Answer types and questions
Native answers are inferred from decisions. Questions are optional and always mean Jev.
Native execution takes decision + state. Answer names and types come from every named decision in that model, including helpers. There is no repeated native questions definition.
Read native answers
| Decision type | Response type | Value field |
|---|---|---|
| Boolean | noul |
.noul: 0 for false, 1 for true with matching facts |
Named choices, such as refund|review|reject |
choice |
.choice |
Ordered levels, such as low<medium<high |
score |
.score and .level |
| Number | number |
.number |
| Date | date |
.date, an ISO date string |
| Other values, objects or lists | value |
.value, preserving the JSON value |
Selected answers from the recorded cancellation example:
{
"answers": {
"refund_eligible": { "type": "noul", "noul": 1 },
"refund_amount": { "type": "number", "number": 20, "confidence": 1 },
"processing_deadline": { "type": "date", "date": "2026-10-28", "confidence": 1 }
}
}
Choice and Score answers also include probabilities. With matching facts, results are deterministic. Jev evidence can introduce uncertainty. See Determinism and confidence.
Questions mean Jev
Top-level questions is optional, and at POST /v1/systemone it always invokes Jev — never model authoring. Jev accepts noul, choice and score; it does not create an aityx model or calculate native Number and Date answers.
{
"questions": {
"damage_probability": {
"type": "noul",
"instructions": "Does customer_email explicitly report damage to the delivered item?",
"criteria": {
"true": "Damage or a defect is explicitly reported.",
"false": "The customer reports no damage."
}
}
},
"state": { "customer_email": "The item arrived broken." }
}
Sent alone, Jev’s answers are the result. Add a native model to calculate outcomes from Jev’s evidence and your account facts; the Decision then returns the consolidated result. Leave questions out entirely to run the Decision alone. See Using Jev with aityx.

