Discover nearby landmarks through your Even G2 smart glasses. Wondereye finds points of interest around you using OpenStreetMap and Wikipedia place data, and shows them right on your display, ranked by distance and significance. Landmark descriptions are written by Grok. Wikipedia is used only to find nearby places.
- Homepage: https://wondereye.app
- Community map: https://wondereye.app/map
Wondereye runs on Even Realities G2 smart glasses and is installed from Even Hub in the Even Realities App on your phone. The glasses show text only; there is no audio.
Developers with Even Hub Developer Mode can also load the current web build by scanning this QR code with Scan QR in the Even Hub tab:
- Automatic landmark detection: up to 5 nearby landmarks found with OpenStreetMap and Wikipedia place data, ranked by distance and significance
- Instant descriptions written by Grok, with full background details on demand
- Compass direction and distance to each landmark, shown right on the HUD
- Voice search: choose Voice Search and say a landmark name; the glasses microphone records the query and the app finds the match
- Adjustable search range — tighten it for dense city centers or widen it to explore
- Community map — every landmark discovered by the community adds to a shared world map at wondereye.app/map; view your own contributions in the app
- Recently viewed landmarks saved for later
- Imperial or metric units, your choice
- Eight languages: English, German, French, Spanish, Italian, Simplified Chinese, Japanese, and Korean (descriptions are written by Grok in the selected language; only English descriptions are added to the community map)
Wondereye gets your position from your phone through the Even Hub SDK (getAppLocation, SDK 0.0.11+) — no manual setup required. If a live fix isn't available yet, it falls back to your last known location, then to a default city. On the simulator you can pass lat/lng query params to simulate a position.
- Tap a landmark to read a snippet, then tap again to load full details
- Scroll up/down to move through the list or paginate details
- Double-tap to go back from any view
Open Wondereye's settings from the Even Hub app to:
- Choose the language
- Set your search radius
- Choose imperial or metric units
- Toggle device location on/off
- View My Map Contributions — the landmarks you've discovered
- Send feedback (bug, idea, or other) — posted publicly as a GitHub issue; no email or account
- Support the project
Your location is obtained on-device via the Even Hub SDK and is used only to look up nearby landmarks — it is never written to storage. Landmark cache keys are rounded to ~110m precision. No name, email, or other personal information is collected or stored anywhere in this system.
npm install
# Create worker secrets — copy to packages/worker/.dev.vars:
# XAI_API_KEY=your-xai-api-key
# ALLOWED_ORIGIN=http://192.168.86.100:5173 (use your dev machine IP)For testing on physical glasses or remote simulators, set your dev machine's IP:
# Edit packages/frontend/.env.local with your network IP
# (e.g., 192.168.86.100 or 192.168.1.100)
cat packages/frontend/.env.local
# Then run all servers:
bash scripts/dev-local.sh
# OR manually in 3 terminals:
cd packages/frontend && npm run dev
cd packages/worker && wrangler dev --ip YOUR_IP --port 8787
npm run qr # scan QR code with Even Hub app on glassesFor desktop/simulator testing with localhost:
npm run dev # starts frontend on localhost:5173
npm run qr # generates QR code# With network IP (physical glasses):
npx evenhub-simulator "http://192.168.86.100:5173/app.html?lat=48.8566&lng=2.3522"
# With localhost (desktop/emulator):
npx evenhub-simulator "http://localhost:5173/app.html?lat=48.8566&lng=2.3522"Pass any lat/lng query params to simulate a location.
POST /api/feedback files filtered feedback as GitHub issues. It is off unless
FEEDBACK_ENABLED=true and needs two Worker secrets (TURNSTILE_SECRET,
GITHUB_FEEDBACK_TOKEN). See docs/feedback.md for the
abuse defenses, every binding, and local preview steps.
cd packages/worker && npm test # vitest: validation, rate limit, Turnstile, dedupecd packages/frontend && npm run pack # builds and packs wondereye.ehpknpm run deploy # worker + frontend
npm run deploy:worker # worker only
npm run deploy:frontend # frontend only© 2026 CMDLAB LLC. Place data © OpenStreetMap contributors.