Pull the text layer out of a PDF and download it as plain text
This reads the text layer and follows the content stream, so multi-column layouts and tables may come out interleaved; scans have no text layer and need OCR first. The file is parsed locally.