What it is
An OpenAI-compatible /v1/chat/completions proxy that gates an agent's shell commands through a calibrated Jev Choice decision with fail-closed semantics.
How it uses Jev
Jev answers one Choice question (approve / deny / escalate) with a calibrated probability distribution. The bridge uses the winning class and thresholds to return APPROVE, DENY, or ESCALATE to the host's guardian call.
Primitives:choice
Technique worth stealing
Adapt the host's approval call rather than patching the host, so upgrades cannot lose the gate.
Try it
git clone https://github.com/oppih/approval-judge-bridge && cd approval-judge-bridge && python3 -m judge_bridge
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-centricno0.20
- Shows a System One patternyes0.71
- Handles uncertaintyyes0.76
- Measuredno0.09
- Runnableyes0.91
- Worth recommendingno0.37
- Model replicano0.30
- Problem scopescore on a 0–2 scale1.49
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.