Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

jevlint

Linter for Jev queries that flags question and state wording the model is documented to handle badly, before you pay.

ProjectbetaMIT
View on GitHub ↗Live demo ↗

Overview

This jevlint lints the request you would send to Jev itself: a state and the questions asked about it. It reports where the query is written in a way the model is documented to handle poorly, naming the defect, its location and what to change. A worked-examples page shows three flawed queries, their reports and corrected versions.

Static checks run with no key and no cost. Optional probe checks ask Jev about the query and cost two calls per question plus two for the query. Findings carry a severity and, separately, a probability, and the docs explain why those say different things, how to accept a finding, which Jev build the rules target and what the tool cannot tell you. It is implemented in PHP with packages for Composer, npm and PyPI and library APIs for PHP, JavaScript and Python. It is for developers writing Jev questions who want a review step before deployment.

What the decision model decides

TypeQuestionAnswers
Yes / noIs this question written in a way Jev is documented to handle badly? (probe checks)(inferred)—
OtherStatic rule checks over the state and question definitions—

How it works

  1. Point jevlint check at a query JSON file
  2. Static rules run offline and report errors, warnings and advisories
  3. Optionally run probe rules that ask Jev about the query
  4. Fix the query using the suggested changes; exit codes gate CI

Strengths

  • Static mode is free and offline
  • Available for PHP, Node and Python
  • Documentation on evidence for each check and limits of the tool

Caveats

  • Probe mode adds API calls and cost
  • Rules are tuned to particular Jev builds
  • Cannot guarantee a passing query gives good answers

Getting started

composer install && ./jevlint check examples/broken-triage.json --static-only (or pip install jevlint / npm install --save-dev @phox-js/jevlint).

At a glance

Type
Project
Status
beta
Language
PHP
Runtime
CLI and library (PHP, Node, Python)
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does it lint?

Jev request bodies: the state and question definitions, not your application code.

Does it cost anything?

Static checks are free; probe checks cost two Jev calls per question plus two for the query.

Which languages can call it?

PHP, JavaScript and Python, via Composer, npm or pip.

Related projects

  • system-one-playground

    SysOneScript language, Go System One client, semlint semantic code checks and a VS Code Studio for building with Jev judgments.

    LibrarybetaGo
  • Agent

    Native macOS agent app wiring 23 LLM providers into a task loop, listed with an optional Jev command-risk advisor via a TypeSafeKit client.

    LibraryproductionSwift
  • JevGate

    Rust code-review gate asking Jev small typed questions about functions and tests, failing CI only on rules measured 80% right.

    CLIproductionRust
  • Abide

    Coding-agent hook that asks Jev one yes/no per project rule on every edit or turn and makes the agent fix any violation.

    HookbetaTypeScript
  • code-quality

    Advisor for coding agents: git and Stop hooks flag test tampering, secrets and skipped hooks, with optional Jev review findings.

    Hookbeta
  • commentcop

    CLI that uses Jev to score JS/TS code comments for accuracy and usefulness and lists the weakest first.

    ProjectbetaTypeScript