Goal: a PDF invoice in, verified accounting JSON out - without trusting an OCR you cannot taste-test.
POST /v1/document-pipeline {"url": "https://.../invoice.pdf"}
-> stages: parse (markdown + ocr_quality) -> fields (invoice JSON)
-> verify (tax math, totals, JP registration check digit)
Flat $0.05 per document (text PDFs to 20 pages, scans to 3). Inputs are validated before any payment settles; every stage outcome is recorded honestly in the response and on the receipt.
POST /v1/pdf-info $0.001 is it scanned? how many pages? POST /v1/parse/quote free exact parse price before paying POST /v1/parse $0.005/doc or $0.012/page (scanned) POST /v1/verify $0.005 arithmetic + registration checks
Japanese invoices are a design focus: Japanese-strong OCR routing and JP registration number (T+13) check-digit validation.