Jev Score questions
A Score rates the state along a rubric you write. criteria is an ordered array of level descriptions. The API wants at least two levels and accepts up to 10. The answer includes score (expectation, may sit between levels), legend, probabilities, and confidence.
This page is for builders who already heard of Jev / System One and want a production-minded mental model. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Confirm facts on docs.typesafe.ai. We do not sell keys.
Score is the primitive for “where on this ladder?” It is not a dollar field, not a star rating you can multiply into revenue, and not a substitute for counting in code. Keep arithmetic in your service. Treat the returned score as a fractional expected level, then decide.
How to use
Write levels as descriptions a person could apply, not as naked integers. “Calm / Frustrated / Very angry” is a rubric. "1" / "2" / "3" is a hope that the model shares your private scale.
{
"frustration": {
"type": "score",
"instructions": "How frustrated is the customer?",
"criteria": ["Calm", "Frustrated", "Very angry"]
}
}Use Score when the action is a threshold: escalate if expected severity is past level 1; hold if it sits between “low” and “medium.” Read the full probabilities map when the runner-up level would change the queue. Confidence is the peakedness of that map — see confidence.
Batch Score with Choice and Noul in the same POST /v1/systemone. Compose atomic questions in code. Do not ask Jev to average three Scores; that is the composite scoring pattern: three questions, weights in your repo.
When not to use Score
- The options are unordered (department, language, document type). Use Choice.
- You need a yes/no probability. Use Noul. A Noul of 0.5 is uncertain, not “medium.”
- You want a physical quantity: refund dollars, minutes of delay, token counts.
jev-1.13levels are “weak in numerical calibration” according to TypeSafe’s jaggedness note. Do not interpolate the expectation into a dollar or a minute. - You have more than ten meaningful rungs. Collapse the rubric or split factors.
- You are treating Score as a price, a KPI, or a billing unit. It is an expected index on your legend.
The jaggedness note is specific: do not use Score expectations to reconstruct the exact magnitude of a number between two levels. You may threshold the expectation (“is this past level 1?”).
Split multi-factor scores
Ticket priority that mixes severity, frustration, and engineer-readiness is three Scores, then weights in code. That is the composite scoring pattern. Keep math in code. If a reviewer later asks why priority was 2.4, the decision trace should show the three expectations and your weights — not a single opaque “priority” question.
Production notes
Pin the model once you fit floors on a Score. Changing a level description is a new decision contract even if the pin stays put. Replay gold before you promote. Label any latency or dollar figures you print next to Score as vendor claims unless you measured them.
Hub: Primitives. Glossary: Score. Official: docs.typesafe.ai/primitives/score.
See also
FAQ
Does Score have confidence?
Yes. Choice and Score include confidence. Noul does not.
Can score be 1.05 on a 0–2 rubric?
Yes — it is an expectation, not an enum index.
Should I ask Jev to average three Scores?
No. Average in code.
Is a Score of 4.2 four dollars?
No. Do not treat score as dollars or any physical unit.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.