Compose Jev questions. Run them against real state. Test them when the model changes. Three tools, one workflow.
OPEN BUILDERBuilt on TypeSafe AI's Jev modelDecisions, not text.
Write Jev questions visually instead of hand-typing JSON. Nouls, scores, and choices with live preview.
Bring your own Jev key and run questions against any state. See the actual probability distribution, not just a decision.
Regression tests for AI decisions. Write cases with expected answers, re-run when the model updates, and see exactly what changed.
You deploy a prompt or a criteria set. Later the model updates, or someone tweaks a definition, and the thing that used to say yes says no. Nobody catches it. Nobody tested it.
Traditional AI has no answer for this. Text output is too stochastic to test.
Jev produces stable, probability-scored decisions. That makes them testable — if you have the right tool.
Suites is that tool. Write cases with expected answers. Run the suite whenever something changes. See exactly which decisions drifted.
AI decisions are treated asmagicinstead of engineering. You ship a prompt, hope it holds, and find out it didn't when a customer tells you.
JCT Labs makes decisions composable, observable, and testable. The same way you'd treat any other piece of software.
Open the Builder with no account. Create one when you're ready to run and test decisions.