Decision model
The zenifra/kev-4b model does not write text: it reads some content (a ticket, an email, a document) and answers closed questions about it, always with probabilities. Use it for triage, routing, moderation and classification where you need to know how sure the model is, not just the answer.
It is based on Kev, an open model under the Apache-2.0 license, and follows the systemone request format, so clients that already use that format work by changing the endpoint and the key.
Before you start
In the console, open AI Models and use a key that has zenifra/kev-4b among its Allowed Models (or the option for all of the organization's models). See Getting started with AI to create the key and set its quotas.
Question types
| Type | What you send | What comes back |
|---|---|---|
choice | 2 to 255 options, each with a description | choice (most likely option), probabilities and confidence |
noul | just the question; yes and no descriptions are optional | noul: probability of "yes", from 0 to 1 |
score | 2 to 10 levels, from lowest to highest | score (mean level, starting at 0), legend, probabilities and confidence |
A request accepts 1 to 128 questions about the same content. Each question is answered independently.
Example
curl https://ai.zenifra.com/v1/systemone \
-H "Authorization: Bearer $ZENIFRA_AI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "zenifra/kev-4b",
"state": "Shoes arrived two weeks late and in the wrong size. Also I see two charges on my card.",
"questions": {
"department": {
"type": "choice",
"instructions": "Which team should handle this?",
"criteria": {
"returns": "Exchanges, refunds, wrong or damaged items",
"shipping": "Delivery status, delays, lost packages",
"billing": "Charges, invoices, payment problems"
}
},
"escalate": {
"type": "noul",
"instructions": "Does this need urgent human attention?"
},
"frustration": {
"type": "score",
"instructions": "How frustrated is the customer?",
"criteria": ["Calm", "Frustrated", "Very angry"]
}
}
}'Response:
{
"answers": {
"department": {
"type": "choice",
"choice": "returns",
"confidence": 0.21,
"probabilities": { "returns": 0.47, "shipping": 0.28, "billing": 0.25 }
},
"escalate": { "type": "noul", "noul": 0.93 },
"frustration": {
"type": "score",
"score": 1.44,
"confidence": 0.34,
"legend": { "0": "Calm", "1": "Frustrated", "2": "Very angry" },
"probabilities": { "0": 0.0, "1": 0.56, "2": 0.44 }
}
},
"usage": { "input_tokens": 101, "output_tokens": 161 }
}The ticket mentions a return, a late delivery and a billing problem, and the department probabilities show it: confidence is low. A common pattern is to decide automatically only the high-confidence cases and send the rest to a person.
confidence measures how concentrated the probabilities are on one answer (0 = spread out, 1 = all probability on one option). It is not a measured accuracy rate.
Price
The model is billed only for input tokens: BRL 0.2244 per million tokens. Output tokens are free. Usage shows up under AI Models in the console, like the other models, and respects the key's quotas. Always check the current price in the catalog (GET https://ai.zenifra.com/v1/models) or in the console.
Limits
- content (
state) up to 65,536 tokens; longer content is rejected, never truncated; - no streaming.
Responsible use
Do not use the model to make automated decisions with consequences for people (credit, hiring, health, access to services, punitive moderation) without human review. The probabilities help you prioritize; they do not replace a person's judgment.
Next steps
Last updated on