Help Center

Your Ngoma deployment package includes the robot unit, a pre-configured tablet, and network credentials. To get started:

• Connect the robot unit to your home network using the provided credentials
• Open the Ngoma client app on the supplied tablet (or download it from the App Store / Google Play / your web browser)
• Enter your Home Assistant base URL and access token when prompted on first launch
• Sign in to SoundCentral from Settings to enable media features

Your account manager will walk you through this during the onboarding call. See also Getting Started in the User Manual.

Ngoma is accessible from:

• Android tablets and phones (Android 11+)
• iPhones and iPads (iOS/iPadOS 15+)
• Web browsers (Chrome, Edge, Safari, Firefox — latest two versions)
• Windows and macOS desktops via the desktop client
• The robot's embedded touchscreen directly on the unit

All clients share real-time state — changes made on one device are reflected instantly on all others.

Yes. User management is handled through the Home Assistant admin interface. Your IT administrator can create additional HA long-lived access tokens for each user, or configure a shared read-only token for display-only devices such as wall-mounted monitors.

For user provisioning assistance, contact your account manager or raise a support ticket via the Contact Support section.

Ngoma has native mobile apps for Android and iOS as well as a full web interface. All three provide identical feature sets. The native apps offer push notifications and background processing for wellness monitoring and security alerts. The web app requires no installation and is suitable for desktop use or quick access from any browser.

Long-lived access tokens are managed directly in Home Assistant:

1. Log in to Home Assistant at your configured URL (e.g., http://192.168.1.100:8123)
2. Click your profile icon in the bottom-left sidebar
3. Scroll to Long-Lived Access Tokens
4. Revoke the old token and create a new one
5. Update the token in Ngoma → Settings → Home Assistant Connection

Go to Settings (⚙) → SoundCentral Account → Sign Out and confirm. Your session credentials are cleared from secure storage on the device. Entertainment features that require a SoundCentral account will be unavailable until you sign in again.

Facial recognition enrolment requires admin access:

1. Go to Settings → Manage Personnel → Add Person
2. Enter the person's name and their role/relation (Employee, Contractor, Guest)
3. Tap Enrol — position the person in front of the robot's camera when prompted
4. The Jetson runs a liveness check and captures the face model using ArcFace

The person will appear in the Security screen's Facial Recognition card immediately after successful enrolment. The Jetson backend must be connected and running for live recognition to work after enrolment.

Yes. SoundCentral is required only for media streaming features (music, video, live TV, live radio, and mood soundscapes). All other features — AI perception, security, smart environment control, wellness monitoring, and voice commands — work independently and do not require a SoundCentral subscription.

This error means the app cannot reach the Home Assistant server. Check the following:

• Home Assistant is running — open a browser on the same network and navigate to your HA URL
• Your device is on the same Wi-Fi network as the HA server
• The base URL in Settings has no trailing slash (correct: http://192.168.1.100:8123)
• Firewall rules on your router or HA machine allow inbound connections on port 8123

The "stale data" banner appears when the app loses its WebSocket connection to Home Assistant. The platform continues serving the last known device states from its cache, but live updates are paused. To resolve:

• Check that your HA server is reachable on the network
• Pull to refresh on the Smart Environment screen to force a reconnection attempt
• Verify the HA access token has not expired (HA tokens do not expire by default, but can be manually revoked)

Ngoma uses the following ports on your local network. Ensure these are not blocked by firewall rules:

Remote access is supported via Home Assistant's Nabu Casa cloud service or a VPN. Configure your preferred remote access solution in Home Assistant, then update the Ngoma base URL in Settings to the remote/cloud URL. SoundCentral and Knights Health API calls already go to cloud endpoints and work from any network. Contact your IT team or raise a support ticket for remote access configuration assistance.

Yes. When the Jetson AGX Thor edge unit is not connected, the perception system enters simulation mode and cycles through realistic data values. All platform features — including mood soundscapes, auto-ducking, and security alerts — remain fully functional in simulation mode. This allows you to test and demonstrate the platform before the Jetson hardware is deployed. Once the Jetson is connected and the MQTT broker is reachable, the platform switches to live data automatically — no app changes required.

Auto-ducking activates when the AI classifies the ambient sound as Conversation. False positives can occur if:

• Background noise patterns (HVAC, machinery) are similar to speech frequencies
• The Jetson is in simulation mode cycling through test values
• The YAMNet audio model is being run on low-quality audio input

To resolve: verify the Jetson is receiving clean audio from the microphone array. If you're in simulation mode, ducking will follow the simulated cycle — deploy the Jetson to use live audio classification. If the issue persists, contact support with the MQTT log output from the perception topic.

People counting uses YOLOv8 running on NVIDIA Jetson edge hardware. In standard office conditions (good lighting, unobstructed sightlines), accuracy is typically 95–98% for individuals within the camera's field of view. Accuracy may be reduced in:

• Low-light conditions (below 100 lux)
• Crowded environments where people overlap in frame
• Areas with significant glass or mirror surfaces

Camera placement recommendations are provided in your deployment guide.

The red error indicates that the sentry_mode_patrol scene entity is not found in Home Assistant. To resolve:

1. Log in to Home Assistant
2. Navigate to Settings → Automations & Scenes → Scenes
3. Create a new scene named exactly sentry_mode_patrol that activates your desired patrol behaviour
4. Return to Ngoma and toggle Sentry Mode again

Your deployment guide includes a sample patrol scene configuration. Contact support if the scene exists but the error persists.

