What it is
zod-jev composes TypeSafe Jev semantic checks into Zod 4 schemas, so TypeScript projects can turn model-only judgments about free text into calibrated probabilities their code can threshold on.
How it uses Jev
Each rule becomes one noul question sent with state { value, context? } in a single request per parse. Jev returns a probability per rule; zod-jev compares it to a threshold and emits Zod issues (rejected, uncertain, or unavailable).
Primitives:noul
Technique worth stealing
Derive criteria.true and criteria.false so "not enough evidence" does not collapse into "no", keeping the uncertain band meaningful.
Try it
npm install zod-jev zod, set TYPESAFE_API_KEY, then call createJevZod() and semantic().
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-centricyes0.84
- Shows a System One patternyes0.92
- Handles uncertaintyno0.28
- Measuredno0.05
- Runnableyes0.96
- Worth recommendingunclear0.41
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.50
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.