orphan <w:comment> entries pruned)
• lo_writer::legacy_doc for the binary .doc reader

Workspace layout (for code edits, not just CLI use)

• lo_core — shared models, XML/PDF/SVG/HTML helpers, raster pipeline
• lo_zip — pure-Rust ZIP + ODF/OOXML packaging
• lo_odf — ODF package serializers
• lo_writer — Writer (DOCX/ODT/HTML/PDF/SVG/TXT), legacy .doc reader, tracked-changes acceptance
• lo_calc — formulas, CSV/XLSX/ODS, recalc check
• lo_impress — slides, PPTX/ODP, native chart rendering
• lo_draw — vector drawings, ODG/SVG/PDF
• lo_math — TeX-style formulas, MathML/ODF/SVG/PDF
• lo_base — typed table model, SELECT/WHERE/ORDER BY/LIMIT, ODB
• lo_uno — UNO-like service runtime
• lo_lok — LibreOfficeKit-like in-process runtime
• lo_app — desktop application surface (windows, autosave, macros)
• libreoffice-rs — CLI binary (libreoffice-pure)

Verifying your output

After a conversion, prefer one of:

• cargo test --workspace for code-level changes
• bash tests/libreoffice_integration.sh (35 assertions; needs soffice)
• bash tests/office_demo_integration.sh (40 assertions, no soffice)
• bash tests/desktop_demo_integration.sh (38 assertions)
• bash tests/benchmark.sh (88 assertions)
• bash tests/quality_benchmark.sh for the full 13-file 1:1 quality comparison vs real LibreOffice

For one-off output sanity-checking, run the produced file through pdftotext or open it in a viewer and compare against the source.

Edge cases & gotchas

• Multi-byte UTF-8 is fully supported in DOCX/XLSX/PPTX/ODT/ODS/ODP/ ODG/ODB import (post-0.3.8 fix). Treat any mojibake as a real bug, not expected behaviour.
• DOCX tracked changes: use accept_all_tracked_changes_docx_bytes if the user wants a clean copy. It keeps w:ins/w:moveTo/w:cellIns, drops w:del/w:moveFrom/ deleted rows, strips formatting-history records, removes w:trackRevisions, and prunes orphan <w:comment> entries.
• Numbered lists in DOCX import respect numFmt (decimal, lowerRoman, lowerLetter) and emit 1., 2., … instead of bullets.
• PPTX charts with date-formatted serial axes intentionally skip tick labels (so Excel epoch numbers don't leak into extracted text).
• XLSX Markdown drops the synthesized workbook title and ## SheetN headers so it matches what soffice --convert-to csv produces.
• Filter strings like pdf:writer_pdf_Export are normalized — pass them through unchanged when mirroring an existing soffice command line.
• Performance: always add --release for files larger than a few hundred KB; debug builds are dramatically slower.
• Not feature-parity: fall back to real soffice for macros, complex page layouts, or any feature not listed in STATUS.md.

Reporting results

When you finish a conversion, summarise:

1. The exact command you ran.
2. Where the output landed (path).
3. Anything you noticed in the output (page count, warnings, recalc status=ok, Jaccard score if you ran a comparison).

If a conversion fails or produces obviously wrong output, capture the stderr and check STATUS.md and CHANGELOG.md to see whether the failing feature is known to be unsupported before filing a bug.