Jev-Quantum
Jev-protocol-compatible random baseline server in Rust: answers noul, choice and score with fast pseudo-random numbers, as a mock or floor.
Overview
Jev Quantum is a tongue-in-cheek but practical random baseline that speaks TypeSafe's System One protocol (noul, choice, score) without reading prompts; its answers come from a fast pseudo-random number generator. The README (Chinese, with an English version) describes it as a 'System -1' model whose accuracy follows a Gaussian distribution. It is meant as a dummy, a mock for tests, or a lower bound when evaluating agent routing.
It runs as a Rust server exposing /v1/systemone, with a demo page that visualises imported benchmark summaries without calling any API. A bench tool compares latency and a maze-navigation task against hosted Jev through a gateway key, and the docs explain how to read in-process, server-handler and client-observed latency numbers. A buffered mode uses worker-local random pools refilled in the background. It is explicitly not a semantic model and its randomness is not cryptographically secure.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Random answers to any noul, choice or score question (baseline) | — |
How it works
- Run the Rust server locally
- Point a Jev client at /v1/systemone
- Receive random typed answers instantly
- Compare against hosted Jev with the bench tool
Strengths
- Useful lower bound for routing evaluations
- Drop-in mock for tests
- Extremely fast
Caveats
- Not a semantic model; answers are random
- Randomness not cryptographically secure
Getting started
cargo run --release -p jev-quantum-server, then POST to http://127.0.0.1:3000/v1/systemoneAt a glance
- Type
- Project
- Status
- prototype
- Language
- Rust
- Runtime
- HTTP server and bench CLI
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
亚微秒级 System-1 模型，准确率服从高斯分布.
hellogumbo/awesome-jev
Frequently asked questions
Is Jev Quantum a real decision model?
No. It returns random answers in Jev's format, as a baseline or mock.
Why use it?
As a lower bound for evaluating routing or as a test double.
Can it compare with hosted Jev?
Yes. The bench tool runs latency and maze tasks against both.
Related projects
Benchmark for agent regression after config edits: a Jev Choice decides which approved answers changed meaning versus wording.
BenchmarkresearchPython- jev-arena★ 122
Side-by-side tool comparing Jev and DeepSeek on labelling the same comments for relevance, sentiment and intent, with replay and reports.
ProjectbetaTypeScript Evaluation suite for decision models with 23,900 pinned rows across nine task families and three primitives, plus a leaderboard.
ProjectbetaPythonOpen-source BYOK arena where Jev faces another model as a judge on your question, with blind voting before costs are revealed.
ProjectbetaTypeScriptpytest-style tool that pins fixtures and expected Jev answers, then fails a model upgrade when answers flip or confidence drops.
ProjectbetaPythonTest harness for any Jev-API decision model: accuracy, calibration and robustness probes for reordering, injection and padding.
ServicebetaPython