User repository on GitHub
Bounded exploratory browser testing with Jev, deterministic assertions, and replayable evidence.
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Demos & Experiments
JevTest — CorieW
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.
- Stars
- 0
- Language
- TypeScript
- Published by
- Community
- Author
- CorieW
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jevtest — joshhu.
About this project
JevTest is a local TypeScript library and CLI for exploratory browser testing. Supply a goal, fixture data, available actions, and exact success assertions. Jev chooses actions and judges observed results; Playwright executes them in isolated sessions and saves replayable evidence. Status: initial working implementation. Source is publicly visible at CorieW/JevTest. Licensing terms are pending adoption; no usage license is granted yet and package publishing remains disabled. See licensing intent.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- Four outcomes separate the causes: passed, failed, incomplete when a limit or the API budget intervenes, and error.
- Ordinary tests, discovery, the baseline and replay make no model calls; only the live demo spends tokens.
- Replay re-checks recorded states, action definitions and assertions without asking Jev to choose again, reporting drift instead of adapting.
- The shipped demo runs 12 labeled flows against a local shop with six deliberately planted bugs.
Watch out
The README describes this as an initial working implementation, and licensing terms are still pending adoption: no usage license is granted yet and package publishing remains disabled, so treat the repository as source to read rather than a dependency to install. An `unexpected` model assessment creates a candidate issue even when every deterministic assertion passed, and it does not prove a bug; weak assessments are downgraded to `uncertain`, with probabilities and confidence retained in the report. Each flow gets a separate browser context, but the README is explicit that browser isolation alone does not isolate a shared database, so `setup`, `cleanup` and the supplied run ID are what isolate backend data. Config files are treated as trusted executable code. Requires Node 24 and pnpm 11, and the live demo shares one 250,000-token admission budget across at most 80 requests. 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.
The design is aimed at exactly that. Every run saves `trace.json`, page captures and masked screenshots, and the CLI has a `replay` command that checks recorded states, action definitions and assertions without asking Jev to choose again. Replay reports state drift rather than silently adapting the trace to what it finds now. Suite output also includes `report.html`, `summary.json`, `graph.json` and `graph.dot`.
Two separate mechanisms. The run command takes `--max-tokens`, with the documented live demo using a 250,000-token admission budget and at most 80 requests while recording actual provider usage. Separately, outcomes distinguish `incomplete`, raised by abort, cancellation, or a step, repetition, time or API budget limit, from `failed` and `error`, so a budget stop is never reported as a product bug. Ordinary tests, discovery, the baseline and replay make no model calls at all.
No, and the README says so directly. An `unexpected` assessment creates a candidate issue even when every deterministic assertion passed, which is the point of pairing model judgment with exact success assertions you write yourself. It does not prove a bug. Weak assessments become `uncertain`, and reports retain the underlying probabilities and confidence so a reviewer can see how much weight the judgment actually carried.
Node 24 and pnpm 11, then `pnpm install`, `pnpm browser:install`, `pnpm verify` and `pnpm demo`. To bring Jev into the loop, copy `.env.example` to `.env.local`, set `TYPESAFE_API_KEY` and run `pnpm demo:live`. On licensing the status line is unambiguous: terms are pending adoption, no usage license is granted and package publishing remains disabled, with the intent documented separately.
It starts its own local shop, runs 12 labeled flows with six deliberately planted bugs, replays every trace, and writes an HTML report plus metrics under artifacts/baseline-*. Expected planted failures do not make the command exit non-zero, so it reads as a benchmark to diff against rather than a gate.
Only the browser, not the data. Each flow gets its own browser context, but you still have to use setup, cleanup and the supplied unique run ID to isolate backend state, and the README is explicit that browser isolation alone does not isolate a shared database.
run drives flows with Jev choosing and judging, discover proposes flows for a named one, replay re-checks recorded states, action definitions and assertions without asking Jev to choose again, and --policy baseline reproduces the recorded behavior. Ordinary tests, discovery, the baseline and replay make no model API calls at all.
More Jev resources like this
98GitHub stars
Self-hosted Docker service and web UI that packages Laya structured-decision inference behind a compatible System One endpoint.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
1Panel-dev/laya-serverPython
1Panel-dev/laya-serverPython23GitHub stars
Browser drone-fleet lab where TypeSafe Jev steers each drone as fast reflexes and an optional planner offers one-use strategy advice when confidence drops, without pausing control.
Jev Demos & Experiments#community#jevbest#catalog-import
23GitHub stars
Chrome extension that sends each X post to Jev with six typed questions in one call near the viewport, then renders a behavior chip under the post plus a session cost and latency panel.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
oso95/x-scannerTypeScript
oso95/x-scannerTypeScript14GitHub stars
Manifest V3 Chrome extension in TypeScript that sends each LinkedIn post to Jev and collapses bait, self-promo, off-topic, and sponsored posts into a one-line bar with reasons and scores.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
stefw/lkcleanTypeScript
stefw/lkcleanTypeScript7GitHub stars
Demo that turns free-text UI descriptions into real shadcn components using only Jev typed Choice, Noul, and Score answers, with no generated code on the decision path.
Jev Demos & Experiments#community#jevbest#catalog-import
8GitHub stars
Claude-driven browser automation where plain-English chat drives Brave or Chrome, Jev makes on-page decisions, and a guard freezes irreversible clicks.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
ranjan2829/AskJevTypeScript
ranjan2829/AskJevTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
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 community
Posts from builders shipping with Jev right now.