TapInsight

§ 1.1Welcome

TapInsight turns your RayNeo AR glasses into a hands-free assistant that can see what you see, hear what you say, and act across the services you already use. It's built around three layers:

This guide walks through every capability — from the first adb install command, through your API keys, all the way to asking TapClaw to help with your homework. If you only read one section, make it How Gemini Routes You; it's the mental model everything else hangs on.

§ 1.2What You'll Need

Hardware

• RayNeo X3 AR glasses (the app targets this device's display and trackpad)
• A USB-C cable for the first-time install (the glasses pair to your computer over ADB)
• Either Wi-Fi on the glasses or a persistent USB tether to your computer

Accounts & Keys (free tiers are fine)

• A Google Cloud project (free to create) — this is the container for all your Google API keys and OAuth credentials
• A Gemini API key — the only required key; it powers the voice assistant
• Optional, but strongly recommended: Calendar, Maps / Directions / Places, Air Quality, and Cloud TTS keys
• Optional: Spotify client ID/secret if you want music playback

§ 1.3Install via ADB

Android Debug Bridge (ADB) is the command-line tool your computer uses to push apps to Android devices. The glasses present themselves to your machine as an Android device, so installation looks exactly like sideloading any APK.

One-time install

adb install TapInsight.apk

Once installed, TapInsight launches from the glasses app drawer. That's all ADB needs to do — unless you want the next, much more convenient, trick.

Connecting to the companion page over USB

The companion app (where you enter API keys, arrange the HUD, and tweak behavior) runs inside the glasses on port 19110. You have two ways to reach it from your phone or laptop browser:

adb forward tcp:19110 tcp:19110

§ 1.4First Boot & Pairing

1. Power the glasses on and launch TapInsight from the app drawer.
2. On your phone or laptop, open the companion page (see previous section).
3. Paste your Gemini API key into Step 2 and hit Save.
4. Double-tap the right temple of the glasses — you'll hear Gemini greet you.
5. Try: What can you do?voice

You're live. Everything else in this guide is optional polish — but the rest of Part I walks through the two concepts that make the app click: API keys and OAuth.

§ 1.5API Keys Explained

API keys are long random strings you get from cloud providers. They tell the provider "this app, owned by this account, is allowed to make requests." TapInsight stores them locally on the glasses — they never pass through any third-party server — and sends them directly to Google / Spotify / etc. when making calls.

Which key unlocks what

§ 1.6OAuth Explained

Some Google features — like writing to your calendar, creating Tasks, or saving Keep notes — need more than an API key. They need your permission, granted via Google's sign-in screen. That's OAuth.

The three-step OAuth setup

Spotify OAuth (similar pattern)

Step 6 of the setup tab does the same handshake with Spotify. The glasses need both a Client ID and Secret, plus a verified Redirect URI (the companion page tells you exactly what URI to register). Once authorized, you get streaming, search, and playback control via voice.

§ 1.7The Companion App — A Tour

The companion page is a mini web app served by the glasses on port 19110. It has several tabs across the top of the page:

Every tab is covered in detail in Part II.

§ 2.1Setup Tab

Every setup-tab section is numbered in the UI. Here's what each one is for, field by field.

Step 1 — Google Cloud Project

Not a form, just instructions: create a project at console.cloud.google.com, name it (e.g. "TapInsight"), enable billing. Every key you'll paste later should come from this project so quotas are consolidated and easy to monitor.

Step 2 — Gemini

• Gemini API Key — required.
• Model Override — leave blank unless you want to pin a specific model. Defaults: gemini-3.1-flash-live-preview for Live (voice + camera), gemini-3.1-pro-preview for regular turns, and gemini-3-flash-preview for fast fallbacks.
• Research Provider — who handles deep-dive research: gemini (default, search-grounded), openai_codex (code-heavy), or groq (fast Llama).
• Research API Key & Model — required if you picked a non-Gemini provider above. Presets are fast / balanced / deep.
• Research Role / Persona — tone of voice for briefs: analyst, engineer, journalist, academic.
• Gemini Readout Model & Voice — which TTS voice reads research aloud (Kore, Puck, Zephyr, etc).
• TTS Language + Director Notes — e.g. "calm, precise, analytical" — shapes how the TTS performs.
• LearnLM Tutor Model — the model LearnLM uses when you say learnlm before a question.

