VerdictThe evaluation engine API

Score evidence with decision models, bring humans into the loop, and apply a policy you control.

Decision models first

Jev and Clef return probabilities over your rubric levels. Shared judge and evidence means one request per rubric scoring batch.

Humans in the loop

Assign reviewers, collect independent ratings, and escalate uncertain decisions through a task inbox.

Reproducible policy

Mix deterministic benchmarks with model scores. Freeze each rubric and reaggregate stored ratings without another provider call.

Evaluate without a provider key

bash

npx @verdict-ai/cli eval -r rubric.yaml -i ./release-notes.txt

Save this as rubric.yaml:

yaml

name: Release notes
criteria:
  - id: migration
    name: Includes migration guidance
    judge: { type: benchmark, method: keywords, keywords: [migration], mode: any }

Use REST, the SDK, or MCP to add the same evaluation step to a product or agent.

VerdictThe evaluation engine API

Decision models first

Humans in the loop

Reproducible policy

Evaluate without a provider key ​

Evaluate without a provider key