User repository on GitHub
Chrome extension for annotating articles, X/Twitter posts, and Threads posts.
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Demos & Experiments
Jevlens
A Manifest V3 Chrome extension that annotates X, Threads and articles with typed Jev decisions, analysing up to three post images locally and sending only sanitized structured evidence to the API.
- Stars
- 0
- Language
- JavaScript
- License
- MIT
- Published by
- Community
- Author
- knowlet
- Last verified
- Oct 2026
About this project
Manifest V3 Chrome extension for annotating articles, X/Twitter posts, and Threads posts with typed Jev / System One decisions. There is no JevLens proxy. The TypeSafe API key is stored in chrome.storage.local.typesafeApiKey and used directly by the service worker. This is convenient for a personal extension, but it is not secret storage. Anyone with sufficient access to the browser profile or extension debugging environment can recover the key.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- JavaScript
Highlights
- Image bytes stay in the browser: an offscreen document produces fixed visual labels, OCR and QR values. Only evidence is POSTed.
- Social posts get ten signal families from politics to adult promotion; general articles are analyzed only when you ask.
- The feed row shows only the three highest-scoring categories, drops anything under 2%, and keeps every remaining label in the hover title.
- A Jev-only release gate exits non-zero on a failed expectation and reads the key from the environment only.
Watch out
There is no proxy: the TypeSafe key sits in chrome.storage.local and the service worker calls the API directly, which the README calls convenient for a personal extension but explicitly not secret storage, recoverable by anyone with enough access to the browser profile or the extension debugging environment. The headline 100% regression figure is a deterministic local baseline and the README says outright that it is not a real-world accuracy estimate, with the synthetic holdout still not a production benchmark. Political and ideological labels are descriptive signals only, never truth ratings, endorsements or claims about an author, and they never trigger automatic filtering; only spam, bot-like, adult-content, adult-promotion and AI content-farm risk drive dim and collapse. Image analysis depends on Chrome's local model, which may need a first-use download, and degrades to whatever alt, OCR, QR and pixel evidence the browser exposes. The README also documents a real Threads false negative that a euphemistic caption slipped past in an earlier version. 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.
In chrome.storage.local, read directly by the service worker, because there is no proxy in front of it. The README is blunt that this is not secret storage and that anyone with sufficient access to the browser profile or the extension debugging environment can recover the key. The evaluation gate is the stricter path: it reads the key from the environment only and never prints or writes it to disk.
No image bytes leave the browser. An offscreen document runs the analysis locally and the raw image URL is never sent; the request carries the post text plus sanitized structured evidence such as alt text, dimensions, OCR text, QR values and fixed visual labels. A test in the suite asserts that no raw image URL reaches TypeSafe.
No, and the README says so in the same breath. That 159/159 field result over 37 cases is a deterministic local baseline, not an accuracy estimate, and the synthetic holdout at 45/50 fields and 15/18 cases is more informative but still not a production benchmark. Known local misses, including English city-council political context and euphemistic adult-promo text, are left visible on purpose rather than patched.
No. Political and ideological labels are informational and never trigger automatic filtering. Dimming and collapsing are driven only by spam, bot-like interaction, adult-content, adult-promotion and AI content-farm risk, and political labels fold into a single chip such as a tilt marker rather than a score you are meant to act on.
It extracts up to three images and analyzes them locally through an offscreen document using Chrome's built-in multimodal Prompt API plus TextDetector, BarcodeDetector and weak pixel statistics. Only sanitized structured evidence and extracted text reach TypeSafe.
An Evaluation Runner runs a local deterministic baseline with npm run eval:local and a Jev release gate with TYPESAFE_API_KEY npm run eval:jev, which exits non-zero on any failed expectation and reads the key from the environment only.
It shows the three highest-scoring enabled categories, gated on the model's choice for politics and risk scores for the rest. A category turned off in settings is removed before ranking, and scores below 2% are dropped.
More Jev resources like this
45GitHub stars
Drop-in OpenAI-compatible cache proxy for repeatable agent workflows: exact hits stay local, and with a TypeSafe key Jev judges whether a cached response is semantically reusable for the current request.
Jev Demos & Experiments#community#jevbest#catalog-import
33GitHub stars
Joke JavaScript helper that asks the Jev decision model whether a number is even and resolves to a boolean.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
wobsoriano/is-jevenJavaScript
wobsoriano/is-jevenJavaScript30GitHub stars
Side-by-side game harness where Laya running locally and Jev over API answer the same typed questions for snake, kombat, runner, flappy, tetris, and human-versus-model matches.
Jev Demos & Experiments#community#jevbest#catalog-import
22GitHub stars
Zero-dependency client that turns any logprobs-capable chat endpoint into typed choice, boolean, and score decisions with abstention.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
9pings/notjevJavaScript
9pings/notjevJavaScript24GitHub stars
Native macOS computer-use agent on Apple silicon that finds controls with local CoreML detection, OCR, and accessibility, then asks Jev to pick the next finite-choice action.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
jcpsimmons/jev-macos-loopJavaScript
jcpsimmons/jev-macos-loopJavaScript19GitHub stars
Six side-by-side web demos where Jev returns typed probabilistic decisions, plain code owns routing, and a switchable small model writes the words for Claude or Kimi tiers.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
mayank953/JevJavaScript
mayank953/JevJavaScriptFrom 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.