POST /api/v1/verify-unified

Agent & microservice path. One endpoint for images and PDFs — automatic routing, unified score envelope, fast runtime decisions.

Choose this when: an AI agent or service mesh needs a single JSON contract (fraud_score, verdict, detector_breakdown) without parsing the full forensic tree. Lineage still applies — low scores can floor to SUSPICIOUS for HITL.

When to use

Autonomous agents making intake-time routing decisions
Microservices that accept mixed image + PDF uploads on one integration
Need detector_breakdown + unified fraud_score in one object
Optional customer_id for cross-type intelligence hooks

For audit-grade custody and richest analyst presentation (including tamper_regions on document scans), use POST /verify.

POST /api/v1/verify-unified
Content-Type: multipart/form-data
Authorization: Bearer {api_key}

Input (multipart)

Field	Required	Description
`file`	one of file / file_url	Image or PDF upload
`file_url`	one of file / file_url	Public HTTPS URL
`detail_level`	no	`basic`, `detailed`, `fast`, `full` — routing selects analyzer mode
`customer_id`	no	Enables cross-type intelligence when set
`document_type`	no	Document class hint
`preserve_exif`	no	Images only
`source`	no	Platform hint

Response shape (top level)

{
  "verification_id": "VER-PDF-20260610-a1b2c3",
  "file_type": "application/pdf",
  "fraud_score": 42,
  "confidence": 0.88,
  "verdict": "manipulated",
  "detector_breakdown": { ... },
  "analyst_summary": { ... },
  "recommendations": [ ... ],
  "red_flags": [ ... ],
  "image_result": null,
  "pdf_result": { ... }
}

Unified verdict bands

< 30   → AUTHENTIC
30–49  → MANIPULATED
50–74  → SUSPICIOUS
≥ 75   → FRAUDULENT

Lineage hits may floor verdict to SUSPICIOUS even on low scores.