Step 3 — Google OAuth

Client ID + Secret + Start OAuth button. One-time; grants access to Calendar, Keep, Contacts, Tasks.

Step 4 — Google Calendar

An API key for read-only calendar queries (overlaps with OAuth but is needed for the HUD widget to pull data without round-tripping OAuth). Calendar ID defaults to your primary.

Step 5 — Maps, Routes & Places

One Maps API key covers Directions (routes + traffic + ETA), Places (nearby search), and the 3D Ask Maps tool.

Phone GPS Bridge (optional)

If the glasses don't have GPS or Wi-Fi-based location working well, toggle this to "Share This Phone's GPS". Your phone becomes the glasses' location source over the local network. Security notice: the phone must be on the same Wi-Fi/hotspot as the glasses, and the network can't have client isolation.

Step 6 — Spotify

Client ID + Secret, plus a Redirect URI the companion page will show you. Click Authorize Spotify to complete OAuth. Once linked, Gemini can search, play, skip, and scrub via Spotify Connect (the Spotify app must remain running on your phone).

Step 7 — Personality & Behavior

See §2.6 Custom System Prompts for the breakdown. Short version: a Personality field for one-sentence vibes, plus four editable prompt sections (Identity, Routing Rules, Behavior, URL Rules) and a Full Override for advanced users.

Step 8 — Audio Settings

• TTS Volume + Mute TTS
• Cloud TTS API Key, Voice Name (Standard / WaveNet / Neural2 / Studio tiers), Language

Step 10 — Skill Manager

View all enabled skills/tools. Add custom skills by scanning a skill QR code.

API Verification

Test buttons for each API: Test Calendar, Test Directions, Test Tasks, Test Places, Test Air Quality, Test Research, Test Location, Test Traffic. Run these after pasting a key to make sure it's working before you take the glasses on the road.

§ 2.2Gemini Live & Voice (Step 8b)

This section is where the assistant's personality on the wire is shaped. Most users can leave the defaults alone — but every field has a reason to exist.

§ 2.3Accessibility (Step 8d)

Everything that makes the app comfortable for the eyes and ears, in one place.

• Display Brightness — 5 %–100 % slider with a live percentage readout. Default 100 % (matches system brightness). Lower for indoor viewing or to save battery. Does not affect your phone's system-wide brightness.
• HUD Font Scale — 0.5×–3.0× multiplier for on-glass text size. Raise it for low vision.
• High Contrast HUD — thicker outlines, bolder colors for bright daylight.
• TTS Speech Rate — 0.5–4.0. Slow it down for learners and language study, speed it up for skimming briefs.
• Auto-Read Responses — on by default. Turn off if you want the HUD cards without voice every time.
• Default Assistant Mode — Audio Only (default) or Camera + Audio when you double-tap to start a session. Double-tap again to swap mid-session; third tap ends it.

§ 2.4HUD Display Settings (Step 9)

Decide what appears on the glasses when you're not actively talking. Each widget is a toggle:

More controls

• HUD Refresh Interval — 5–300 s. Lower = more current, higher = lighter on battery.
• Display Order — drag-reorder Calendar, Tasks, and News to decide what's on top.
• Enabled Calendars — if you have multiple Google calendars, pick which ones feed the widget.

§ 2.5Timeouts (Step 8c)

Every network request has a timeout. If a model is slow or your connection is spotty, raising these stops Gemini from giving up too early. 0 means use the default.

