Jev directory
CommunityJev Repos & SDKs
Open Jev — npmjs.com
Browser TypeScript library that runs open Jev-shaped System One models through Transformers.js. Takes one state plus typed questions and returns a calibrated probability distribution per question, on-device via WebGPU or WebAssembly.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
This is one of several Jev listings for the same project. The canonical entry is Open Jev — nico-martin.
Highlights
- Ships three model aliases: kev-0.6b, kev-4b, and open-jev (DeBERTa-v3-large).
- Runs on-device with WebGPU when available and WebAssembly otherwise.
- Question builders are choice, score, and noul, with option literals flowing into typed answers.
- open-jev allows 2 to 255 choice options and 2 to 10 score levels; kev models allow 2 to 255 levels.
- Install with npm install open-jev @huggingface/transformers; Transformers.js is a peer dependency.
Questions about this entry
Short answers drawn from the linked project or page.
Run npm install open-jev @huggingface/transformers. Transformers.js is declared as a peer dependency so your app and the library share one copy of it and one model cache.
No. Everything happens on-device, using WebGPU when available and WebAssembly otherwise. Model weights are downloaded from Hugging Face once and then read from the local cache.
The aliases kev-0.6b (default, Qwen3-0.6B-Base), kev-4b (Qwen3-4B-Base), and open-jev (DeBERTa-v3-large), or any Hugging Face repo id whose config.json carries an open_jev or kev section.
choice picks one option and accepts optional per-option descriptions, score rates on an ordered scale of levels, and noul asks whether a statement holds for the state. Plain objects with type, instructions, and options also work.
open-jev allows 2 to 255 choice options and 2 to 10 score levels with a 512-token context. The kev models allow 1 to 255 options and 2 to 255 levels with an 8192-token context.
No. One forward pass returns a calibrated probability distribution per question, so an answer is always one of the options you supplied rather than free-form generated text.
kev-0.6b is 0.34 GB at q4f16 and runs in the browser. kev-4b is 2.3 GB and needs a capable GPU. open-jev also ships fp16 (0.88 GB) and fp32 (1.75 GB) variants.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Text-to-screenshot (Malay Vasa)
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
X: Google's MediaPipe Decision Maker, tested in the browser
More Jev resources like this
A Node and TypeScript package that runs Convai's Laya checkpoint through ONNX Runtime, with the same system_one shape as the Python reference.
Jev Repos & SDKs#community#typescript#open-models
Community
receptron/layaTypeScript
receptron/layaTypeScriptZero-dependency TypeScript client for TypeSafe AI's Jev, providing typed questions in and typed answers out over the TypeSafe API or Cloudflare Workers AI. Shared client behind the jev-tools packages; unofficial and not affiliated with TypeSafe AI.
Jev Repos & SDKs#typescript#sdk#github
Community
A TypeScript package for one-line local AI decisions: calibrated choice, noul and score answers from open models such as kev-0.6b and kev-4b, with no API keys and no cloud, and models auto-downloaded on first use.
Jev Repos & SDKs#community#github#typescript
Community
amatelic/jevTypeScript
amatelic/jevTypeScriptThe AI Toolkit for TypeScript: Vercel's free, open-source AI SDK for building AI-powered applications and agents, with GitHub topics covering OpenAI, Anthropic, Gemini, React, Svelte, Vue, and Next.js.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
vercel/aiTypeScript
vercel/aiTypeScriptA filesystem-first framework for durable AI agents whose auto model router defaults to Jev through Vercel AI Gateway and whose evaluate helper asks typed Choice, Score, and Boolean questions inside tools.
Jev Repos & SDKs#community#typescript#agents
Communityvercel/eveTypeScript
vercel/eveTypeScriptA 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-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.