CONCEPTS

Transparency & Evals

A verdict you can verify is only half the promise. The other half is whether the verdict is any good. Fidacy publishes that too: accuracy, the direction of its errors, and how much of the decision runs on the deterministic hot path — on a public endpoint anyone can read, and every verdict carries the policy_version it was produced under.

The public transparency endpoint

GET/v1/transparency

Returns the engine’s current evaluation metrics, computed from a labeled, human-reviewed ground-truth set. No authentication, no account.

{
  "labeled_cases": 6,
  "accuracy": 1,
  "cost_weighted_error": 0,
  "error_favors_fail_safe": true,
  "hot_path_share": 1,
  "model_version": "fidacy-risk-0.1.0",
  "generated_at": "2026-06-28T09:46:52Z"
}
labeled_cases

Size of the human-reviewed ground-truth set the metrics are computed against.

accuracy

Share of cases where the engine’s decision matched the human label.

cost_weighted_error

Error weighted by how costly each mistake is — a wrongful approve counts far more than a wrongful review.

error_favors_fail_safe

Whether the remaining errors lean toward the safe side (review/deny) rather than wrongful approve. The invariant the whole engine is built around.

hot_path_share

Fraction of the decision made by the deterministic, replayable hot path (versus the optional reasoning layer).

model_version

The version label that also appears as policy_version in every signed verdict, so you can tie a payload to a published metric set.

Errors that favor fail-safe. A model that is wrong by wrongly approving is far worse than one that is wrong by asking for review. cost_weighted_error and error_favors_fail_safe make that direction public, not just the headline number.

How the engine improves: the eval loop

Capture

Real assessments are captured as cases — structured signals only, no payment credentials or PII.

Label

A human reviews a case and assigns the correct outcome, building a ground-truth set.

Evaluate

A pure, deterministic eval engine replays the cases and computes the metrics, including cost-weighted error and the fail-safe direction.

Propose

A change to the rules is proposed against the ground-truth, with its expected effect. A human approves or rejects it; nothing ships itself.

Version

An accepted change bumps the policy_version, stamped into every signed verdict from that point on.

Tie a verdict to a published version

Every signed Risk Payload carries the policy_version it was produced under. Combined with the public metrics, an auditor or counterparty can take a verdict from months ago, read which model version produced it, and see the accuracy and error profile of that version — without contacting Fidacy. The signature proves who decided and that nothing was tampered; the transparency endpoint shows how well the decider performs and which way it errs.

← PREVIOUS
Payment Firewall
NEXT →
API Reference