Jev Decisions · by TypeSafe, on CompanyFabric

Decide first.
Generate second.

Jev is a decision model. It never writes text. You send a situation and typed questions, and it returns one calibrated probability per question in about 100 ms, for $0.042 per million tokens. Put it in front of anything expensive.

Jev Decisions playgroundtypesafe/jev-latest · $0.042 / M input · output free

Threat, harassment and spam as probabilities, with the action as a choice.

Sign in to run

Free starter credits cover thousands of decisions. Or copy the code below and run it with your own key.

Answers arrive here as probabilities, one card per question, in about 100 ms. No text to parse, nothing invented outside the options you gave.

Runs in front of a comment form; the chat model is only called to draft the notice.

What to do with the numbers

  • severity ≥ 2 or threat > 0.7 → hide now, human reviews.
  • Never auto-suspend: remove_and_suspend is a recommendation for the reviewer.
  • Store the full distribution with the case; it is the audit trail.

This exact request, in code

curl https://api.companyfabric.com/v1/decisions \
  -H "Authorization: Bearer $COMPANYFABRIC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "typesafe/jev-latest",
  "state": "If anyone from that group shows up at the meetup again I will personally make sure they regret it.",
  "questions": {
    "threat": {
      "type": "noul",
      "instructions": "Does this content threaten harm to a person or group?"
    },
    "harassment": {
      "type": "noul",
      "instructions": "Is this targeted harassment rather than general rudeness?"
    },
    "severity": {
      "type": "score",
      "instructions": "How severe is the policy violation?",
      "criteria": [
        "None",
        "Borderline",
        "Clear violation",
        "Credible threat of harm"
      ]
    },
    "action": {
      "type": "choice",
      "instructions": "What should the system do?",
      "criteria": {
        "allow": "Publish as is",
        "hold_for_review": "Hide pending a human decision",
        "remove": "Remove and notify the author",
        "remove_and_suspend": "Remove and suspend the account"
      }
    }
  }
}'

Needs a CompanyFabric key, or add your own TypeSafe key in the BYOK vault and pay them directly at 0% fee. Cost comes back in usage.cost_usd.

How it works

One call.
Many answers.

Same key and same balance as the chat models. BYOK with your own TypeSafe key routes at 0% fee. The welcome credit covers thousands of decisions.

01

Send a state

A ticket, a brief, a transcript, a JSON record. Text or JSON, up to about 32k tokens. It is sent once, whatever you ask about it.

02

Ask typed questions

noul is a yes/no probability. choice picks one label from criteria you describe. score places on a rubric you define, from zero. Ask several in one call.

03

Threshold the numbers

Answers are probabilities and distributions, never prose. Act above a threshold, hand to a human below it, and treat a noul near 0.5 as unsure.

Put a decision
in front of the spend.

POST /v1/decisions from your code, or let your agent call the decide tool on our MCP server before every expensive step.