anytomd¶

Any document to clean Markdown, for LLMs. The Python binding of anytomd-rs, a pure-Rust engine by developer0hye.

pip install anytomd

import anytomd as ad

document = ad.convert("report.docx")
print(document.markdown)

# Quarterly Report

## Summary

Revenue grew by **18 percent** and costs fell by *4 percent*.

| Region | Sales |
|---|---|
| EMEA | 1200 |

Three verbs, each with its async twin:

Sync	Async	Does
`ad.convert(source)`	`await ad.aconvert(source)`	Document → Markdown
`ad.detect(source)`	`await ad.adetect(source)`	Which format is this?
`ad.describe(image)`	`await ad.adescribe(image)`	Image → description, with a vision model

Extras¶

The base package has no Python dependencies. Extras add a capability, and once installed they work on their own:

Extra	Adds
`anytomd[pdf]`	PDF conversion
`anytomd[ocr]`	The text inside images
`anytomd[vlm]`	Image descriptions from any OpenAI-compatible vision model, and `describe`
`anytomd[full]`	All of them

pip install "anytomd[full]"

Format by format → Formats · every call → API · what can fail → Errors

anytomd-rs

Conversion quality, format coverage and performance belong to anytomd-rs. Wrong Markdown is an upstream issue; anything about the Python API is ours.