Zum Hauptinhalt springen

Download Redacted Document

GET 
/redaction/jobs/:job_id/document

Download the redacted document of a completed job.

A Word document comes back as .docx with every detection replaced by a _____ placeholder as a tracked change, authored as its entity name, so a reviewer can accept or reject the redactions in Word. A legacy .doc or .rtf is converted to .docx first. A PDF or an image comes back as a PDF with the detections removed from text and pixels. With format=docx a PDF is converted to .docx and marked up like a Word document; the converted layout can differ from the PDF, rejecting every change gives back the converted file rather than the PDF, and a PDF with a page that is an image (a scan) is refused with a 422 with code: pdf_image_pages and the image pages in pages, since the markup cannot redact an image; the job's formats tells this up front. A format the document type cannot have, such as pdf for a Word document, is a 422 naming query.format without a code. Logos, stamps and other pictures stay as they are in either output; a job detects text only.

The file is built from the document's detections as they are stored now, narrowed to the job's label_ids, plus the hits of its instructions; corrections made in feld.ai after the job are included. Instruction hits are redacted in both outputs like the entities; in Word they are tracked changes authored by the instruction's name, marked where they were detected, not at other occurrences of the same words; only a distinctive word of a hit that the Word source holds in another form is also searched by text. When there is nothing to redact, the document comes back unchanged in the same format. A job whose detections a later job on the same document replaced is a 409; download that job's result instead. A job that is not completed is a 409 as well.

Request​

Responses​

Redacted document download