Skip to content

Jev directory

JudgeJev

An educational lab for DeepEval plus Jev: replay 17 recorded Jev evaluations of fictional support cases, inspect raw requests and responses, adjust release gates and drift playback, or run fresh evaluations with your own key.
Stars
1
Language
Python
License
MIT
Reading time
3 min
Published by
Community
Author
shyamsridhar123
Last verified
Oct 2026

Highlights

  • The public site replays 17 real recorded Jev evaluations and makes no model calls; its drift sequence is reconstructed and labels reused samples.
  • It opens with a false shipping guarantee that scored 15.1% beside a correct control at 95.5%, showing why a five-pair candidate fails a release gate.
  • Exercises LLMTestCase fields and native JevEval with Noul, ordinal Score and credited Choice questions, plus weights, thresholds and strict mode.
  • Release gates require complete applicable pairs, a minimum pass rate and a maximum mean drop; monitoring separates quality change from ticket-mix change.
  • The repo's Python app makes fresh provider calls, streams progress and stores a local evidence ledger; Python 3.12 and your own Jev key.

Quickstart

bash
python3.12 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
cp .env.example .env
python serve.py

Watch out

Grounded in the truncated README and GitHub metadata reviewed on 2026-10-15; the README was cut off after the local-app startup section, the hosted demo makes no live model calls, the gate and drift rules are application logic rather than a monitoring service, and Python 3.12 is required.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai