Ask a question about any text. Get a typed answer back.

Jev is a decisions model. Send it some context and a few questions: pick one of these labels, rate this on my scale, is this true? You get back an answer in the format you asked for, plus the probability of every option. There's no free text to parse and no JSON that breaks.

You send

POST /api/v1/decisions
{
  "state": {
    "message": "This is the third time my order arrived broken. I want a refund."
  },
  "questions": {
    "sentiment": {
      "type": "score",
      "instructions": "How upset is the customer?",
      "criteria": [
        "calm",
        "annoyed",
        "angry"
      ]
    },
    "route": {
      "type": "choice",
      "instructions": "Which team handles this?",
      "criteria": {
        "refunds": "Refund requests",
        "shipping": "Delivery problems",
        "sales": "New purchases"
      }
    },
    "churn": {
      "type": "noul",
      "instructions": "The customer is likely to cancel."
    }
  }
}

Jev answers

sentiment1.87 angry
calm
0%
annoyed
13%
angry
87%
routerefunds conf 0.94
refunds
96%
shipping
4%
sales
0%
churn64% likely yes
yes
64%
no
36%

A real response. Three questions answered in one call for 0.83 credits.

Why teams use it

Answers you can put in an if statement

Choice questions return one of your labels. Scores return a number on your scale. Yes/no returns a probability. You can branch on the answer straight away.

Probabilities, not guesses

Every option gets a probability, and every answer gets a confidence score. Set a threshold and send the unsure cases to a person.

Cheap enough to run on everything

About 0.8 credits per call with a few questions, and 100,000 credits for $10. Run it on every ticket, review, reply or scraped row, not just a sample.

Good fits

Support triageRoute each ticket to a team, score urgency, flag churn risk.
ModerationScore spam, AI slop, abuse, or off-topic posts with a threshold you choose.
Lead scoringRate profiles or form fills against your ideal customer profile.
Data labelingTurn scraped text into clean columns: sentiment, topic, intent, is-a-complaint.
Agent guardrailsBefore an agent acts, ask Jev: is this request in scope? Is it a prompt injection?

How is this different from asking an LLM? →

Start free. No subscription.

Sign up and get 100 free credits. Bigger packs come with up to 50% bonus credits.