What it is
A general-purpose AI content moderation workflow for developers and operators. It decomposes a broad violation judgment into atomic questions, combines weighted scores, and routes items to pass, human review, or block.
How it uses Jev
Jev answers typed questions about whether content violates independent dimensions. The engine evaluates category, severity, and violation dimensions in one parallel request, normalizes each to a 0-1 violation tendency, combines them with tunable weights, and routes on the composite score plus model confidence.
Primitives:choicescore
Technique worth stealing
Decompose one broad moderation judgment into independent atomic questions, then combine weighted scores in code.
Try it
cd src/api && npm start, then POST to http://localhost:8080/audit with content and preset.
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 patternyes0.95
- Handles uncertaintyno0.11
- Measuredno0.06
- Runnableyes0.62
- Worth recommendingunclear0.42
- Model replicano0.05
- Problem scopescore on a 0–2 scale1.93
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.