How to read Jev answers: Choice, Score, and Noul

Read question IDs first

One request may contain several independent questions. Answers are keyed by your IDs, such as department and is_urgent. Check each ID, type, and required field rather than relying on position. IDs index the response for code.

Choice: selection and distribution

choice is the selected key. probabilities maps candidate keys to probabilities. confidence is a separate reported field. Neither the maximum probability nor confidence is a measured accuracy guarantee.

Score: a defined rubric

score is a value along your criteria scale; probabilities shows the distribution, and legend names the levels. Read the rubric before mapping a level to an application action.

Noul: probability of a statement

noul is a 0–1 value for the statement. It has no confidence field. The meaning changes if you reverse the statement, so keep the wording and test examples together.

Map output to safe actions

A validated threshold may allow reversible routing. Missing fields, uncertainty, and errors should reach a review path. Refunds, deletion, and access changes still require deterministic controls.

Practice

Imagine a Choice answer of shipping with confidence 0.64 and a threshold of 0.80 chosen on your labeled data: send it to review. These numbers are illustrations, not a measured Jev result or recommended threshold.

An annotated excerpt of the official response

These fields and values are excerpted from the official quickstart support example to explain the shape. They are not a run by this site. Code still owns the next action.

{
  "answers": {
    "department": {
      "type": "choice",
      "choice": "technical",
      "confidence": 0.78,
      "probabilities": {
        "technical": 0.85,
        "sales": 0.0,
        "billing": 0.15
      }
    },
    "is_urgent": {
      "type": "noul",
      "noul": 1.0
    }
  }
}

Official docs