• Gemini Live Idle — how long Live stays open with no speech. Default ~120 s.
• Standard Gemini — per-turn timeout for non-Live calls. Default 45 s.
• Research — long briefs can take a while. Default 45 s; raise to 90–120 s for deep reports.
• LearnLM Tutor — same idea for tutoring responses.

§ 2.6Custom System Prompts (Step 7)

The system prompt is the instructions Gemini reads before every conversation. TapInsight ships a solid default but lets you swap any part of it. Expand the "System Prompt Sections" panel in Step 7 to see four editable blocks:

• Identity — "You are TapInsight, a proactive AI assistant integrated into RayNeo X3 AR glasses…" Change this to shift how Gemini refers to itself.
• Routing Rules — the decision tree Gemini uses to pick which tool to call for a given request. Edit only if you want to change default behavior (e.g. make research a different trigger word).
• Behavior — proactive rules, HUD output length (1–6 lines), privacy rules, "don't say 'I think' when stating facts," etc.
• URL Rules — when to send a direct URL vs. a search link. This is where YouTube, the video-list picker, and other browser routing get calibrated.

Personality

A separate short field injected after the system prompt. Try "dry, technical, precise — no pleasantries" for a focused assistant, or "warm, encouraging, always offer a next step" for a coach vibe.

§ 2.7Dashboard Tab

Your TapBrowser homepage is a grid of launch tiles grouped by category: Navigation, Social, Music/Streaming, Creative, Productivity, AI/LLMs, AI Music, AI Video, Voice, Utilities. The Dashboard tab is where you edit that grid — add, remove, reorder, rename. Changes push live to the glasses.

§ 2.8Browser Tab

• Desktop Mode — TapBrowser identifies as a desktop browser (no mobile redirects).
• Force Dark Mode — CSS filter that darkens light-themed sites. Default on.
• Show Browser Status Overlay — Wi-Fi, battery, and time strip. Toggle off for a cleaner view.
• Pointer Sensitivity — 0.4 (slow, precise) to 1.8 (fast, flicky). Default 1.0.
• Login Sync — paste cookies from a desktop browser to sign the glasses into YouTube, Gmail, Reddit, Twitter/X, Radio Garden, etc. The UI walks through where to copy each site's cookies from.

§ 2.9Radio Tab

Search and favorite internet radio stations — the results drive the tapradio tool. Data comes from the public Radio Browser API (30,000+ stations). You can favorite stations from the companion page and they sync to the glasses for instant voice playback.

§ 2.10OpenClaw Tab

Pair TapInsight with an OpenClaw agent running elsewhere (laptop, home server, cloud VM). Covered in depth under § 4.24 — TapClaw / OpenClaw.

• Scan QR / paste setup URL — one-step pairing if the OpenClaw host is reachable.
• Gateway URL — ws://…:18789 (local) or wss://… (Cloudflare Tunnel / Tailscale).
• Gateway Token — bearer token from openclaw config get gateway.auth.token.
• Session ID — default main.
• Heartbeat Update Interval — 10–120 s. How often TapClaw streams progress to your HUD.
• Test TapClaw Connection — ping the gateway.
• Vision Relay — view status, saved images, install instructions for Mac / Linux / Windows.

§ 3.1Gemini 3.1 Flash Live — the Brain

"Live" means a persistent, bidirectional audio stream between the glasses and Google's servers. You speak; Gemini starts generating speech within a few hundred milliseconds; the two of you can talk over each other like humans. Unlike text chat, there's no "send" button — turn-taking is detected from pauses.

What Flash Live is good at

• Latency. Replies start streaming while you're still finishing your sentence.
• Multimodal grounding. If camera mode is on, Gemini sees a live video feed alongside your voice. "What bird is that?" works.
• Tool calling. Gemini knows about every tool TapInsight has registered and picks the right one for your question.
• Search grounding. For current-event questions, Gemini automatically cites Google Search results — you get "today's" weather, not 2024's.

