Document Conversion Web App · 2026

MD2TXT: Markdown-to-Document Engine

Turns Markdown and TeX from AI study notes into faithful documents, with rendered math and chemistry instead of leftover LaTeX commands, exported to clipboard, TXT, PDF, or DOCX.

View on GitHub Live demo

[ EVD ]

Engineering evidence

My contribution: Solo · Product design, conversion architecture, frontend, and export service
Outcome: Live, MIT-licensed, with opt-in real-browser export tests
Decision record 01: Used a shared parsed document model instead of scraping the rendered DOM, so preview, clipboard, TXT, PDF, and DOCX keep the same meaning.

Live preview

[ ARC ]

Architecture

One parsed document tree drives every output, so the preview and each export agree. The server reparses the source rather than trusting the client.

01
Parse (web worker)
- unified + remark-gfm
- remark-math, mhchem
- Diagnostics with source links
02
Document tree
- Shared MathJax 4 adapter
- Unsupported input kept visible
03
Browser outputs
- Formatted preview
- Rich clipboard
- Unicode TXT
04
Export service
- Playwright Chromium → PDF
- docx + equation images → DOCX
- 422 on hidden errors

[ KEY ]

Technical highlights

01
Treated Markdown structure and TeX notation as separate parsing concerns, normalised into one model for every format.
02
Blocked raw HTML and remote resource fetching, isolated each export in its own browser context, and made fallback exports an explicit user decision.
03
Supports English, Bangla, Greek, and scientific symbols across outputs.

[ STK ]

Stack

TypeScript
Vite
Unified / Remark
MathJax 4
Playwright
docx
Node.js
Vercel