What it is
A Swift 6 package that integrates TypeSafe AI's Jev System One decision model into Apple's Foundation Models framework, letting developers evaluate strongly typed @Generable structs and enums against application state with calibrated probabilities and no hallucinations.
How it uses Jev
Jev evaluates typed questions derived from @Generable schemas against application state. The bridge maps Bool to noul, enum to choice, and bounded Int to score, then uses Jev's answers and confidence to route decisions via RoutingPolicy.
Primitives:choicescorenoul
Technique worth stealing
Map @Generable schema fields to Jev primitives (noul/choice/score) and use calibrated confidence for routing.
Try it
Add package dependency, set TYPESAFE_API_KEY, create JevLanguageModel, and use LanguageModelSession.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.93
- Shows a System One patternunclear0.58
- Handles uncertaintyno0.06
- Measuredunclear0.48
- Runnableyes0.94
- Worth recommendingunclear0.49
- Model replicano0.05
- Problem scopescore on a 0–2 scale0.99
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.