Real examples

§ 3.2Voice Sessions — Starting and Ending

The glasses wake on a double-tap of the right temple. That starts a Live session in the mode you picked in Accessibility:

• Audio Only (default) — just mic + speaker. Lightest on battery.
• Camera + Audio — adds the forward camera so Gemini can see what you see.

During a session:

• Double-tap again → swap between audio-only and camera+audio mid-conversation.
• Double-tap a third time → end the session. Gemini says goodbye and the mic closes.
• No speech for the configured idle timeout (default 120 s) also ends the session automatically.

§ 3.3How Gemini Routes You — The Mental Model

When you speak, Gemini doesn't just generate text. It chooses from three paths:

The routing rules in the system prompt tell Gemini when to pick a tool instead of just talking. A small set of high-priority triggers:

For anything not matched by a routing rule, Gemini either answers directly or invokes Google Search grounding to pull live web results. See the cheat sheet for the full list.

§ 3.4Barge-In & Interruption

Barge-in lets you interrupt Gemini mid-sentence by speaking. It's on by default — just start talking and Gemini yields. Two settings tune it:

• Barge-In Sensitivity — higher = more eager to cut Gemini off. Lower if you find it interrupting itself on background noise.
• Silence Threshold — audio below this level counts as "quiet." Raise in noisy environments.
• Interrupt Mode: Never — Gemini always finishes its current sentence before listening. Useful on stage, in a museum tour, or anywhere you will have background speech you don't want misread.

§ 3.5Camera vs Audio-Only Mode

The camera is a power-hungry sensor; you don't always want it on. TapInsight lets you pick per-session, with global default.

Commands that implicitly need camera will still work with audio-only mode — they'll just ask you to flip to Camera + Audio first, or they rely on the OpenClaw Vision Relay for a recent still frame (see §4.24).

§ 4.1HUD — Your Heads-Up Display

The HUD is the always-on display layer. It shows the widgets you enable in Step 9 — calendar, tasks, news, traffic, notifications — and is also where Gemini's replies surface as concise text cards (1–6 lines per card, by default).

What Gemini puts on the HUD

• Reply cards — short text summaries of what Gemini just said.
• Tool results — calendar events, route ETAs, place cards, research briefs, translation output.
• Progress indicators — when Gemini calls a slow tool (research, TapClaw), a status line appears so you know something is still happening.

Commands that modify the HUD

§ 4.2TapBrowser — The AR Web Browser

When Gemini needs to show you something bigger than a card — a YouTube video, an article, a map, a recipe — it opens TapBrowser. It's a full Chromium-based browser tuned for the glasses.

What's in it

• Dashboard homepage with grouped quick-launch links (edit in the Dashboard tab).
• Reader mode for clean article reading.
• Picture-in-picture for videos.
• TTS read-aloud for articles and anything selectable.
• A night-sky audio visualizer (see next section) you can bring up in dim mode while music or video is playing.
• Desktop or mobile mode — site compatibility lever for difficult pages.
• Force dark mode — easier on the eyes at night.
• Volume and brightness sliders via gestures or a settings overlay.
• Text zoom independent of overall UI scale.
• QR code scanner — quick URL entry without typing.
• Bookmarks — add favorites, open from the dashboard.
• Cookie import — paste cookies from your desktop browser to log sites in on the glasses.

Common voice paths into the browser

§ 4.2ALibrary — Local Media & Playlists

The Library is TapInsight's on-device shelf for media you own: local music, videos, text files, and playlists stored directly on the glasses. It is designed for offline or low-friction playback when you do not want to depend on YouTube, Spotify, or a network connection.

Native video playback new in 0.4

Local and SMB videos play through a native ExoPlayer path. A folder queue with previous/next controls lets you move through everything in a folder, and the media controls stay up longer so you have time to pick subtitles. A CC language picker lists the available subtitle tracks and supports sidecar .srt files placed alongside the video. HDR display correction is tuned for the X3 Pro panel so colors and brightness look right on the glasses.

