jev-lab
Zero-dependency single-page Jev classifier demo showing request JSON, probabilities, confidence, latency and token usage.
Overview
Jev Lab (q93304989-bit) is a minimal single-page classifier demo, documented in Chinese, that lays out how a Jev call works. You open the page, paste an API key and run choice questions on a piece of text, then see the full request JSON, the probability for each category, confidence, round-trip time, token usage and the model version that answered, plus an equivalent curl command that updates as you edit.
It has templates for sentiment, topic and ticket priority plus a blank one, lets you add and remove questions and categories, and validates before sending. It needs only Python 3.8+ standard library; a local proxy avoids browser CORS restrictions, and the key is kept only in the tab's sessionStorage. The README explains common pitfalls: categories are entirely yours, question keys do not affect inference, confidence is not accuracy, one Choice picks one category, and extra questions add little latency but do add tokens.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What overall sentiment does this review express? | positive · neutral · negative |
How it works
- Start server.py and open the page
- Paste a key and pick or edit a template
- Send the request through the local proxy
- Inspect probabilities, confidence, latency and tokens
Strengths
- No dependencies
- Shows the raw request and curl
- Clear explanation of pitfalls
Caveats
- Choice only in the demo
- Chinese documentation
Getting started
`python server.py` (or run.sh / run.cmd) and open http://127.0.0.1:8000.At a glance
- Type
- App
- Status
- prototype
- Language
- Python
- Runtime
- Standard-library local server with web page
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
This repository provides a single-page classifier demo that sends text with choice questions to Jev and displays the request JSON, probability distribution, confidence, latency, and token usage.
logicrw/awesome-jev-projects
Frequently asked questions
Why is there a local proxy?
Browsers block direct calls due to CORS, among other reasons.
Is confidence the same as accuracy?
No, it measures how concentrated the distribution is.
Where is my key stored?
Only in the tab's sessionStorage.
Related projects
Colab notebook experiment that analyzes a movie review with one Noul, one Choice and one Score question to Jev.
ProjectprototypePythonLocal Python toolkit and web dashboard for trying Jev via OpenRouter or TypeSafe, with review analysis, topic routing and request history.
ServiceprototypePython- Building with Jev★ 155
Agent skill for writing and improving programs that call Jev: question design, state structure, thresholds and diagnosing wrong answers.
Agent skillproduction Server behind api.typesafe.pro, a free anonymous System One-compatible endpoint that forwards to TypeSafe on the operator's key.
ServiceproductionPortable Agent Skill teaching coding agents to use Jev as a bounded judgment layer, plus an optional MCP bridge.
MCP serverbetaTypeScriptIndependent agent skill that makes a coding agent design, build, test and tune a bounded Jev decision layer with a contract first.
Agent skillbeta