Skip to content

jev-cli

Node/Bun CLI for asking Jev boolean, choice and score questions over text or JSON, with direct or Vercel AI Gateway providers.

CLIbetaMIT

Overview

jev-cli by jtsang4 is a command-line interface for Jev with English and Chinese documentation. You give it a state and typed questions and get structured JSON back. The README example evaluates a support exchange, asking whether a refund was issued (boolean probability 0.99), which tone the agent used (warm at 0.98 versus curt), and how well it was handled on a poor/acceptable/excellent scale (score 1.78). It positions Jev as a good fit for classification, routing, rubric scoring and automated verification and a poor fit for writing.

It installs globally from npm or Bun as @jtsang/jev-cli and runs on Node 22+ or Bun. Configuration is set up with config init and config set, with the default provider calling TypeSafe directly and an alternative routing through the Vercel AI Gateway. A doctor command sends one tiny evaluation to confirm the key and model work. It is MIT licensed.

What the decision model decides

TypeQuestionAnswers
Yes / noWas a refund issued?yes · no
ChoiceWhat tone did the agent use?warm · curt
ScoreRate how well this was handled.poor · acceptable · excellent

How it works

  1. Install and run config init
  2. Set the provider key; run doctor
  3. jev-cli eval with state and questions
  4. Consume JSON output

Strengths

  • Simple setup with health check
  • Two provider options
  • Clear JSON output

Caveats

  • Requires a TypeSafe or Vercel AI Gateway key

Getting started

`npm install -g @jtsang/jev-cli`, then `jev-cli config init` and `jev-cli doctor`.

At a glance

Type
CLI
Status
beta
Language
TypeScript
Runtime
Node 22+ or Bun CLI
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does jev-cli return?

JSON with a probability, choice or score for each question.

Can I use Vercel AI Gateway?

Yes, set provider to vercel.

How do I check my setup?

Run jev-cli doctor.

Related projects

  • Rust terminal coding agent for local models whose swarm mode can use Jev to referee parallel attempts and pick a typed winner.

    LibrarybetaRust
  • Declarative compiler that turns state, decisions and labeled examples into a measured, frozen JSON program of typed Jev questions.

    LibraryprototypePython
  • Vercel Labs terminal CLI for AI tasks that can use Jev as the evaluation model for its evaluate command.

    CLITypeScript
  • Tool that recomputes Jev's confidence identities on outputs published in public repositories to spot answers Jev never actually produced.

    ServiceresearchPython
  • TypeSafe's official drop-in replacement for the SDK's system_one call backed by OpenAI, Anthropic or Gemini, for comparing Jev with LLMs.

    LibraryproductionPython
  • Hack Club's lightweight LLM proxy, listed for adding a Jev forwarding endpoint behind its auth, limits and usage logging.

    MCP serverproductionTypeScript