Home /Benchmarks, evals and calibration /jev-harness

jev-harness

Confidence gates, shadow mode, recipes, and evals; reports Claude CLI at 48.9 s against Jev at 1.3 s on the same row-filter job.

From the readme

jev-harness A small TypeScript library that turns TypeSafe Jev answers into actions you can ship. Jev returns structured decisions (choice / score / noul) with confidence scores. This harness adds the missing pieces around that API: a policy, a confidence gate, shadow mode, reusable recipes, and an offline eval CLI. It is not affiliated with TypeSafe. You …

Details

Section
Benchmarks, evals and calibration
owner
AntonioCoppe
stars
5
forks
0
pushed
2026-09-18
Language
TypeScript
License
MIT

What it is

Form
Library or SDK
Host agent
Standalone
Audience
Agent builders
Maturity
Docs
reports measured numbers

Useful for

Best intent matches