What is jev-systemone-lab?

A lab built around a dispatcher skeleton for TypeSafe System One (Jev), with a browser harness that measures the dispatcher against ground truth and scripts for stress and calibration tests.

How does the dispatcher turn types into questions?

Python type hints become question shapes: Literal becomes a choice, list[Literal] becomes a set and bool becomes a flag, after which the dispatcher builds one batched request per tick.

What does the browser harness do?

It snapshots a live page into a state and action space with clickable refs, watches the dispatcher play in a headed browser, and measures the results against known ground truth.

What are the sample and hard room sets?

sample/scenarios.py has 20 obvious rooms that saturate at 100% and is described as a smoke test; sample/hard.py has 12 near-miss rooms with headroom for real evaluation.

How do I run the demo and tests?

The README lists python play_chrome.py -n 10 for a headed run, python stress.py --set hard for the four-way comparison and python calib.py for confidence versus truth.

Does it need an API key or extra dependencies?

It reads TYPESAFE_API_KEY from the environment or the user registry. The core jev package uses only the standard library and urllib; only the browser layer needs Playwright.

What is missing from the captured README?

The design-commitments section cuts off mid-sentence, so the action-space construction and scoring details are incomplete, and no measured results or licence details beyond MIT metadata are shown.

Jev directory

CommunityJev Demos & Experiments

Jev Systemone Lab

A Python dispatcher skeleton for TypeSafe System One (Jev) that turns typed signatures into batched questions, plus a Playwright browser harness that measures the dispatcher against ground truth.

See highlights Open github.com

Stars: 0
Language: Python
License: MIT
Reading time: 2 min
Published by: Community
Author: monkey1sai
Last verified: Oct 2026

monkey1sai/jev-systemone-lab

User repository on GitHub

View on GitHub

TypeSafe System One (Jev) dispatcher skeleton: typed signatures to batched questions, with browser and PR-evidence-gate specs on one engine

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity github python playwright

Works forJev for Classification Jev for Realtime control Jev for Evaluation

Used inJev for Software & DevTools

Languages: Python
Platforms: Playwright CLI
Capabilities: Open sourceRequires API key

Highlights

Converts Python type hints into question shapes: Literal becomes choice, list[Literal] becomes set, bool becomes flag.
Sends one batched request per tick from a page-derived state and action space, returning a typed call.
Core jev/ package uses only the standard library and urllib; only the browser layer needs Playwright.
Ships 20 obvious rooms that saturate at 100% and 12 near-miss rooms with headroom for real testing.
Includes stress.py for four-way comparison and calib.py for checking whether confidence is worth gating on.

Quickstart

bash

python play_chrome.py -n 10
python stress.py --set hard
python calib.py

Watch out

README reviewed 2026-10-17; the captured layout and run commands are visible, but the design-commitments section cuts off mid-sentence, so scoring internals, API-key setup details and measured results are unavailable. Metadata shows 0 stars.

Questions about this entry

Short answers drawn from the linked project or page.

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

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai