User repository on GitHub
Turn any off-the-shelf LLM into a Jev -like decision layer
- Forks
- 10
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
LitJev
A reproduction of Jev that turns any Qwen checkpoint into a decision model serving the same /v1/systemone schema (Choice, Score, Noul), with no training and no generated answer text.
- Stars
- 47
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- zhengxuyu
- Last verified
- Sep 2026
About this project
A reproduction of Jev: turn any Qwen model into a fast decision model. LitJev reproduces the decision layer of Jev on top of off-the-shelf Hugging Face checkpoints. Define questions and options, load a Qwen model, and get typed choices with probability distributions from a single API call. No training, no generated answer text: scores are read directly from the model's output head. LitJev 是对 Jev 的复现：把 Qwen 全系列模型变成像 Jev 一样的快速决策模型。 定义问题和选项，加载模型，即可通过 API 或浏览器前端获取选择结果与概率分布， 无需训练，也无需生成回答文本。.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
Highlights
- Serves POST /v1/systemone with the same model, state, and questions body and the same choice, score, and noul question types as the hosted API.
- Scores are read directly from the model's output head, so no answer text is generated, and up to ten questions fit in one request.
- Default checkpoint is Qwen/Qwen3.8-27B on one H100 80 GB; vision checkpoints are required for screenshot decisions.
- Probabilities are not calibrated by default, and confidence values differ from the hosted Jev service.
Quickstart
bash
git clone https://github.com/zhengxuyu/litjev.git
cd litjev
uv run --locked litjev --model Qwen/Qwen3.8-27BWatch out
Apache-2.0 for original code with adapted examples under MIT. Needs uv, a local GPU, and a Qwen checkpoint; not affiliated with or endorsed by TypeSafe AI.
Questions about this entry
Short answers drawn from the linked project or page.
Same wire contract, different model and different internals. It serves the same systemone path with the same model, state and questions body and the same choice, score and noul types, so a client only changes base URL, auth header and model name. The README states that internals, confidence values and performance are not identical and that probabilities are not calibrated by default.
The default and most-tested checkpoint is a 27B Qwen on one H100 80 GB. The full Qwen family is supported at any size, and you can point at a local checkpoint path and pin a device. It is not on PyPI yet, so you run the checkout. It is a research reproduction, not affiliated with or endorsed by TypeSafe AI.
Both, but not through every backend. Vision checkpoints are required for screenshot decisions and the Qwen vision checkpoints are supported; other model families are not guaranteed to work. Under SGLang it is text states only for now, and screenshot decisions still need the transformers backend.
The two backends round differently, and SGLang has open issues that can attach a hidden state to the wrong request in a batch, which is why batched questions are off by default there. A question whose routing gain sits close to the routing threshold can go one way on transformers and the other on SGLang, so the README says compare numbers only within one backend. A verify module exists to cross-check them.
The full Qwen family (Qwen3.x text and vision, any size), defaulting to Qwen/Qwen3.8-27B, with Hugging Face transformers or SGLang as inference backends; vision checkpoints are required for screenshot decisions.
It prefills the shared state, then appends each option's own words after 'Answer:' and teacher-forces them; the summed log-probability of an option's tokens is its score and a softmax over scores is the distribution.
Only three things change: the URL (127.0.0.1:8000 versus api.typesafe.ai), the auth header (none versus Bearer API_KEY) and the model name; the JSON contract is identical, though probabilities differ.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
Open multilingual System 1 decision models with published checkpoints for choice, score and noul questions, plus a router that dispatches each request to the right checkpoint in one forward pass.
Jev Repos & SDKs#community#python#open-models
Community
NandhaKishorM/layaPython
NandhaKishorM/layaPythonA local server and Python package that runs open models with Hugging Face Transformers or PyTorch and returns Choice, Score, and Noul answers read from next-token logits instead of generated JSON.
Jev Repos & SDKs#community#python#open-models
Community
featherless-ai/simple-jevPython
featherless-ai/simple-jevPythonAn open, non-autoregressive System One model with published weights and a TypeSafe-compatible /v1/systemone API; the 395M-parameter OptionMarker checkpoint scores discrete and ordinal questions locally.
Jev Repos & SDKs#community#python#open-models
Community
wfzyx/vonPython
wfzyx/vonPythonTurns local text and vision models into Jev-style decision models by reading probabilities from prefill logits alone, with SGLang, Transformers, and Apple Silicon MLX backends.
Jev Repos & SDKs#community#python#open-models
Community
Yinsongxu/LLM2JevPython
Yinsongxu/LLM2JevPythonTogether AI's open recipe and weights for a Jev-inspired decision model fine-tuned on Qwen3.5-4B, with the full data pipeline, training config, and saved benchmark results.
Jev Repos & SDKs#community#python#open-models
Community
togethercomputer/tev1Python
togethercomputer/tev1PythonAn open 0.6B System One decision model that reads unstructured state and typed questions in one forward pass, returning calibrated distributions with zero decoded tokens; weights are published on Hugging Face.
Jev Repos & SDKs#community#python#open-models
Community
malevrigns/agent-jevPython
malevrigns/agent-jevPythonFrom 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.