Actions10
Document → Extract Data From Invoice
AI-generatedSummary
Extract structured data from invoices using Veryfi's OCR API.
Inputs
- fileSource — Where the document comes from: binary data from a previous node or a public URL.
- binaryPropertyName (required) — The name of the input binary field that contains the file (PDF, JPEG, PNG, or TIFF). Required when fileSource is 'binary'.
- fileUrl (required) — Public URL of the document. Veryfi will fetch and process it server-side. Required when fileSource is 'url'.
- fileName — Optional file name to send to Veryfi. If left empty, the file name is taken from the binary input (or omitted for URL input).
- confidenceDetails — Whether to include per-field confidence scores in the response. Veryfi will return each field as { value, score, ocr_score }. Available for extractInvoice, extractReceipt, extractCheck, extractBankStatement, extractW9, extractW2.
- flattenConfidence — Whether to flatten { value, score, ocr_score } objects into field, field__score, and field__ocr_score sibling keys for easier downstream mapping. Only active when confidenceDetails is true.
- additionalOptions — Optional collection of additional options including auto_categorize (default true), boost_mode (default false), categories, external_id, and tags.
Output shape
a single record
Returns structured data extracted from the invoice. When confidenceDetails is enabled and flattenConfidence is true, nested fields are flattened using '__' as a separator.
Examples
Example 1: Extract invoice data from a binary file uploaded from a previous node
Set fileSource to 'binary', provide the binaryPropertyName, and optionally set confidenceDetails to true.
Example 2: Extract invoice data from a publicly accessible URL
Set fileSource to 'url', provide the fileUrl, and optionally set confidenceDetails to true.