Where the files live

TapInsight creates an app-private media folder on first run:

/storage/emulated/0/Android/data/com.rayneo.visionclaw/files/Media/

• Music/ — default audio folder.
• Videos/ — default video folder.
• README.txt — a built-in reminder of the folder rules and supported formats.
• Nested folders — fully supported, so you can organize by artist, album, topic, course, project, or anything else.

This storage is visible over USB / MTP, requires no runtime permission prompts, survives reboot, and is normally wiped only if you fully uninstall the app.

Supported formats

• Audio — mp3, m4a, aac, ogg, opus, wav, flac, weba
• Video — mp4, webm, mkv, mov, m4v, 3gp, avi
• Text / TTS — txt, md, markdown, log, csv, json, xml, html, rtf, ini, cfg, conf, yaml, yml, toml
• Playlists — m3u, m3u8

How .m3u playlists behave

Playlist paths are resolved relative to the playlist file's own directory. That means a playlist can reference files next to it, in child folders, or in sibling folders using standard relative paths such as ../Videos/clip.mp4.

#EXTM3U
#EXTINF:180,Morning Coffee
coffee.mp3
#EXTINF:240,Subway Ride
../Videos/commute.mp4

The companion Library page can also auto-generate a playlist for the current folder and open a visual editor so you can add, remove, and reorder entries without hand-editing the file.

How to use it

• From the companion app — open the Library tab to upload media, browse folders, generate playlists, or edit existing .m3u files.
• On the glasses — open the Media Library tile from the TapLink Dashboard. If it is missing, the companion Library page can add it back for you.
• For reading — open a text file in the library and let Gemini TTS read it back like a local audiobook or spoken note.

§ 4.3The Audio Visualizer

In dim mode, swipe up on the right side of the touchpad to bring up the night-sky audio visualizer; swipe down to close it. It's a single calm scene — captions and titles stay out of it — that breathes at a coherent-breathing pace (roughly 4s in, a short hold, 6s out) while reacting to whatever is playing.

The visualizer is frequency-reactive: it splits the audio into bands and maps them to the scene. Sub-bass glows up from the bottom of the frame, bass pulses the central orb, and bass, mids, and highs ripple through three distinct ribbons — bass broad and heavy, mids articulated, highs tight and quick. Highs also shimmer as stars, with an "air" band setting how fast they twinkle. When no audio is captured, it degrades gracefully into a pure breathing exercise with gentle synthetic motion.

§ 4.4TapRadio & Podcasts

TapRadio is a first-class tool — tapradio — that handles internet radio and podcasts. It pulls from the Radio Browser public API (30,000+ stations) and standard podcast feeds. The routing rule makes it the highest-priority responder for anything involving the word "podcast" — Gemini never makes up podcast names from memory; it always calls the tool first.

Radio

Podcasts

Favorites

Favorite stations from the Radio tab of the companion app sync to the glasses. "Play my favorite station" picks the most-recent favorite, or you can name one — "Play my Radio Paradise favorite."

Song Identification — AudD (optional) new in 0.4

Ask "what song is this?" while listening to TapRadio. With an AudD token, TapInsight asks AudD to fingerprint the current stream first so stale station metadata doesn't win. Without a token, TapInsight falls back to the free track metadata broadcast by many stations. Get a free token (300 IDs/month) at audd.io and add it in the companion app.

§ 4.5YouTube Workflow

YouTube is one of TapInsight's most polished flows because it's routed through Gemini in two phases — describe first, play on command — so you never end up with a video playing when you just wanted to know what's out there.

Phase A — Describe & send

When you ask an informational question, Gemini gives a short verbal summary, sends a tappable list of picks to your glasses, and ends with "tap one to play, or hit the red button to open them all as a playlist."

Phase B — Play

When you say a playback verb — "play that," "play the first one," "watch the Deviant Ollam one" — Gemini opens the matching video in TapBrowser with autoplay.

