jevbooks

← All projects

Jevest

tincke10/Jevest

Automated PR review pipeline using Jev as a millisecond decision layer.

Classification & Routing86%Runner-up: Evaluation & Benchmarking

What it is

Jevest is an automated PR review pipeline for developers. It uses Jev (TypeSafe AI) as a millisecond decision layer over an LLM reviewer to decide what to review, how much, and what to publish. Status: phase 0/1 skeleton.

How it uses Jev

Jev answers surface/recognition questions about code hunks, never "does this have a bug?" after H0 failed. It flags change_kind, touches_public_api, touches_error_handling, touches_async, touches_io, and filters findings via is_real_defect, severity, is_style_only, actionable. Results are used in the pipeline via DecisionPort adapters (fake, recorded, typesafe).

Primitives:choicescorenoul

Technique worth stealing

Pivot after H0 failure: use Jev only for surface/recognition questions, not defect judgment.

Try it

pnpm install && pnpm test && pnpm typecheck; then pnpm spike --mode record with TYPESAFE_API_KEY.

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-centricyes0.74
  • Shows a System One patternyes0.89
  • Handles uncertaintyno0.12
  • Measuredno0.05
  • Runnableyes0.81
  • Worth recommendingno0.27
  • Model replicano0.05
  • Problem scopescore on a 0–2 scale1.93
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.