eKYC Suite Identity Verification

Problem

In remote onboarding, financial face-signing, and compliance review, engineers often need AI agents to inspect images or videos: check whether two faces match, whether a selfie is live, what a document contains, and whether a scene shows fraud risk. Text-only Q&A cannot inspect visual evidence, while calling raw APIs is easy to break on signing, front/back parameters, async queries, and error explanation. eKYC Suite wraps these calls into a triggerable skill: the input is an image or video, and the output is structured data plus a business-readable interpretation.

How It Works

• Face and liveness: face comparison, photo liveness, and video liveness return similarity, riskLevel, and riskTag for anti-deepfake and fraud screening.
• Document OCR: ID cards, bank cards, driver licenses, and vehicle licenses are supported, with explicit side flags such as side=0/1 for ID cards and side=1/2 for vehicle licenses.
• Image labeling: detects masks, hats, multiple people, car interiors, hotel rooms, and other labels, up to five label codes per request.
• Execution: it calls scripts/ekyc_api.py subcommands with separate KYC and label auth scripts; signatures must remain 40-character uppercase SHA1, and responses must be parsed from both the top level and the nested result field.
• Interpretation: the skill explains scores, risk levels, OCR fields, and label hits instead of returning raw numbers.

Boundaries

It only processes image or video files and does not accept text PII such as name, ID number, or phone number. It supports business decisions but does not replace human review. Use test credentials before production to avoid billing. Driver-license OCR supports only the front page, video liveness has size and duration limits, and image labeling is asynchronous with low default concurrency.