User repository on GitHub
Dark Pattern HUD — system-wide overlay that flags manipulative UI via TypeSafe Jev
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Dark Pattern HUD (UXRay)
An Android heads-up display that reads visible on-screen text through an AccessibilityService, asks TypeSafe's Jev whether a manipulative UI pattern looks likely, and shows a floating badge above 0.9 confidence — password fields are never sent and card or PIN screens are skipped.
- Stars
- 0
- Language
- Kotlin
- Reading time
- 3 min
- Published by
- Community
- Author
- zohaib-md
- Last verified
- Oct 2026
- Capabilities
- Open sourceRequires API key
Highlights
- Confidence bands: 0.9 and up shows a floating badge naming the pattern(s); 0.6–0.9 writes a quiet review log entry; below 0.6 is discarded.
- The service never sends password-field values and skips the scan entirely on screens that look like card or PIN entry.
- Setup: install on a device with min SDK 26, paste a Jev API key, then enable the accessibility service and the display-over-other-apps overlay.
- Session stats (screens scanned, median latency, cost from input_tokens at $0.042 per million input tokens) stay in memory.
- Logcat tag UXRay logs the first real Jev response in full so the Noul probability field can be confirmed.
Watch out
Grounded in the repository README and GitHub metadata reviewed on 2026-10-15; the metadata lists no license and shows the repo created and last pushed on 2026-09-21, the test plan is manual and targets Indian shopping and food-delivery apps, and no accuracy or latency benchmarks are published beyond in-memory session stats.
Questions about this entry
Short answers drawn from the linked project or page.
It is an Android overlay: an AccessibilityService reads visible on-screen text and sends it to TypeSafe's Jev API, and a floating badge appears when a manipulative UI pattern looks likely. Tapping the badge shows detail naming the pattern, which you can dismiss.
Probability 0.9 and above shows the floating badge naming the pattern(s); 0.6 to 0.9 logs a quiet review entry in the app without interrupting; below 0.6 the screen is discarded. The README says the middle band is the interesting result, not only the obvious catches.
It never sends password-field values, and it skips the scan entirely on screens that look like card or PIN entry. API errors are logged and swallowed, so the service never crashes the foreground app.
Open the project in Android Studio and install it on a device or emulator with min SDK 26, launch Dark Pattern HUD, paste a Jev API key and save it, then enable the service under Accessibility and allow Display over other apps. The accessibility toggle cannot be granted from the app itself.
Session stats show screens scanned, median latency, and cost computed from input_tokens at $0.042 per million input tokens, all kept in memory. The README gives no other pricing, so check TypeSafe's current pricing before long sessions.
Open Indian shopping and food-delivery apps — Flipkart, Amazon.in, Swiggy, Zomato, Meesho — and scroll product, cart, and checkout-adjacent screens, stopping before entering card or UPI PIN. Screenshot where the badge fires and note the 0.6–0.9 review-log rows.
It is zohaib-md's Kotlin Android project, attributed to Mohammad Zohaib on the Made with Jev gallery, created and last pushed 2026-09-21 with 1 fork and no license listed. Accuracy is unproven in the captured material, so judge it on your own labeled screens.
More Jev resources like this
Chrome extension that classifies X posts with Jev and blurs chosen expression categories — discrimination, provocation, hatred, calls to attack, cynicism, resignation — behind a translucent overlay you can reveal with one click.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
eijiaraki/toxic-filterJavaScript
eijiaraki/toxic-filterJavaScriptA Chrome extension that hides posts you do not want on your X timeline: tap noise bubbles or type rules in plain words, and Jev judges each tweet one-shot before it scrolls into view. Good hide and Shouldn't have hidden answers stay local as an eval set.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
icpmacdo/signal-and-noiseHTML
icpmacdo/signal-and-noiseHTMLA browser extension that classifies web text with Jev: hover, pin, drag or double-click to judge blocks or whole articles as slop, or define your own questions. No build step; the key stays in storage.local and only api.typesafe.ai is reachable.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
kmcheung12/sniffslopJavaScript
kmcheung12/sniffslopJavaScriptModeration for communities and apps, powered by Jev: every message gets a probability for spam, scam, harassment, nsfw, doxxing and more, plus rules in plain English. Thresholds and actions are yours, and flag-only is the default.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ohernandezdev/jevmodPython
ohernandezdev/jevmodPythonA Manifest V3 Chrome extension that sends each LinkedIn feed post's text to TypeSafe's Jev with a rubric of narrow yes/no and 0–2 questions, then blurs only the posts that plain code marks as filter — uncertain and failed checks stay visible, with one-click reveal.
Jev Tools & CLIs#community#catalog-import#madewithjev
A schema-guided document-to-graph pipeline: local parsing feeds entity mentions and bounded candidate pairs to Jev, which chooses among closed-set relations, and every edge lands as proposed, review, or rejected with page-level evidence.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
chenmingtang830/jevgraphPython
chenmingtang830/jevgraphPythonFrom 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.