YouTube-specific voice commands

§ 4.6Research & Deep Dives

The research_topic tool produces long-form, search-grounded briefs that Gemini reads verbatim — no paraphrasing, no summarizing. It's for when you actually want to learn something, not just get a one-liner.

Triggers

Don't use "research" for casual "tell me about," "analyze," "brief me on," or "explain" — those route to normal Gemini and come back much faster.

Presets (set in Step 2)

• Fast — a tight one-pager, ~15–30 seconds.
• Balanced — default. 2–3 paragraphs with citations.
• Deep — multi-section brief with structured headings, up to a minute or two.

Readout voice

The brief is read aloud with the voice and director notes you set in Step 2. Try "calm, precise, analytical" for a podcast feel, or "energetic, curious, warm" for a morning briefing.

Alternate providers

Change Research Provider in Step 2 to openai_codex (for code-heavy topics with GPT-5.2-codex) or groq (Llama 3.3 70B — very fast). You need that provider's API key pasted below.

§ 4.7LearnLM — The Tutor

LearnLM is Google's tutoring model. It teaches rather than answers: it asks probing questions, scaffolds concepts, and remembers your past cards across sessions. Activate it by prefixing your request with learnlm.

§ 4.8Google Search Grounding

Gemini 3.1 has native Google Search grounding. You don't toggle it — TapInsight leaves it on so any current-world question (weather, stock prices, recent news, sports scores, "who won the Nobel Prize this year") pulls fresh search results and cites them in the reply.

Grounding is different from the research_topic tool:

§ 4.93D Maps — Ask Maps

The ask_maps tool combines Google's 3D photorealistic map tiles with landmark data and directions. It's the "show me" for geography.

Ways to invoke it

Camera integration

Point the glasses at a landmark with Camera + Audio mode on, and "What is that?" triggers Ask Maps to ID the building, park, or natural feature in view, with a short card of info and an option to navigate there.

§ 4.10Directions & Traffic

Backed by Google Directions. The tool is google_routes. Origin defaults to your current location unless you specify otherwise.

§ 4.11Places Nearby

Tool: google_places. Google Places Nearby Search + ratings + open/closed status.

§ 4.12Calendar

Tool: google_calendar. Read and write with OAuth; read-only with an API key + Calendar ID.

§ 4.13Tasks & Reminders

Tool: google_tasks. Needs Google OAuth with Tasks scope. Also drives the Tasks widget on the HUD.

§ 4.14Notes (Google Keep)

Tool: google_keep. Quick-capture notes that sync to Keep on your phone.

§ 4.15Contacts Lookup

Tool: google_contacts. Use it to retrieve saved phone numbers and email addresses from your Google contacts. TapInsight does not place calls or send texts directly.

§ 4.16Spotify

Tool: spotify_player. Two actions: search (fact lookups — who sings, what album, when released) and play (streaming via Spotify Connect).

Premium gets full tracks; free tier gets 30-second previews. The Spotify app must be running on your phone — if the phone isn't listed as a Connect device, open Spotify, play any track for a second, and retry on the glasses.

§ 4.17Sonos

Tool: sonos_control. If you have Sonos speakers on the same network, voice commands can target them instead of the phone.

§ 4.18Weather & Air Quality

Weather is rolled into the Daily Briefing tool and answered via Google Search grounding. Air quality is a dedicated tool: google_air_quality.

§ 4.19News Headlines

Tool: google_news. Uses the public Google News RSS feed — no API key needed.

§ 4.20Daily Briefing

Tool: daily_briefing. One call, many tools — calendar + weather + traffic + air quality + proximity + parking + public events, rolled into a two-minute morning rundown.

Combine with a quick action macro named "good morning" so a single trigger phrase runs the whole thing plus your music.

§ 4.21Camera Actions

Tool: camera_action. Save photos, read QR codes, start/stop audio recording. Camera frames are also used by:

• Live sessions in Camera + Audio mode — streaming visual context to Gemini.
• Ask Maps landmark ID.
• Translate's camera source for reading signs and menus.
• The OpenClaw Vision Relay — the remote agent gets the latest still frame.

