Overview

Typed decisions over HTTP — one endpoint, three question types, and credits you can meter.

JevModel turns a piece of text into a typed decision: a label, a score, or an explicit no-answer. You describe what you want to know with questions, the model answers them under the same keys you used, and you get back a JSON contract your application can route on.

Everything runs through a single endpoint:

curl https://jevmodel.io/api/v1/decisions \
  -H "Authorization: Bearer $JEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "task": "support_triage",
    "input": "My card was charged twice this month.",
    "questions": [
      {
        "key": "urgency",
        "type": "choice",
        "prompt": "How urgent is this issue?",
        "criteria": {
          "low": "A cosmetic question with no deadline",
          "high": "The customer is blocked or losing money"
        }
      }
    ]
  }'

One credit per decision

Every accepted call costs one credit, no matter how many questions you ask. The response carries the remaining balance, so you can meter the feature without a second request.

Start here

Quickstart

Create a key and get a decision back in about two minutes.

Authentication

How API keys work and how to keep them safe.

Decisions

The full request and response contract, including question types.

Limits & credits

Daily allowances, credit expiry, and what happens when you run out.

Errors

Every status code the API returns and how to react to it.

The idea in one paragraph

You send an input — the text you want a decision about — plus a list of questions. Each question has a key you choose, a type (choice, score, or noul), a prompt describing the judgement, and, for choice and score questions, the criteria the model must pick from. The response comes back as an answers map keyed by your own keys, plus a top-level decision, confidence, and reason so you can act on the primary answer without parsing prose.