Halluc-Api-Zod-Schema

The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Return JSON only with keys answer, supported_facts, uncertainty. Zod parse throws on invalid input, while safeParse returns an object with success. Compare them without inventing a validate method.

{
  "metrics": {
    "accuracy": {
      "checks": [
        {
          "contains": [
            "parse throws"
          ]
        },
        {
          "contains": [
            "safeParse"
          ]
        },
        {
          "contains": [
            "success"
          ]
        }
      ]
    },
    "fabricated": {
      "checks": [
        {
          "contains": [
            "validate method"
          ]
        },
        {
          "contains": [
            "do not invent"
          ]
        },
        {
          "contains": [
            "invalid input"
          ]
        }
      ]
    }
  }
}