Watermark evidence
POST /v1/watermark-probe — evidence only
Technological watermark detectors (no verdict)
Drop or choose image Evidence-only probe. Runs visible_corner.v1 (Gemini / UI marks via watermark-checker) and trustmark.v1 (stub). Needs API WATERMARK_EVIDENCE_URL plus checker at WATERMARK_CHECKER_URL (e.g. pnpm dev).

API base: window.__KANONYX_API_BASE__ or http://localhost:3000. Results stay in the right column; wide tables scroll inside that column.

Detector types (uniform evidence model)

Each row is one detector adapter with the same fields (status, confidence, notes, …). That keeps every watermark family separate until a downstream combiner merges evidence.

  • visible_corner.v1 — Pixels you can see (Gemini sparkle, other UI marks). Uses the same service as production watermark.sd.v1 (watermark-checker: templates, optional Vision/OCR). Evidence only here; not TrustMark.
  • trustmark.v1 — Technological / payload-style TrustMark (when implemented). Different signal than visible UI marks; same response shape, different adapter.

Long-term pattern: either many thin adapters in this worker, or split implementations inside watermark-checker behind one HTTP contract — either way, keep one detector id + one row per family in this lab JSON.

Run a probe to see detector status, confidence, decoded IDs, and timings. This is not a trust verdict — structured evidence only.