What it is
A minimal router for pi that ranks the whole OpenRouter catalog per task, filters to fitting models, computes a Pareto frontier over quality, cost and latency, then picks the knee point. For pi users who want automatic model selection.
How it uses Jev
src/classifier.ts makes one Jev call with five questions to classify the task. The result feeds the router's filtering and selection. Jev is used only for classification; the README does not specify the exact question types.
Technique worth stealing
Pareto frontier knee point selection without weights, falling back to a weighted value function when the frontier is too small or flat.
Try it
npm install; set OPENROUTER_API_KEY; copy config/models.openrouter.json to ~/.pi/agent/models.json; start pi.
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-centricyes0.63
- Shows a System One patternunclear0.46
- Handles uncertaintyunclear0.41
- Measuredno0.05
- Runnableyes0.79
- Worth recommendingno0.27
- Model replicano0.37
- Problem scopescore on a 0–2 scale1.05
- About Jevyes0.73
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.