How is a distilled head different from just calling Jev every time?

Once trained the head answers in tens of milliseconds with no tokens or round trip and hands unsure cases back to the provider. Multi-field objects learn one head per field and serve locally only when every field is confident.

I have no Jev key — can I still run my typesafe-sdk client?

Yes. Point the client base URL at the daemon with any placeholder key and the base checkpoint answers typed questions. Add labelled rows later via paid upstream relay or manual import to train heads.

How do I know when a site is safe to enable locally?

Use status to see captures, train per site, then report for holdout agreement and calibration at the operating point. Enable warns under 50% coverage in local mode and live fields keep comparing against provider answers.

Will it work as a local Claude Code approval hook?

That is an intended use: a PreToolUse gate answering whether a command may run in about 50 ms without sending every command to a provider, with hook examples shipped under examples/devtools.

What does the keyless local mode look like?

Run pip install 'stuntd[train,jev]', then stuntd config init, then stuntd serve, which listens on http://127.0.0.1:8787 and serves the base Laya checkpoint. Point a typesafe-sdk client at that base URL with any placeholder key and typed questions answer locally with no provider and no training.

Which APIs does the proxy speak?

The Jev System One protocol (POST /v1/systemone with choice, score and noul), the OpenAI Decisions API (POST /v1/decisions), OpenAI Chat Completions and Anthropic Messages. Typed requests are recorded and learned from; anything else is relayed byte for byte, streaming included.

How does local serving decide when to answer itself?

It answers locally only when every field's head is live and confident; otherwise the provider answers and every field is recorded. Multi-field objects learn one head per field named site.field, and the X-Stuntd header reports what served the request, for example mode=local with zeroshot=1 when the base checkpoint answered.

CommunityJev Demos & Experiments

Stuntd

Self-hosted proxy that records typed decisions, distils them into small heads on a frozen encoder, and answers locally when confident.

See highlights

Open github.com

Stars: 70
Language: Python
License: APACHE-2.0
Published by: Community
Author: bladedevoff
Last verified: Oct 2026

bladedevoff/stuntd

User repository on GitHub

View on GitHub

Local proxy that learns your app's typed LLM decisions and answers them with a Laya head. Jev and OpenAI compatible.

Forks: 5
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

stuntd is a local, self-hosted proxy that records the typed decisions your app already makes and learns to answer them itself. It speaks the Jev System One protocol (POST /v1/systemone, with choice, score and noul questions), the OpenAI Decisions API (POST /v1/decisions), the OpenAI Chat Completions API and the Anthropic Messages API, distils each decision site into a small head on a frozen Laya encoder or a stock sentence encoder, and serves that answer locally with calibrated confidence, handing anything it is…

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Speaks Jev systemone plus OpenAI and Anthropic typed-decision shapes, relaying anything else byte for byte.
Runs keyless on the base checkpoint, or learns from OpenAI, Anthropic, or paid Jev traffic with import, train, and enable.
Per-site confidence thresholds with novelty gating, fallback policies, and report scoring what is actually served.

Watch out

Early release where zero-shot quality trails Jev on some label sets and low-coverage sites fall back to the base checkpoint or provider. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

open-jev-typed-decision-engine

Open 150M encoder that reproduces the typed noul, choice, and score interface in one forward pass, trainable on a free GPU and runnable locally for nothing.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

intikhab49/open-jev-typed-decision-enginePython

TypeLLM

941

Library that adds schema-guaranteed typed outputs to existing autoregressive LLMs without changing weights, served through SGLang or a hosted API.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

TypeLLM/TypeLLMPython

jevify

Toolkit that makes a model behind an OpenAI-compatible endpoint or running in-process answer typed choice, score, and yes/no questions from next-token probabilities in Jev API shape.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

fidecastro/jevifyPython

jev-forge

164

End-to-end toolkit that synthesizes decision data, trains and calibrates candidate scorers, evaluates fixed splits, and serves Jev-compatible inference for interactive web decisions.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

zwliJay/jev-forgePython

open-jev

Open harness that renders allowed JSON answers on a diffusion canvas, denoises for a fixed step budget, then picks the top allowed tokens from final logits with no training.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

JoshuaSP/open-jevPython

jev-studio

Python kit pairing a Jev CLI for verify, screen, classify, extract, match, route, ask, find, rerank, compact, and batch with an MCP server exposing the Jev ruleset over stdio.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

utk2103/jev-studioPython

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai