01 / C
C
Choice
Pick one option from a list you define. Useful for routing, intent, and tool selection.
intentbillingconfidence · 0.94
FIELD GUIDE · INDEPENDENT
Jev makes a decision.
Your software takes it from there.
A practical guide to TypeSafe AI's Jev model: what typed decisions are, where they fit, and what to verify before you build.
Independent resource. Jev is a product of TypeSafe AI; this site is not affiliated with or endorsed by TypeSafe AI.
DECISION FLOW / 01● EXAMPLE
STATE
“Charged twice and need a refund before Friday.”
3 questions · 1 requesttyped output
CHOICEroutebilling0.94
SCOREurgencyhigh2.6 / 3
NOULreply_todayyes0.89
Illustrative output · not a live model call↗
01 / 06
DECISION LAB · STATIC EXAMPLES
See the shape of
a Jev decision.
Choose a workflow to inspect its input state, typed questions and possible application branch. All outputs below are illustrative fixtures, not live model results.
01 / STATE
Support ticket
A short support message becomes the shared context for multiple independent decisions.
“I was charged twice for my annual plan. Please refund one charge before Friday or I’ll dispute both payments.”
02 / QUESTIONS3 typed questions
03 / TYPED ANSWERSFIXTURE · NO API CALL
APPLICATION BRANCH↳
Inspect response shape +
THREE PRIMITIVES
One state. Three ways
to make a decision.
Jev is designed for decisions with a defined output space, rather than open-ended text generation. The key question is whether a typed answer helps your application act more reliably.
02 / S
S
Score
Place a state on an ordered rubric, such as severity, quality, or lead fit.
severity2.6 / 3probability by level
03 / N
N
Noul
Estimate the probability of a yes/no statement so your code can set its own threshold.
escalateyes · 0.89your threshold · configurable
Illustrative examples only. Output shape and supported fields should be checked against the current official API reference.
WHERE IT FITS
Too fuzzy for an `if`.
Too small for a chat.
Use a decision model for narrow judgments inside a larger system. Keep a generative model for the parts that need language, reasoning, or a conversation.
Explore use cases ↗START WITH A QUESTION
Build an understanding
before a dependency.
01What is Jev AI?System One, typed output, and the distinction from generative modelsCONCEPT↗02Jev AI use casesRouting, classification, scoring and agent guardrailsPATTERNS↗03Jev API guideRequest anatomy, typed questions and integration checksDEVELOPERS↗04Benchmarks & limitationsHow to read speed, cost and accuracy claims carefullyEVIDENCE↗
J.
A GOOD FIRST TEST
Start with your hardest branch.
Write down the decision, the available outcomes, and what happens when the answer is uncertain. Then compare Jev with the simplest alternative on your own examples.