WWhatJustShipped

Repo tutorials · Open source

Markitdown

github.com/microsoft/markitdown

A full course on markitdown. We installed it, ran it on something real, and reported what happened.

Start here, in order· 4 parts, about 43 min total

1
1. Installing MarkItDown, and the Python 3.9 Trap That Installs 2.6 KB of Nothing
On a stock Mac, pip reported success in 6.18 seconds and installed a 2.6 kB placeholder with no code in it. With uv, the real thing took 26.97 seconds and converted a Word file in 0.46.
10 min
2
2. A Tax Form, a Famous Paper and a Scanned Page, and the Scan Came Back as One Newline
We fed MarkItDown eight real inputs. A scanned PDF returned 1 byte with exit code 0, and the Attention paper came back with 397 words glued together. YouTube, oddly, was flawless.
9 min
3
3. We Converted 698 Pages of IRS Guidance and MarkItDown 0.1.8 Scrambled Every Publication
Our batch of 12 IRS publications converted with zero errors. A canary check then found MarkItDown 0.1.8 kept 1 to 9 of every 50 sentences intact. Version 0.1.5 kept 39 to 48.
10 min
4
4. MarkItDown Against Pandoc, Docling and pdftotext, and Who Should Actually Adopt It
Docling read our scanned page (922 words) where MarkItDown returned 1 byte, but took 108 seconds on a file pdftotext finished in 0.10. Pandoc kept 30 bullets MarkItDown dropped.
8 min