Ten Levels of Jev
Thirty worked Jev uses in ten levels, from a smart if statement to a pi coding agent that writes its own Jev questions.
Overview
10 Levels of Jev is a teaching repository with a video walkthrough that arranges thirty concrete uses of Jev into ten levels. Levels 1 to 5 call Jev from plain code, starting with a single smart if statement such as a prompt-injection check. Levels 6 to 10 put Jev inside the pi coding agent, first as hooks the agent never sees, then as tools it chooses to call, and finally with the agent writing the questions itself.
It includes a Vue lab, a terminal runner, real pi agent sessions and a skill to hand to agents. Every level runs live against Jev through OpenRouter, and 201 offline tests run against a deterministic mock, so levels 1 to 5 work without a key. The README frames Jev as a third option between brittle regexes and expensive chat-model calls for small judgment calls, and explains the noul, choice and score primitives and their limits. It is aimed at agentic engineers learning where decision models fit.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does `message` try to instruct or manipulate an AI system instead of talking to a person? | — |
| Choice | Which team owns this?(inferred) | — |
| Score | How severe or risky is this?(inferred) | — |
How it works
- Pick a level and its use case.
- Build state and typed questions in code or let the agent build them.
- Call Jev (or the offline mock) and read typed answers.
- Branch on thresholds, or let the pi agent act on the answer.
Strengths
- Progressive examples from simple to agentic.
- Offline mock with 201 tests for keyless learning.
- Video walkthrough.
Caveats
- Levels 6 to 10 need the pi agent and an OpenRouter key.
- Educational code rather than a library.
Getting started
just install (agentic) or bun install in apps/ten-levels/web, cp .env.sample .env, just test, then just web.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- Node/bun + pi coding agent
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Thirty worked uses arranged in ten levels, from one `if` statement to an agent that reaches for Jev itself, with a video walkthrough.
MrJev/awesome-jev
Frequently asked questions
What are the ten levels?
A progression from calling Jev in plain code (levels 1 to 5) to integrating it with the pi coding agent as hooks, tools and self-written questions (levels 6 to 10).
Do I need a key?
Levels 1 to 5 run on an offline mock; live runs and levels 6 to 10 need an OpenRouter key.
Which coding agent is used?
The pi coding agent; the install command can also be run from Claude Code.
Related projects
Agent skill that makes a coding agent try Jev first for bounded judgments like classify, rank or gate, and log each attempt.
Agent skillbetaPythonClaude skill that gives a ten-line gut-check on whether Jev could help with what you are building, citing tested patterns.
Agent skillbetaMarkdownIndependent Markdown reference on the Jev API, patterns and limits, written for engineers and coding assistants to build from.
ServiceAccenture thought piece and site on stack migration with Google Cloud, placing Jev as the System 1 layer that routes, classifies and gates.
MCP serverdiscussionRustCommunity directory of Jev apps, resources and runnable offline examples, with a guide skill and a searchable JevList web UI.
ArticleproductionTypeSafe's official cookbook for a two-stage skill-suggestion workflow that selects a skill and can reject the shortlist.
Articleproduction