User repository on GitHub
Ten levels of Jev, from one smart if statement to a coding agent that reaches for Jev on its own
- Forks
- 62
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Cookbooks
10 Levels of Jev
A hands-on lab that arranges thirty real uses of Jev into ten levels: levels 1–5 call Jev from plain code, and levels 6–10 run inside the pi coding agent, ending with an agent that writes its own Jev questions.
- Stars
- 237
- Language
- TypeScript
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- disler
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
- Platforms
- CLIOpenRouter
- Capabilities
- Open sourceRequires API key
Highlights
- Ten levels from one smart if statement to a coding agent that reaches for Jev on its own; thirty concrete uses are arranged into them.
- Every level runs live against real Jev and every option is tested; 201 offline tests run against a deterministic mock.
- README framing: send the System One model a state and typed questions; typed answers with probabilities arrive about 300 ms later for a fraction of a cent.
- Prereqs: Node 24, bun, just, pi 0.85.1+ for levels 6–10, and an OpenRouter key for live runs; levels 1–5 run on the mock without one.
- just install runs the /install command in Claude Code or Pi; just test runs the offline suite and just web opens the lab at 127.0.0.1:4399.
Quickstart
bash
cd apps/ten-levels/web && bun install && cd ../../..
cp .env.sample .env
npm install -g @earendil-works/pi-coding-agent
just test
just webWatch out
README reviewed 2026-10-17 and truncated mid-install, so the later prerequisites and level walkthroughs are not in the digest. The ~300 ms latency and cost-per-call framing are the README's own, not independent benchmarks, and the digest metadata shows last-push equal to creation date (2026-09-27).
Questions about this entry
Short answers drawn from the linked project or page.
A live lab arranging thirty concrete uses of Jev into ten levels for agentic engineers: levels 1–5 call Jev from plain code, and levels 6–10 place Jev inside the pi coding agent, up to agents that write the questions themselves.
You send TypeSafe's System One model a state and typed questions, and roughly 300 ms later receive typed answers with probabilities your code can branch on. Every level in the lab runs live against real Jev.
Agentic install is just install, which runs an /install command in Claude Code or Pi. Manual install needs Node 24, bun, just, pi 0.85.1+, and an OpenRouter key, then bun install, cp .env.sample .env, and npm install -g the pi-coding-agent.
For live runs, yes. Without one, levels 1–5 still run on the offline deterministic mock. The /install command checks the key without printing it and verifies the sandbox and offline suite.
Jev inside the pi coding agent: first as hooks the agent never sees, then as tools the agent chooses to call, until the agent writes the questions itself.
201 offline tests on the deterministic mock, per the README. just web then builds the lab and opens it in a browser on http://127.0.0.1:4399.
disler's MIT-licensed TypeScript repo, with a YouTube walkthrough linked. Digest metadata: 234 stars, 61 forks, created and last pushed 2026-09-27.
More Jev resources like this
Six small Python files that build one coding agent step by step on DSPy's RLM sandbox, with Jev answering closed questions — command safety, tool choice, history relevance, delegation — and plain Python owning the thresholds.
Jev Cookbooks#community#github#python
A coding agent for your terminal: you chat, and when you ask for a change it reads, edits and runs your code itself. Jev helps on the side with a few quick calls that spot a simple question so replies come back faster, and its key is optional.
Jev Agents & MCP#community#catalog-import#logicrw
Community
coasty-ai/jevcodeTypeScript
coasty-ai/jevcodeTypeScriptKorean-language Fastcampus seminar repo that teaches attaching Jev to a DeepAgents (LangGraph) shopping-mall support agent across 13 notebooks and a React sample app: tool selection, guardrails, risk gates and grounded RAG.
Jev Cookbooks#community#github#korean
A Claude Code plugin from cth9191 that gives compaction to Jev: it keeps tool output that is still needed word for word and moves the rest to a drawer file, taking 0.43 to 0.74 s instead of about 35 s on one 591,489-token session. A fork of fast-jev-compaction.
Jev Tools & CLIs#community#github#claude-code
An MCP server that gives coding agents only the JS/TS code a task needs: the TypeScript compiler follows calls, types, callers and tests, Jev judges relevance, and results come back within a token budget.
Jev Agents & MCP#community#catalog-import#logicrw
Community
123wwwa/jevtraceTypeScript
123wwwa/jevtraceTypeScriptA TypeScript/Bun launcher that routes coding-agent user turns through JEV, the TypeSafe classifier, with model and reasoning-effort selection: codex-jev wraps the Codex CLI, claude-jev wraps Claude Code, and a Pi extension adds a jev/auto virtual model.
Jev Tools & CLIs#community#github#cli
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.