What it is
A CLI that takes a JSON array of candidate documents from stdin and outputs a JSON array to stdout, using Jev to rerank, filter, or compress them for LLM context.
How it uses Jev
Jev makes separate Noul checks per candidate: rerank scores relevance, filter scores evidence, compress selects sentences/lines. Scores are added as rerankScore or evidenceScore; compressedText holds retained passages.
Primitives:noul
Technique worth stealing
Run rerank, filter, and compress as independent Jev judgments, piping one invocation into another with the same query.
Try it
npm install --global jev-reranker, export TYPESAFE_API_KEY, then pipe a JSON array with --query.
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-centricyes0.94
- Shows a System One patternno0.32
- Handles uncertaintyno0.07
- Measuredno0.05
- Runnableyes0.93
- Worth recommendingunclear0.41
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.42
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.