What it is
FastJev is an open-source Python SDK and System One-compatible API for running Jev decisions on your own infrastructure, supporting Choice, Boolean, and Score questions on pinned open models across Torch, vLLM, MLX, llama.cpp, and WebGPU.
How it uses Jev
The SDK loads a pinned model, validates 2–16 options, performs scoring, and returns typed results with probabilities, token usage, timing, model revision, and prompt version. Jev makes Choice, Boolean, and Score decisions on a provided state; results are used directly in Python code.
Primitives:choicescore
Technique worth stealing
Direct logit scoring in one pass with zero output tokens, returning calibrated probabilities and provenance.
Try it
pip install 'fastjev[torch]' then use FastJev.from_pretrained with a pinned model revision.
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-centricyes0.93
- Shows a System One patternunclear0.40
- Handles uncertaintyno0.05
- Measuredno0.15
- Runnableyes0.94
- Worth recommendingunclear0.44
- Model replicayes0.92
- Problem scopescore on a 0–2 scale1.25
- About Jevyes0.97
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 22 Sept 2026.