Jev Cookbook (Chinese)
Datawhale's Chinese Jev course: notebooks on the three primitives, translated recipes, evaluation and local Laya fine-tuning.
This repository is mostly documentation or write-up rather than runnable code.
Overview
Jev Cookbook from Datawhale is a Chinese-language course on Jev and System One decision models. It contains eleven runnable Jupyter chapters, a community translation of the official TypeSafe documentation and a knowledge base of more than 500 files. Topics run from the Choice, Score and Noul primitives and confidence, through architecture patterns such as speculative fan-out and confidence gating, to 18 replicated official recipes.
Later chapters include a voice-driven 3D smart-home demo, a benchmark comparing the open Laya model with Jev on 231 public questions across accuracy, Brier score and ECE, 12 runnable application projects, agent integrations in Pi and DSH, and a full pipeline to fine-tune Laya with RLCD and serve it as a local Jev-compatible service. Offline examples let learners follow without an API key; a key is optional.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | How to phrase Choice, Score and Noul questions and act on their probabilities(inferred) | — |
How it works
- Read the introductory chapters on primitives and mechanisms
- Run notebooks offline or with a TYPESAFE_API_KEY
- Work through recipes, the smart-home lab and the Laya vs Jev benchmark
- Optionally fine-tune Laya locally and serve a Jev-compatible endpoint
Strengths
- Fills a gap in Chinese-language material on decision models
- Runs without a key using offline examples
- Covers both hosted Jev and the open Laya model
Caveats
- Documentation translation is unofficial; English docs are authoritative
- Mostly educational content
Getting started
Read online at https://datawhalechina.github.io/jev-cookbook/ or clone the repo and run `./setup_env.sh` in main/ to execute the notebooks.At a glance
- Type
- Article
- Status
- production
- Language
- Python
- Runtime
- Jupyter notebooks
- Has code
- yes
Use cases and integrations
How curated lists describe it
⚡ 适合中国宝宝的 Jev 入门教程｜手把手带你了解关于 Jev 的一切——Jupyter Notebook 轻松实验，从三种问题原语到 18…
wh000wh000/awesome-jev-liveJev 模型（TypeSafe AI）官方使用文档的中文翻译 | Unofficial Chinese translation of the official Jev (TypeSafe AI) docs — https://docs.typesafe.ai.
hellogumbo/awesome-jevDatawhale's Chinese course built on the official documentation: the three primitives, the official recipes, evaluation and local fine-tuning.
MrJev/awesome-jev
Frequently asked questions
Who is it for?
Chinese-speaking developers and students with basic Python who want to learn decision-model programming.
Do I need an API key?
No, offline examples cover every code path; a key lets the same code call the real model.
Does it cover open models?
Yes, it benchmarks and fine-tunes Laya, an open Jev-like model, and serves it locally.
Related projects
Bilingual research report on what Jev is, how its claims compare with independent evidence, and the open-source alternatives.
ProjectresearchJev-compatible open System One engine with a Claude Code-style terminal REPL showing live probabilities, plus an experiment harness.
ProjectprototypePython33-chapter Chinese tutorial on agentic RL and GRPO, including a chapter on when Jev-style external decision models beat training with RL.
ProjectresearchPythonJapanese research deck, article and source ledger tracking TypeSafe's Jev and System One models over time.
ProjectdiscussionMarkdownHead-to-head eval page where TrustedRouter's gev-1.0 outperforms Jev on the same JevBench questions and also accepts images.
ArticlediscussionHacker News thread on Decision Index 0.2.1, which compared about 70 open reproductions of Jev's decision model.
Discussiondiscussion