Upload a document outside the context of a capture session. Optionally attach it to an existing session or identity. The document is stored and available for classification, extraction, and verification via the sub-resource action endpoints.
Required scope: documents:write
Organization API key. Keys are prefixed with vk_live_ (production) or
vk_sandbox_ (sandbox).
API version date string (e.g. 2026-02-01). If omitted, the version
pinned to your API key is used.
The document image file (JPEG, PNG, WebP, or PDF). Max 10 MB.
Optional document classification (e.g. passport, drivers_license).
Document ID (doc_*).
Detected or assigned document classification (e.g. passport, drivers_license).
Number of pages (for multi-page documents).
Arbitrary key-value metadata.
How the document was uploaded (api, capture, dashboard).
HMAC-signed download URL (when available).
Server-side JPEG preview URL for page 1. Works for both PDFs (rendered via PyMuPDF) and images. Same HMAC signature as download_url. Null for redacted documents.
Structured field extraction produced by Verifa’s document
extractor. Populated when the document’s matched
document_type has non-empty extraction_hints. Each key is
one of the field names from the type’s extraction_hints,
and each value is either the extracted string or null if
the field was not visible in the document. null until the
verification pipeline has run.
Per-field token-presence confidence (0.0-1.0) for the
corresponding entries in extracted_fields. Computed by
checking whether each extracted value’s tokens actually
appear in the underlying OCR text. Values below ~0.5 are
likely hallucinations and should be flagged for reviewer
attention. null until extraction has run.