User repository on GitHub
最简 Jev 调用演示器：单页分类器，把请求 JSON、概率分布、confidence、耗时与 token 都摊开给你看
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev Lab — q93304989-bit
A no-dependency single-page classifier that unpacks a Jev call: paste an API key, edit a choice question, and see the request JSON, per-category probabilities, confidence, latency, token usage and the equivalent curl.
- Stars
- 1
- Language
- HTML
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- q93304989-bit
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Lab — jammaru.
- Languages
- Python
- Capabilities
- Open sourceRuns locallyRequires API key
Highlights
- Python 3.8+ standard library only: run ./run.sh or python server.py and open 127.0.0.1:8000, with no pip install at all.
- One page shows request JSON, raw Jev response, a live-updating curl command, round-trip milliseconds, token usage and the answering model version.
- Four templates (sentiment, topic, ticket priority, blank) with validation for empty questions, fewer than two categories and duplicate names.
- Explains that confidence is not accuracy: it measures distribution concentration, which is why it can gate auto-handling versus human review.
- server.py exists because api.typesafe.ai sends no CORS headers and a key in a static page is public; it forwards POST /api/systemone.
Quickstart
bash
python server.py # then open http://127.0.0.1:8000 and paste your keyWatch out
Grounded in the repository README (primarily Chinese) and GitHub metadata reviewed on 2026-10-15; the README documents a single local demo flow and reports no accuracy benchmarks, warns that Jev is English-first so CJK results differ, and says the local-proxy design must not be exposed publicly as-is.
Questions about this entry
Short answers drawn from the linked project or page.
A single-page classifier and teaching tool for Jev: you edit a Choice question, hit send, and the page lays out the full request JSON, per-category probabilities, confidence, latency and token usage for one call.
You need Python 3.8+ and nothing else; the README stresses there are no dependencies. Double-click run.cmd on Windows, run ./run.sh on macOS or Linux, or start python server.py, then open http://127.0.0.1:8000 and paste a key from console.typesafe.ai/keys.
Two reasons from the README: api.typesafe.ai returns no Access-Control-Allow-Origin header for any origin, so a browser fetch always fails, and an API key inside a static page would be public. server.py serves the page and forwards POST /api/systemone with the Authorization header.
The page keeps it in the tab's sessionStorage, so closing the tab clears it. The proxy holds it in memory per request, listens on 127.0.0.1 only, and the README says the key is not written to disk or logs. Do not deploy this design as-is for public access.
The README says confidence is not accuracy: it measures how concentrated the probability distribution is. Categories close together push it down, which the demo suggests using to decide between automatic handling and sending the case to a human.
No; it deliberately uses only choice questions to stay small. The README notes Jev also has noul (yes/no probabilities) and score (positions on ordered levels) types and links to the official primitives documentation for those.
Jev is trained mainly on English; the README states CJK input works but officially is not equivalent, and suggests tracking the confidence distribution on Chinese data rather than judging by accuracy alone.
More Jev resources like this
A workshop of small, inspectable Python experiments for TypeSafe System One models: ten critical bays, from fraud screening to emergency dispatch, each giving Jev a realistic state while deterministic policy keeps authority over the next action.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
jgalego/jevs-garagePython
jgalego/jevs-garagePythonA local proof of concept where Jev picks which of two conflicting signal phases gets the green light at a real Hong Kong intersection, using near real-time keyless open-data traffic sensors. A phase-timing state machine enforces minimum and maximum green, amber and all-red around the model's answer.
Jev Demos & Experiments#community#catalog-import#logicrw
An AI companion that replies only with facial expressions: Jev judges each message and picks one of 16 moods while blobatar morphs its face, with no text replies and no visible chat history.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
appchainai/jevatarTypeScript
appchainai/jevatarTypeScriptA one-page playground and test bench for TypeSafe Jev: pick a backend, load an example and run typed questions against Jev through OpenRouter or TypeSafe, an LLM baseline, or a keyless mock — all speaking the same /v1/systemone contract, with an inspector for the answer distributions.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
creativoma/here-we-go-jevTypeScript
creativoma/here-we-go-jevTypeScriptGenerative painting instrument where Jev picks paint materials and colour palettes for 105 procedural sketches: you select part of a sketch, and the chosen material flies in and paints itself. FastAPI + React + TypeScript.
Jev Demos & Experiments#community#catalog-import#logicrw
A minimal real-time Clash Royale-style game engine in Node.js that pits a Jev System One bot against a traditional Ollama LLM bot, with an MCP server and a canvas web client. The Jev bot answers four tactical questions in one parallel call.
Jev Demos & Experiments#community#catalog-import#logicrw
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.