MarkItDown in production — what the docs don't cover

[Jakeshadow]

Been using MarkItDown in production for document processing. Here's what I hit that the official docs don't cover:

1. Encrypted PDFs fail silently — no error, just empty strings
2. Large files have no timeout — need a wrapper
3. Complex tables lose structure — pre-processing helps
4. PDF output is noisy — CID markers, duplicates, flat structure
5. Dependency fragility — unpinned versions silently break

Each fix and production setup guide here: https://markitdown-pro.com/projects/markitdown/

Also built a Docker + FastAPI wrapper with file size limits and timeout handling. What production issues have others hit?

[hasski0x]

UI for uses,

[krickert]

My PR will emit the pages as they are done with PDFs and PPT.. I can add further support but waiting to see if I can get it merged. That could eliminate your PDF timeout woes.

[hasski0x]

I did not push that yet! I did i locally