folio-docx 0.2.1

Markdown to polished .docx output, without the cleanup pass.

Folio

Markdown to polished .docx output, without the cleanup pass.

English · 简体中文

Folio is a cross-platform desktop application and Rust workspace for turning Markdown into Microsoft Word documents with native Word structures instead of brittle export artifacts. The focus is practical fidelity: editable equations, sane image sizing, readable tables, predictable heading hierarchy, and output that looks intentional when it opens in Word or LibreOffice.

Quick Start

For most users

The simplest way to use Folio on both macOS and Windows is to download a prebuilt installer from GitHub Releases.

At the moment, the published desktop releases are still experimental and may not work reliably on every machine. If you need the most dependable path today, use the CLI from source instead.

• macOS: download the .dmg that matches your CPU.
• Apple Silicon: choose the aarch64 / arm64 build.
• Intel Mac: choose the x64 build.
• Open the .dmg, then drag Folio into Applications.
• Windows: download the NSIS .exe installer for the simplest setup.

Because Folio is still pre-alpha and may be unsigned, the first launch may show a platform warning:

• macOS: right-click Folio in Applications and choose Open the first time if Gatekeeper blocks it.
• Windows: if SmartScreen appears, click More info -> Run anyway.

If you want to build from source

Use the development setup below only if you want to contribute or run the app from source.

Recommended CLI Usage

If the desktop release does not work for you yet, the CLI is the recommended fallback.

macOS:

cd /path/to/Folio
cargo run -p scribe-cli -- "/absolute/path/input.md" -o "/absolute/path/output.docx"

Windows (PowerShell):

cd C:\path\to\Folio
cargo run -p scribe-cli -- "C:\absolute\path\input.md" -o "C:\absolute\path\output.docx"

You can also omit -o, and Folio will write <input>.docx next to the source Markdown file.

Python (preview)

Folio is also published as a Python wheel for users who want to embed the conversion in scripts, notebooks, or pipelines.

pip install folio-docx

import folio

# Markdown string -> .docx bytes
docx_bytes = folio.convert("# Hello\n\n$E = mc^2$")

# Markdown file -> .docx file (relative image paths are resolved
# against the input file's parent directory)
folio.convert_file("paper.md", "paper.docx")

# Live HTML preview
html = folio.preview_html("# Title\n\nbody")

The wheel ships as cp38-abi3, so a single download works on every CPython 3.8+ on macOS (Apple Silicon + Intel), Linux x86_64 + aarch64, and Windows x64. The conversion runs entirely in Rust — convert() releases the GIL, so multiple threads can render in parallel.

Templates

The output's fonts, sizes, headings, code style, and paragraph spacing all come from a Word style sheet. Folio ships built-in themes for the common cases and accepts your own reference .docx for everything else.

Built-in themes

# Pick a theme by name. Use --list-themes to see what ships.
scribe-cli paper.md --theme academic   -o paper.docx
scribe-cli paper.md --theme thesis-cn  -o paper.docx
scribe-cli --list-themes

import folio
folio.convert(md, theme="academic")
folio.list_themes()  # ["academic", "thesis-cn"]

Theme	Use it for
academic	English academic paper. Times New Roman 12pt body, 1.5 line height, classic heading hierarchy.
thesis-cn	中文学位论文。宋体正文 + 黑体标题，1.5 倍行距，首行缩进 2 字符。
report	Modern business report. Calibri sans-serif, blue heading accents, tighter line spacing.

Bring your own template

Pandoc-style: hand a styled .docx to Folio and your output adopts its styles wholesale.

scribe-cli paper.md --reference-doc thesis-template.docx -o paper.docx

folio.convert(md, reference_doc="thesis-template.docx")

Anything you can do in Word — fonts, colors, paragraph spacing, table styles, list bullets — flows through verbatim. Open the template in Word, edit the styles, save, and Folio's next run picks them up. --reference-doc and --theme are mutually exclusive.

Why Folio

Most Markdown-to-DOCX workflows break down in the last 10%:

• math is flattened into images or mangled XML
• lists and tables need manual repair
• images overflow, shrink unpredictably, or lose aspect ratio
• report and academic formatting still needs hand cleanup in Word

Folio is built to close that gap with a pure Rust pipeline and a desktop UX around it.

Sample Export

The samples below are generated from the comprehensive regression fixture in test/folio-comprehensive.md, exported to test/output/folio-comprehensive.docx, and rendered from the corresponding PDF.

