What it is
A learning and benchmarking record for Jev (TypeSafe System One), the judgment model that returns typed probabilistic answers instead of free text. Aimed at developers evaluating Jev for triage, guardrails, reranking, or function calling.
How it uses Jev
Demos send a state plus typed questions to the single /v1/systemone endpoint and use the returned probabilities and confidence in code. Example: 01_triage.py asks 8 questions in one call to run a whole ticket-triage decision tree; 04_function_calling.py uses low Choice confidence to detect multi-intent requests.
Primitives:choicescorenoul
Technique worth stealing
Speculative fan-out: ask every question you might need in one call, since 40 questions cost the same latency as 1.
Try it
Set TYPESAFE_API_KEY in .env, install requirements, then run bash demo/run_all.sh.
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-centricyes0.91
- Shows a System One patternyes0.78
- Handles uncertaintyno0.06
- Measuredunclear0.52
- Runnableno0.34
- Worth recommendingyes0.73
- Model replicano0.04
- Problem scopescore on a 0–2 scale0.64
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.