Actions10
Document → Extract Data From Bank Statement
AI-generatedSummary
Extracts account holder details, balances, statement period, and transactions from a bank statement document using Veryfi's OCR service.
Inputs
- fileSource — Source of the document file: 'binary' for uploaded files or 'url' for a public URL Veryfi fetches directly.
- binaryPropertyName (required) — Name of the input binary field containing the file (PDF, JPEG, PNG, or TIFF) when fileSource is 'binary'.
- fileUrl (required) — Public URL of the document when fileSource is 'url'. Must be accessible without authentication.
- fileName — Optional custom file name to send to Veryfi. If omitted, the detected filename from the binary input is used.
- confidenceDetails — Include per-field confidence scores (value, score, ocr_score) in the response.
- flattenConfidence — Flatten confidence objects into sibling keys (e.g., field, field__score, field__ocr_score) for easier mapping. Only applies when confidenceDetails is true.
- auto_categorize — Enable Veryfi's automatic document categorization.
- boost_mode — Enable boost mode to return extracted data without storing the document image.
- categories — Optional list of categories to constrain the auto-categorizer.
- external_id — Optional external identifier to associate with the document.
- tags — Optional tags to attach to the document.
Output shape
One record per input item containing extracted bank statement data (account holder, balances, period, transactions) and optional confidence details.
When confidenceDetails is true, fields are returned as objects with 'value', 'score', and 'ocr_score' unless flattenConfidence is enabled, which flattens them into sibling keys. Confidence details are not supported for this operation.
Examples
Example 1: Process a bank statement PDF uploaded from a previous 'Read Binary File' node.
File Source: Binary, Input Binary Field: data, Confidence Details: false
Example 2: Extract data from a bank statement hosted at a public URL with confidence scores flattened.
File Source: URL, File URL: https://example.com/statement.pdf, Confidence Details: true, Flatten Confidence Scores: true