Evaluate an LLM output

Scored by the deterministic layer and routed by the use-case profile. Every run is written to the tamper-evident audit trail.

Use case PSPV Case Preprocessing (auto-approve 0.92) Site Identification (auto-approve 0.90 default) Document Authoring Capability (auto-approve 0.90 default) HAQ/RTQ Contract Extension (auto-approve 0.95) Generic (0.90 default)

Mode deterministic demo (simulated LLM) full (real judge — needs key)

LLM output (required) Input / prompt (optional) Reference / ground truth (optional) Operator id

Run evaluation