RayJackson30/court_filing_reflower
PleadingReflow
Court filings (PDF) into clean, readable text: paragraphs joined and de-hyphenated, the printed 1–28 line numbers stripped, headings marked as # / ## / ###, footnotes separated — driven by a small decision-tree model trained on the geometry and typography of public California filings.
Test tool. Use public court filings only — do not upload confidential documents. Your file lives only in the session's temporary space while you work with it — destroyed when you upload another document or the session ends; nothing is stored or logged.
For confidential documents there is a desktop version that never sends files anywhere.
Scanned filings are read with on-device OCR on this Space's CPU, which is slow — expect a few seconds per scanned page. Born-digital filings are fast.
Pages marked as-is are output line for line as printed — with the pleading gutter line numbers removed. Page markers ( [START PDF PAGE 00X] ) are on by default.
