jevbooks

← All projects

perch

lakeday-org/perch · Homepage

Semantic code linting with Jev.

Evaluation & Benchmarking85%Runner-up: Verification & Guardrails

What it is

Perch is a semantic code linter for developers. It scans code units, evaluates them against configurable Jev questions, and reports problems with severity, type, confidence, and method.

How it uses Jev

Perch evaluates code units against configurable Jev questions defined in perch.yaml. Each rule specifies where to look, what to check, and a minimum confidence. Jev returns probabilities that become issue confidence scores and rankings.

Primitives:choicescore

Technique worth stealing

Define custom semantic lint rules in perch.yaml with where, each, min, and ensure to evaluate code units against Jev questions.

Try it

npm install -g @lakeday/perch && export PERCH_API_KEY=... && perch scan <file>

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-centricyes0.83
  • Shows a System One patternunclear0.43
  • Handles uncertaintyno0.05
  • Measuredyes0.90
  • Runnableyes0.98
  • Worth recommendingno0.20
  • Model replicano0.08
  • Problem scopescore on a 0–2 scale1.93
  • About Jevyes0.78

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