Are the latency and cost gaps real or tuned prompts?

The digest reports one round trip staying in the hundreds of milliseconds even as questions grow, while the same batch to the comparison model took seconds and far more cost. Rerun the 1, 5, 15, and 30 question timing page on your own connection before trusting the gap.

Why does rewording the same intent change the scores?

Yes, and that is the point. Jev judges literally, so the digest shows one phrasing scoring low and a rewritten intent scoring high, plus Chinese cases where punctuation moves the calm probability. You edit questions on the page until wording matches intent.

What is the minimal setup before the live probability bars move?

Copy .env.example to .env with an OpenRouter key shared by both models, uv sync, then uvicorn on port 8010 for the page or main.py for the CLI. The key stays in app.py on the server and never reaches the browser.

How does Jev travel over OpenRouter without a models-list entry?

Jev uses a separate POST systemone endpoint absent from the models list, with IDs like jev-latest accepted. The official SDK only needs its base URL pointed at OpenRouter, and an optional model override selects the Jev and comparison models.

How do I run the jevtest demo?

It needs uv: copy .env.example to .env and fill OPENROUTER_API_KEY, run uv sync, then uv run uvicorn app:app --port 8010 for the web page or uv run python main.py for the command line; the key stays on the server.

Which models does jevtest compare?

Jev, defaulting to jev-latest and overridable with JEV_MODEL, against a comparison LLM that defaults to google/gemini-3.5-flash-lite and is set by OPENROUTER_MODEL, both through one OpenRouter key.

What kinds of questions does the jevtest demo ask?

Three scenarios, a partner, a boss and a friend, ask a Choice for the real meaning, a Score for anger on four levels, and a Noul for whether you should care right now, and the questions can be edited on the page.

CommunityJev Agents & MCP

Jevtest — joshhu

Chinese-language demo contrasting Jev probabilities with a generative model on the same relationship message. Type a sentence, watch choice, score, and noul distributions move, then compare latency, cost, and output shape.

See highlights

Open github.com

Stars: 7
Language: HTML
Published by: Community
Author: joshhu
Last verified: Oct 2026

joshhu/jevtest

User repository on GitHub

View on GitHub

情緒測謊器：嘴上說「好」，心裡真的好嗎？用 TypeSafe Jev（System One 模型）透過 OpenRouter 即時判斷，並與一般 LLM 對照

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity awesomejev catalog-import html

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Highlights

Choice, score, and noul questions share one request with probabilities and confidence.
Live typing view plus per-character trajectory and adjustable confidence thresholds.
Same prompt sent to Jev and a generative model for time, cost, and format contrast.
Three editable scenarios with server-side key handling and CLI plus web entry points.

Watch out

Demo wording is literal and fragile; paraphrases can shift scores, and the key must stay server-side. 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.

Other listings of this project

Forks and copies of this project in the directory. This entry is the canonical one.

jevtest

Semantic test matchers for Vitest and Jest where plain-English expectations become single Noul judgments, with probabilities mapped to pass, fail, or ambiguous.

Jev Practices & Patterns#community#awesomejev#catalog-import

Community

realZachi/jevtestTypeScript

JevTest

Local TypeScript library and CLI for exploratory browser testing: Jev picks actions and judges observed results while Playwright executes them in isolated sessions, with trace files, reports and replay that needs no model calls.

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

Community

CorieW/JevTestTypeScript

More Jev resources like this

jev-code

TypeScript SDK and CLI coding agent where Jev only picks options while your code runs effects, building programs as syntax trees.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

rhighs/jev-codeTypeScript

muse-jev-playbook

Agent-oriented playbook for using Jev as a cheap, fast decision layer that triages, classifies, scores, and gates work before expensive browser, research, retry, or subagent steps.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

Bodila51/muse-jev-playbookPython

pi-jev-router

Pi extension that classifies each task with Jev over OpenRouter and routes it to a Pareto-optimal model balancing measured quality, cost, and latency under a role policy.

Jev Agents & MCP#community#jevbest#catalog-import

Community

philippdubach/pi-jev-routerTypeScript

jevyoumean

Shell wrapper that intercepts unknown subcommands, reads the target CLI help output, and asks Jev which documented subcommand matches the intent behind the typo.

Jev Agents & MCP#community#jevbest#catalog-import

Community

syumai/jevyoumeanGo

claude-jev

Coding-agent plugin that routes prompt classification, subagent tiering, edit rule checks, and fast Jev-backed compaction through small typed judgments.

Jev Agents & MCP#community#jevbest#catalog-import

Community

0x7067/claude-jevTypeScript

jevcache

Local OpenAI-compatible proxy that skips repeat model calls on same-intent paraphrases and retried agent turns, judged by Jev or a local System One model.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

kushals256/jevcacheTypeScript

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai