Changelog

Verify is a new capability for detecting document fraud. Submit one file and get back a document-level riskLevel (LOW, MEDIUM, or HIGH), a plain-language summary, and the detected signals, each with typed, locatable evidence you can show a reviewer. See the Verify overview.

• POST /verify — verify a file and wait for the result (sync; 5-minute timeout. Use POST /verify_runs for verify_agentic)
• POST /verify_runs — start a verify run (async)
• GET /verify_runs/{id} — poll a verify run
• GET /verify_runs — list verify runs, filtering by status, source, sourceId, riskLevel, or fileNameContains
• POST /verify_runs/{id}/cancel — cancel an in-progress verify run
• DELETE /verify_runs/{id} — delete a verify run and the data derived from its document

Configure a run with config.engine (verify_agentic, which parses the document and runs every check through an agent harness optimized for tamper detection, or verify_light for faster checks that skip the agentic content and OCR checks), and config.advancedOptions.requireParsing to add the parse-backed checks to a light run. Verify is in beta: the endpoints return 404 unless Verify is enabled for your organization. Contact support@extend.ai to request access. Inputs are PDF, JPEG, and PNG files, supplied by URL or file ID; webhook events for verify runs are not available yet, so poll for results.