How do people reliably parse PDFs with agents into valid JSON?

Slothilism · reddit · 2026-07-21

A Reddit user asks how others reliably extract structured data from PDFs with agents when plain text dumping and regex fail on spacing, broken line breaks, or image-based documents.

The workflow goal is to pull fields such as emails and MD5 hashes from vendor PDFs into valid JSON for downstream tools. The post asks whether people rely on OCR, pass the document itself instead of extracted text, or use preprocessing/postprocessing tools from GitHub.

Original post →

More from coding & agent

coding & agent channel →