Extract fields from a document
Invoices, contracts, IDs and bad scans, returned as the fields the next system needs.
POST
Extract fields from a document
Reads the paperwork a process is waiting on and returns structured fields (amounts, dates, names, ID and iqama numbers, VAT numbers, line items) from Arabic and English on the same page, including mixed handwriting.
Reading the result
Labels are always English so a field can be mapped to a system even when its value is Arabic. Values are verbatim. Nothing is translated and no number or date is reformatted, because a value that has been helpfully tidied is a value you can no longer reconcile. Confidence is honest.high means printed clearly, medium means legible but ambiguous, low means inferred from a poor scan or from context. Route low to a person rather than into a ledger.
Warnings are the things that would make a person distrust the page: a cut-off edge, a handwritten correction, two conflicting totals, a stamp across a field. An empty array is a meaningful result.
Limits and cost
PDF, PNG, JPEG or WebP, up to 8 MB. Four cents a document. A read that fails is not charged.Authorizations
API token from the Voho console. Begins with voho_sk_live_.
Body
Send multipart/form-data with the file attached, or JSON with the file base64 encoded.
The file itself. Accepted types: application/pdf, image/png, image/jpeg, image/webp.
Response
What the document says.
Example:
"Supplier invoice"
Anything that would make a person distrust the result: a cut-off page, a handwritten correction, two conflicting totals.
What this call cost, in cents. Also returned as the X-Voho-Cost-Cents header.

