jevbooks

← All projects

ask-jev

logicrw/ask-jev

Bounded advisory decisions and verbatim reading views for AI agents.

What it is

ask-jev is a zero-dependency Python CLI that routes small semantic judgments to Jev for AI coding agents and CLI pipelines. It supports closed-set selection, evidence checks, ordered ratings, mixed questions over one state, and verbatim passage selection.

How it uses Jev

Jev makes Choice, Noul, and Score judgments on a supplied state (observations and instructions). Results are returned as probabilities with confidence; the calling agent retains reasoning, verification, and action authority. Without both environment gates, judgments return answer: null and purification returns the full original text.

Primitives:choicenoulscore

Technique worth stealing

Use separate Noul questions for labels that can all apply; use Choice only for mutually exclusive alternatives.

Try it

Set TYPESAFE_API_KEY and HARNESS_JEV_ALLOW_REMOTE=1, then run python3 scripts/ask_jev.py choose/check/score/purify/batch.

View on GitHub

judged by Jevjev-1.13.0

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 patternyes0.72
  • Handles uncertaintyunclear0.45
  • Measuredno0.06
  • Runnableyes0.96
  • Worth recommendingunclear0.54
  • Model replicano0.04
  • Problem scopescore on a 0–2 scale1.25
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 22 Sept 2026.