SchemaLock

SchemaLock — parse a contract for its key terms as structured JSON data from a PDF or image (parties, dates, term, payment terms, termination clause, governing law, obligations, signatures). Unlike generic OCR or text-only extraction APIs, this takes the raw file directly and validates date logic (expiration after effective date), one automatic retry. Still fails after that, you aren't charged. Per-field confidence included.

SchemaLock — extract structured JSON data from an invoice PDF or image (vendor, line items, subtotal, tax, total, dates, payment terms). Unlike generic OCR (raw text) or text-only extraction APIs, this takes the raw file directly and validates against source math: line items checked against subtotal/tax/total, dates and currency checked, one automatic retry. If it still fails after that, you aren't charged — settlement is skipped, not just flagged.

SchemaLock — extract structured JSON data from any PDF/image or raw text/HTML, converting it into a JSON Schema you supply, not limited to invoice/receipt/contract/resume. Unlike the fixed invoice/receipt/contract/resume endpoints, this accepts any JSON Schema — you define the fields. Takes a PDF/image file OR raw text/HTML you already have (e.g. a scraped page). Schema-conformance, not fact-checking. A rejected schema is a 400 and costs nothing.

SchemaLock — extract structured JSON data from a receipt PDF or image (merchant, items, subtotal, tax, tip, total, payment method). Unlike generic OCR (raw text) or text-only extraction APIs, this takes the raw file directly and validates against source math: line items checked against subtotal/tax/total, dates and currency checked, one automatic retry. If it still fails after that, you aren't charged — settlement is skipped, not just flagged.

SchemaLock — extract structured JSON data from a resume PDF or image (contact info, work experience, education, skills). Unlike generic OCR or text-only extraction APIs, this takes the raw file directly and validates email format and work-history date order, one automatic retry. Still fails after that, you aren't charged. No content stored after the response returns.