§ 4.22Translation

Tool: translate_text. Works on speech and camera-visible text (signs, menus, documents). 40+ languages.

§ 4.23Battery Saver

Tool: battery_saver. Cuts HUD refresh, pauses camera analysis, switches to voice-only mode, and defers non-essential network calls. Auto-enables under the threshold you set in Step 8g (default 20%, set to 0 for manual-only).

§ 4.24Quick Actions / Voice Macros

Tool: quick_action. A macro is a chain of tool calls triggered by a phrase. TapInsight ships four pre-built ones — good morning, leaving work, heading home, meeting mode — and lets you define your own via JSON in Step 8h.

Custom action JSON

[
  {
    "name": "gym time",
    "description": "Directions to gym + a hype playlist",
    "steps": [
      { "tool": "google_routes",   "args": { "destination": "Planet Fitness" } },
      { "tool": "spotify_player",  "args": { "action": "play", "query": "workout mix" } }
    ]
  }
]

§ 4.25TapClaw / OpenClaw — Your Remote Agent

TapClaw is an agent that runs on a computer outside the glasses — your laptop, a home server, a cloud VM — and has access to your files, emails, and whatever else you plug into it. When you prefix a request with tapclaw, TapInsight bundles the request (plus the latest camera frame if relevant) and sends it to the agent. The agent does the work and streams results back to your HUD.

Why it matters — real-life use cases

Pairing the agent

Vision Relay

A tiny HTTP service on port 18790 that receives live camera stills from the glasses. On macOS: bash tools/install-relay.sh. On Linux or Windows: python3 tools/image_relay.py. The relay rotates the frame 90° clockwise and saves it to ~/.openclaw/workspace/camera_frame.jpg, archiving each frame under saved_photos/YYYY-MM-DD/. The agent can read the latest frame via GET /latest.

TapClaw voice commands

§ 5.1Voice Command Cheat Sheet

Every trigger phrase that has a dedicated routing rule. Gemini matches loosely — phrasing variations of these all work.

§ 5.2Every Native Tool

Under the hood, Gemini sees these tools by name. Useful if you're editing the routing rules or building custom quick actions.

§ 5.3Trackpad & Gestures

The glasses have a touchpad on the right temple. Gestures:

§ 5.4Troubleshooting

Gemini doesn't respond

• Check the Gemini key in Step 2. Hit Test in API Verification.
• Check connectivity — Wi-Fi or USB-tether, whichever the glasses use.
• Check Gemini Live Idle Timeout — if zero, a very brief pause can end the session.

"Spotify not available" when trying to play

• Open the Spotify app on your phone and play any track for ~1 second. That registers the phone as a Spotify Connect target.
• Retry on the glasses.

Companion page won't load

• If over USB: re-run adb forward tcp:19110 tcp:19110. Then try https://localhost:19110 first and accept the certificate warning. Check adb devices shows the glasses.
• If over Wi-Fi: verify the IP still matches (DHCP may have changed it).

TapClaw connection fails

• Check the gateway token in the OpenClaw tab. Copy it fresh from openclaw config get gateway.auth.token.
• If using a Cloudflare Tunnel, verify cloudflared tunnel list shows it as connected.
• Try the Test TapClaw Connection button.

HUD widgets show "—"

• Calendar / Tasks: OAuth expired. Re-authorize in Step 3.
• Traffic: Maps API key missing or quota exceeded.
• News: Network down (uses public RSS, no auth).

Gemini opens YouTube when you just wanted info

• Default routing uses Phase A (describe + send list). If it jumps straight to playing, check Step 7 → URL Rules — make sure you haven't overridden the default.

Display too bright / too dim

• Accessibility → Display Brightness slider. Default 100%; lower for indoor comfort. Use Reset to default to snap back.