ADVANCED / STAGE 1 OF 6
Parse structure and figures
Read structured text and tables, with selected visual descriptions.
Why this stage exists
Allows table structure and selected visual descriptions to enter the searchable text.
How it works
Structured document text is combined with available text recognition and figure descriptions.
Tech used in this stage
Docling
Parses structured text and tables from the local PDF.
libraryRapidOCR
Runs local text recognition during structured parsing.
libraryPyMuPDF
Prepares selected figure images from PDF pages.
libraryOpenRouter
Can describe selected figures for searchable text.
service