jev-handson
Korean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.
Overview
jev-handson contains the hands-on code for a three-part Korean live class on Jev. Lesson 1 covers concepts and calling Jev through OpenRouter with noul, choice and score, plus a comparison site (RouteLab) that pits a general LLM against Jev on 100 toxic comments, with extras for direct TypeSafe calls, a 471-item LLM comparison, skills, hooks and MCP. Lesson 2 builds a message router with 40-item scoring, service and cost comparisons, plus agent-specific routers for OpenClaw, Codex and Claude Code. Lesson 3 covers Jev-like projects: jev-ultrafast scripts, a local decider server, API versus local scoring, and use cases in computer use, games and robots, with a Colab notebook dissecting openjev.
A bonus lesson uses OneJev, an open decision model that accepts video, to judge 13 CAVIAR CCTV clips for falls, fights and abandoned objects. Exercises are done by pasting prompts to an agent, and all figures were measured on 23 September 2026. It needs Node 20+, Python 3.10+ and a TypeSafe or OpenRouter key.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this comment toxic? | — |
| Choice | Where should this message be routed? | — |
| Yes / no | Does this CCTV clip show a fall, fight or abandoned object? | — |
How it works
- Follow each lesson's prompt file in an agent
- Call Jev via OpenRouter or TypeSafe
- Build and score routers and hooks
- Compare with LLMs and open decision models
Strengths
- Structured curriculum with measured results
- Covers hosted Jev and open alternatives
- Agent-driven exercises
Caveats
- Korean-language material
- Some parts need extra tools (uv, Chrome, Ollama)
Getting started
Copy each folder's .env.example to .env with a TypeSafe key and follow the lesson's 프롬프트.md in Claude Code, Codex or OpenClaw.At a glance
- Type
- Project
- Status
- production
- Language
- JavaScript
- Runtime
- Node and Python scripts, Colab notebooks
- Has code
- yes
Use cases and integrations
How curated lists describe it
Jev 특강 실습 코드(1~3강): TypeSafe Jev를 Claude Code·Codex·OpenClaw에 붙이는 스킬·훅·라우터·활용 예제
logicrw/awesome-jev-projectsJev 특강 실습 코드(1~3강): TypeSafe Jev를 Claude Code·Codex·OpenClaw에 붙이는 스킬·훅·라우터·활용 예제.
hellogumbo/awesome-jev
Frequently asked questions
What does it teach?
How to attach Jev to coding agents as skills, hooks and routers, and how it compares with LLMs and open models.
What language is it in?
Korean.
Do I need a TypeSafe key?
A TypeSafe or OpenRouter key; lesson 1 uses OpenRouter.
Related projects
Companion code for the ZazenCodes course on decision models: primitives notebook, use-case examples and a Pi model-router build.
ProjectproductionPythonFree 12-notebook course building 13 agent tools with Jev as the fast decider and an LLM as the slow writer, for under $0.20.
ProjectproductionPythonBeginner manual, video script and three small examples for building an agent harness with Jev via langchain-typesafe.
ProjectprototypePythonGuide and small prototype for using Jev with Herdr to route tasks to coding agents like Claude, Codex and Hermes under code-owned policy.
ProjectprototypeChinese translation of 'Jev Engineering for coding agents', design notes on a harness where Jev makes every per-turn decision.
ProjectdiscussionVercel's decision guide on when to use a System One model like Jev versus a frontier model like GPT-6 Astra.
Articlediscussion