User repository on GitHub
Open, Jev-compatible System One decision server on DiffusionGemma
- Forks
- 64
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Repos & SDKs
OpenJev — razorback16
An Apache-licensed System One decision server that reads Choice, Score, and Noul probabilities from DiffusionGemma 26B-A4B through vLLM on NVIDIA or MLX on Apple silicon, so TypeSafe SDKs work unchanged.
- Stars
- 721
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- razorback16
- Last verified
- Sep 2026
About this project
Fast, calibrated, typed decisions from an open model. OpenJev is an open-source "System One" decision server. Send it a state and typed questions (yes/no, choice, score). It returns a probability and a confidence for each answer in tens of milliseconds. It reads the answers directly from the model's probabilities and parses no text, so an answer cannot go off-schema. Questions can also ask about images.
Quoted from the project's README. Read the full README.
Works forJev for Scoring
Used inJev for Software & DevTools
Highlights
- Speaks the same wire API as Jev, so TypeSafe's SDKs work unchanged, and both backends also serve text generation on /v1/chat/completions.
- It reads answers straight off the model's probabilities with no text generation or parsing, and questions can also ask about images.
- A choice can have at most 128 options versus Jev's 255; a one-option choice or single-level score is answered directly and bills no tokens.
- confidence is 1 - H(p)/ln K, and usage.output_tokens stays 0 unless the optional think setting is used.
- Hosted free on Codiv with 100M input tokens on signup; question types support noul, choice, and 2-10 score levels.
Quickstart
python
from typesafe_sdk import TypeSafeClient
client = TypeSafeClient()
r = client.system_one(
"Everything is down and we have a demo at noon.",
{"urgent": {"type": "noul", "instructions": "Need a reply within the hour?"}},
)
print(r.nouls["urgent"].noul)Watch out
Apache-2.0; not affiliated with TypeSafe. The NVIDIA path pins an unmerged vLLM branch with extra fixes, choices cap at 128 options, and answer quality is that of DiffusionGemma 26B-A4B, so evaluate it on your own tasks.
Questions about this entry
Short answers drawn from the linked project or page.
Yes, it speaks the same wire API and also accepts jev-latest and jev-preview so SDK defaults resolve. The catch is version pinning: model ids are OpenJev's own, and a pinned version such as jev-1.13.0 returns 400 Unknown model. OpenJev extensions like images, samples and think are additions your SDKs never send.
It runs DiffusionGemma 26B-A4B through vLLM on an NVIDIA GPU or MLX on Apple silicon. The vLLM part is a specific pull request, vllm-project/vllm#57250, which added seeded canvases, read-only steps and pinned canvas positions; openjev/engine.py adapts that PR's own example server rather than carrying a private fork.
It is derived, not self-reported: confidence is 1 minus the normalised entropy of the distribution, so 1 when certain and 0 when uniform, taken from the model's own probabilities on the answer slots. A slot above 0.1 entropy triggers three more reads that are averaged. Nothing in the docs reports accuracy benchmarks, so judge answer quality on your own tasks.
Not necessarily. Besides your own vLLM or MLX deployment, the project is hosted on Codiv with 100M input tokens on signup and no card. The same endpoint works either way, including /v1/chat/completions for text generation with the diffusiongemma model id.
openjev-latest, its alias openjev-0.1, DiffusionGemma for text generation, and the small encoder models laya-1.0, verdict-1.4, clm-v0.1 and jevk5-0.2 when they run. All weights are Apache-2.0, and the server also accepts jev-latest and jev-preview.
images for up to 8 images, steps 1 to 8 for denoise passes, samples 1 to 32 to average reads, think 0 to 4096 tokens for a written thought, and sequential for long question lists. A request without them behaves exactly like Jev.
Each response carries a Server-Timing header splitting model, server and total time, where model is time on the model summed over reads and server covers schema compile, tokenization, validation and serialization, excluding your network.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev 1.14 is now on OpenRouter
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 NewsOther listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
Express and Vite playground for System One-style decisions, taking a state plus typed questions and returning structured answers. Three swappable backends: per-option LLM micro-scorers, one structured JSON call, or a locally served decider model.
Jev Models & Reimplementations#community#jevbest
Community
SiliconLabAI/OpenJevTypeScript
SiliconLabAI/OpenJevTypeScriptOne-pass option scoring with a local Gemma 3 4B on Apple silicon via MLX: it prefills the context once, expands that KV cache across the option batch, and scores every option in a single padded forward pass with no decoding.
Jev Practices & Patterns#community#python#open-models
Community
daseinlabs/open-jevPython
daseinlabs/open-jevPythonAn open-source, local alternative to Jev: System-One decision models that answer the POST /v1/systemone shape — choice, yes/no, or score — for general decisions and browser-agent steps. Models ship as OpenJev-1.7B, 4B, and 8B, with a pip-installable Python package.
Jev Models & Reimplementations#community#github#python
Community
alanhuangyoo/openjevPython
alanhuangyoo/openjevPythonPython decision client exposing Choice, Noul, and Score primitives with isolated order-invariant scoring, a deliberative flywheel that crystallizes hard cases into fast paths, and local or hybrid deployment.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
zhangcy122/OpenJevPython
zhangcy122/OpenJevPythonOpen reimplementation of the Jev decision-model idea: masked-logit softmax turns any small local LLM or OpenAI-compatible API into a type-safe decision engine with raw probabilities. Also ships a WeChat/QQ/Telegram crush radar with a verdict and next-move advice.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
lookski/openjevPython
lookski/openjevPythonIndependent Jev-inspired decision API with choice, score, and noul primitives, shipping a local mock server plus Python and JavaScript SDKs for interface prototyping.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
xingwudao/OpenJevPython
xingwudao/OpenJevPythonOpen-source Jev-compatible decision engine with a Claude Code style REPL, answering Choice, Score, and Noul questions in one forward pass over open models.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
GPT-AGI/OpenJevPython
GPT-AGI/OpenJevPythonOpen local classifier that reproduces a hosted Jev job-posting decision with a frozen CPU encoder plus small per-question heads, answering the same posting questions offline at zero marginal cost.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
DECRUX9812/openjevPython
DECRUX9812/openjevPythonA public Jev API funded by creator fees from a token, with a playground, docs, use cases, and a smart treasury dashboard.
Jev Tools & CLIsPlayground#community#playground#api
Community
More Jev resources like this
A 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-jevPythonA server that implements the TypeSafe/Jev HTTP API with Qwen3.6-35B-A3B on SGLang 0.5.19, deployed on Modal with radix caching and breakable prefill CUDA graphs, plus a separate FastAPI process.
Jev Repos & SDKs#community#python#open-models
Community
ekzhang/openjev-sglangPython
ekzhang/openjev-sglangPythonOpen 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/layaPythonTurns 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/LLM2JevPythonAn 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-jevPythonA DSPy fork whose @typesafeify decorator routes decision-shaped signature fields to Jev, so bool, Literal, and scored outputs come back typed while freeform fields still use the LM.
Jev Repos & SDKs#community#python#sdk
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.