Page 1 Formatting, lists, code, and tables	Page 2 Math and raster logo embedding	Page 3 SVG assets and footnotes

Download DOCX · Open PDF · View Source Markdown

The current sample exercises:

• heading hierarchy
• inline emphasis, code, and links
• unordered, ordered, and task lists
• blockquotes and code blocks
• aligned tables
• inline and display LaTeX math rendered as editable OMML
• raster and SVG image embedding
• footnotes

WPS Spot Check vs Pandoc

The comparison below uses the same source note rendered as .docx, then opened in WPS Office for a practical end-user check.

Folio	Pandoc

In this note, Folio is materially stronger than Pandoc in the places that matter most for technical writing:

• formula-heavy sections stay closer to the intended reading order instead of collapsing into inline-like fragments
• image + heading + paragraph blocks hold together more predictably across pages
• long structured notes stay denser and cleaner instead of ballooning into many extra pages

What Works Today

Folio is still pre-alpha, but the core conversion path is already structured as a serious multi-crate system rather than a prototype script.

Implemented today:

• Markdown parsing with CommonMark and GFM features
• typed AST for downstream transforms
• LaTeX -> MathML -> OMML conversion
• image loading, normalization, and SVG rasterization
• DOCX emission with styles, numbering, footnotes, tables, and images
• HTML preview rendering for the desktop app
• Tauri desktop shell with a Svelte frontend

Not complete yet:

• reference-template ingestion from user reference.docx
• richer academic presets and style packs
• cross-references and auto-numbered figures, tables, and equations
• batch conversion UX polish
• higher-fidelity preview parity

Technology Stack

Core conversion engine

• Rust stable workspace
• pulldown-cmark for Markdown parsing
• latex2mathml plus a custom MathML -> OMML transformer
• docx-rs for OpenXML document generation
• image and resvg for raster and SVG asset handling
• syntect for syntax highlighting
• zip and quick-xml for DOCX package post-processing

Desktop app

• Tauri 2 for the native desktop shell
• Svelte 5 + SvelteKit for the frontend
• Vite for frontend build tooling
• Tailwind CSS 4 for styling
• TypeScript for the UI layer

Output model

• native Word paragraphs, runs, numbering, and footnotes
• editable OMML equations instead of screenshots
• embedded raster media for portable image rendering
• HTML preview generated from the same AST used for DOCX emission

Repository Layout

The product is branded as Folio. Internal crate and package names still use the historical scribe-* prefix for now so the workspace can evolve without a broad package rename.

crates/
  scribe-ast        Typed Markdown AST
  scribe-parser     Markdown -> AST
  scribe-math       LaTeX -> MathML -> OMML
  scribe-images     Image loading and sizing
  scribe-highlight  Code highlighting
  scribe-template   Template/style plumbing
  scribe-docx       AST -> .docx emission
  scribe-preview    AST -> HTML preview
  scribe-core       Shared orchestration API
  scribe-tauri      Desktop shell and frontend bridge
scribe-cli/         CLI wrapper for fixture testing
fixtures/           Focused sample inputs
test/               Comprehensive regression fixture and export artifacts
docs/               Design docs and README assets

Development

Requirements

• Rust stable
• Node.js 20+
• pnpm

Install frontend dependencies

pnpm --dir crates/scribe-tauri/frontend install

Run tests

cargo test --workspace
pnpm --dir crates/scribe-tauri/frontend check

Start the desktop app

cd crates/scribe-tauri
cargo tauri dev

Build the frontend only

pnpm --dir crates/scribe-tauri/frontend build

Shipping Releases

For end users, GitHub Releases should be the default distribution channel.

1. Bump the workspace version in Cargo.toml.
2. Create and push a tag such as v0.1.2.
3. GitHub Actions will build and publish:
   • macOS Apple Silicon .dmg
   • macOS Intel .dmg
   • Windows NSIS .exe

The release workflow also generates platform icon assets from crates/scribe-tauri/icons/icon.png before bundling.

Fixture-Based Testing

The repository includes focused fixtures under fixtures/ and a comprehensive all-in-one regression fixture under test/.

Useful local commands:

cargo run -p scribe-cli -- fixtures/english/m2-kitchen-sink.md -o /tmp/folio-m2.docx
cargo run -p scribe-cli -- fixtures/english/m3-math.md -o /tmp/folio-m3.docx
cargo run -p scribe-cli -- test/folio-comprehensive.md -o test/output/folio-comprehensive.docx
soffice --headless --convert-to pdf --outdir test/output test/output/folio-comprehensive.docx

If you have LibreOffice installed, rendering generated .docx files to PDF and PNG is a practical way to inspect layout regressions before shipping changes.

Design Notes

Longer-form planning and design material lives here:

• docs/superpowers/specs/2026-04-17-scribe-md-to-docx-design.md
• docs/superpowers/plans/2026-04-17-scribe-v1-plan.md

GitHub Activity

License

MIT. See LICENSE.