Live facial recognition requires the Jetson backend to be running:

• Confirm the Jetson AGX Thor unit is powered on and connected to the same network
• Verify the Mosquitto MQTT broker is running (check via HA → Settings → Add-ons if using the HA Mosquitto add-on)
• Confirm the ArcFace pipeline is running on the Jetson — check the Jetson console or contact your deployment engineer

Enrolment data stored locally will be used automatically once the Jetson is connected — you do not need to re-enrol them.

Common causes and fixes:

• Obstacle in path: The Nav2 stack has detected an unpassable obstacle. Remove the obstruction or adjust the patrol waypoints to route around it.
• Motor fault: A motor error is indicated in the navigation status message. Power cycle the robot unit and retry. If the fault persists, contact hardware support.
• Navigation timeout: The robot could not find a valid path within 120 seconds. This may indicate a mapping issue — contact your deployment engineer to re-run the SLAM mapping routine.
• MQTT disconnection: The Jetson lost its broker connection during patrol. Check network stability and restart the MQTT broker.

Patrol waypoints are configured on the Jetson during the initial deployment mapping session (SLAM-based mapping with Nav2). The default configuration supports up to five waypoints. To add, rename, or reposition waypoints, contact your deployment engineer or raise a support ticket — this requires a re-mapping session and waypoint file update on the Jetson.

Wellness monitoring requires three things to be in place simultaneously:

1. You must be signed in to SoundCentral (Settings → SoundCentral Account)
2. The deployment must include KNIGHTS_BASE_URL and KNIGHTS_API_KEY environment variables — contact your account manager if these are missing
3. The person wearing the KnightGuard device must have the Wear OS companion app installed and authenticated with their Knights Health account

If all three are in place and you still see "Not Configured," contact support with your deployment ID for investigation.

The Wellness screen displays data exactly as received from the Knights Health API. Unusual readings may be caused by:

• Poor wrist contact — ensure the KnightGuard wearable is worn snugly on the wrist above the wrist bone
• Motion artefacts — brief high readings during vigorous movement are normal and will self-correct within seconds
• Connectivity lag — data is polled every 30 seconds; the displayed value may be slightly behind the actual current reading

For persistent anomalous readings, advise the wearer to remove and re-wear the device, then wait two polling cycles (60 seconds) for the reading to stabilise. Always consult a healthcare professional for any clinical concerns.

The current Wellness screen displays data for one KnightGuard wearable at a time, associated with the account linked to your SoundCentral sign-in. Multi-person wellness dashboards are on the roadmap — contact support for details.

"Unavailable" is the state reported by Home Assistant when it cannot communicate with the device. Common causes:

• Zigbee: Device is out of range, battery is low, or the Zigbee coordinator USB stick has disconnected — check HA → Integrations → Zigbee Home Automation
• Z-Wave: Similar range/power issues — check HA → Integrations → Z-Wave JS
• Network devices: The device's IP address may have changed — update in HA if using static IP assignments

Tap the Refresh button on the Devices screen after resolving the underlying issue in Home Assistant.

New smart devices are added through Home Assistant, not directly through Ngoma:

1. Pair the device with your Zigbee or Z-Wave coordinator following the device manufacturer's instructions
2. Home Assistant will automatically discover and add the device
3. Give the device a clear friendly name in HA (this is what appears in Ngoma and in voice commands)
4. Return to the Ngoma Devices screen and tap Refresh — the new device will appear

This error means the scene entity ID configured in Ngoma does not match an existing scene in Home Assistant. To fix:

1. In Home Assistant, go to Developer Tools → States and search for your scene
2. Note the exact entity ID (e.g., scene.office_morning_routine)
3. In Ngoma, go to Settings → Entity IDs and update the relevant scene field
4. Tap Validate entities to confirm the ID is found (✅)

Ngoma requires microphone permission to use voice control. If you denied the permission on a previous launch, you must re-enable it manually:

• iOS: Settings → Privacy & Security → Microphone → Ngoma → toggle ON
• Android: Settings → Apps → Ngoma → Permissions → Microphone → Allow
• Web: Click the lock icon in the browser address bar → Microphone → Allow

Ambient (always-on) voice mode requires the wyoming-whisper service to be running on the Jetson at the configured address. Check:

• Jetson unit is powered on and connected to the network
• The wyoming-whisper service is running (default port: 10300 on 192.168.1.200)
• Your device can reach that address (ping test from the same Wi-Fi network)

Tap-to-talk continues to function independently using the platform's built-in speech-to-text and does not require the Jetson service.

Voice commands match device names by their friendly name as set in Home Assistant. If a command like "Turn on the boardroom lights" isn't working:

• Check the exact friendly name in HA → Integrations → device list
• Keep friendly names simple and distinct — avoid special characters and abbreviations
• Speak clearly and use the full friendly name (e.g., "Turn on Boardroom Main Lights" not "boardroom")

The Entertainment screen requires an active SoundCentral session to load content. Steps to resolve:

1. Go to Settings → SoundCentral Account and confirm you are signed in
2. If signed in, check your internet connection — SoundCentral content is fetched from cloud servers
3. Pull to refresh on the Entertainment screen to retry — the client uses exponential backoff, so allow 30–60 seconds between attempts
4. If the issue persists for more than 5 minutes, check SoundCentral's service status or contact support

Live TV streams use HLS (HTTP Live Streaming). The player retries up to three times automatically. If all retries fail:

• Try a different channel — some streams may be temporarily unavailable at the source
• Check your internet connection speed (live HLS typically requires 3–8 Mbps per stream)
• Return to the original channel after 5–10 minutes — broadcast interruptions are usually brief