Whole-Home Smart Living
Scenario details
Voice chat that knows the family, climate-linked AC and lighting, camera and door alerts.
Pain Points & Challenges
Generic smart speakers cannot differentiate between family members, mixing personal profiles and schedules; uploading household audio and facial biometric data to public clouds poses grave home privacy leak hazards, worsened by sluggish network lag.
Edge AI Solution
Deploy smart touch terminals with on-device vision and speech NPUs that match registered family faces locally for personalized greetings, parsing voice intents offline to orchestrate smart home devices with zero biometric data leakage.
Core Advantages
- On-device face recognition, up to 20 people
- Greets recognized family by name
- Conversations cast to a TV or monitor
Pain Points & Challenges
Residential HVAC and lighting run on rigid timers or single-wall thermostats, chilling empty rooms for hours and driving up electric bills; fragmented wireless smart protocols isolate sensors from thermostats, blocking unified multi-room automation.
Edge AI Solution
Install mmWave presence radars and multi-parameter climate sensors in every room alongside split-core CTs in circuit panels, allowing a local edge hub to orchestrate micro-zoned HVAC and lighting triggers over local LAN.
Core Advantages
- Temperature, humidity, CO₂
- Power per circuit
- AC and lighting automation
Pain Points & Challenges
Standard PIR and basic security cameras trigger incessant false alarms from moving pets, billowing curtains, and lighting shifts; continuously streaming live residential surveillance to the public cloud leaves private spaces vulnerable to hacking.
Edge AI Solution
Deploy edge AI cameras running on-device human pose filtering to distinguish real intrusions from pets, dispatching event flags only to local smart hubs to pair with sub-GHz door and leak sensors for secure, zero-image-leak home security.
Core Advantages
- On-camera person detection
- Door and window events
- Water leak alerts
Deployment & Selection
Pick the home hub, then voice terminals, sensors and actuators per room.
Not sure which to choose? See the full selection guide →
Scenario
Select the smart living package best suited to your deployment
Inputs route to local gateways and hubs with closed-loop device triggers
Voice Terminals maps to Inputs, Home Hub to local compute, and Sensing, Control, Add-ons covers field acquisition and actuation & output; optional compute is the add-on HVAC gateway, voice host or cloud dialog.
Core Advantages
- Voice and sensors come in first
- Rules run on the local hub
- Output drives lights and HVAC
| Image | Product | Connectivity | Use Case |
|---|---|---|---|
 | Home Assistant Green | Ethernet, Zigbee, Thread, Matter | Home: Ethernet plus consumer protocols (Zigbee / Thread / Matter) |
 | reComputer R1125-10 | WiFi, BLE, Ethernet, RS485, CAN bus | Industrial-protocol gear: RS-485 / RS-232 / CAN plus DI/DO |
One local hub per home running all automation rule scheduling
One HA hub per home: Ethernet for the network, and Zigbee, Thread and Matter devices through USB radio dongles (see Sensing, Control, Add-ons). When the site has professional gear on RS-485, RS-232 or CAN — HVAC controllers with BACnet or Modbus, for instance — add a reComputer R11 series gateway beside the hub.
Core Advantages
- Pre-installed smart home hub ready to run out of the box
- Local automation rules run offline even without internet
- Industrial edge gateway bridges legacy HVAC controllers
| Image | Product | Connectivity | Use Case |
|---|---|---|---|
 | Home Assistant Voice Preview Edition | WiFi, BLE | Voice control through HA |
 | Watcher W1-B White Enclosure | WiFi, BLE | Voice assistant with camera |
 | reTerminal D1001 8" LCD Touch Display | 4G LTE | 8" touchscreen voice terminal |
 | reSpeaker XMOS XVF3800 | USB, I2S | Far-field pickup for a local host |
 | reComputer RK3576-30 | Ethernet, WiFi, 4G LTE, LoRaWAN, Wi-Fi HaLow, Bluetooth | Runs speech and chat models locally |
Voice assistants for daily routines, local LLMs for private dialogue
For voice control of appliances only, choose Home Assistant Voice. For a camera or touchscreen, choose SenseCAP Watcher or reTerminal D1001. To run the chat model locally, connect a reSpeaker XVF3800 mic array to an RK3588 + RK1828 or Jetson Orin NX 16GB host; an RK3576 host runs speech recognition and synthesis locally while the chat model stays in the cloud.
Core Advantages
- Voice assistant terminals connect to the hub directly
- Touchscreen displays and smart cameras recognize faces
- Noise-cancelling mic arrays with local edge chat models
| Image | Product | Connectivity | Use Case |
|---|---|---|---|
 | SONOFF SNZB-02P Zigbee Temperature and Humidity Sensor | Room climate | |
 | SONOFF SNZB-01P Zigbee Switch | Scene button | |
 | SONOFF SNZB-04P Zigbee Door/Window Sensor | Door/window | |
 | Sonoff SNZB-03-Zigbee Motion Sensor | Motion | |
 | reCamera 2002 8GB | SSH, HTTP, DNS | On-device person detection |
 | LWL02 Water Leak Sensor | LoRaWAN | Water leak (LoRaWAN) |
 | Indoor Gateway(SX1302) | LoRaWAN, Ethernet, WiFi, BLE | LoRaWAN gateway for leak sensors |
 | XIAO 2-Channel Wi-Fi AC Relay | WiFi | On/off control |
 | Indicator D1 | WiFi, BLE, LoRa | Wall or desk display panel |
 | Home Assistant Connect ZBT-2 | Zigbee, Thread, Matter | Zigbee / Thread devices |
 | Home Assistant Connect ZWA-2 | Z-Wave devices |
What each room senses, switches and adds, in one place
Zigbee sensors cover climate, doors and windows, motion and buttons. reCamera detects people on the device. LWL02 leak sensors use LoRaWAN; the SenseCAP M2 gateway forwards to an MQTT broker (or TTN / ChirpStack) that Home Assistant subscribes to. Wi-Fi relays switch lights and appliances; SenseCAP Indicator is the display panel. The hub has no radio: add Connect ZBT-2 for Zigbee / Thread, Connect ZWA-2 for Z-Wave.
Core Advantages
- Low-power sensors cover climate, doors and motion
- Relays and touch displays execute appliance control
- Multi-protocol RF expansion links wireless ecosystems
FAQ
Automations on the Home Assistant host run locally and keep running without internet. Wake-word dialogue on Watcher and reTerminal D1001 goes through the Xiaozhi service and needs the network; face recognition runs locally on the terminal. For offline dialogue, use the fully local package of the Conversational Voice AI reference design (RK3588 + RK1828 or Orin NX 16GB), which runs offline once the models are cached.
Face enrollment and recognition run on the terminal; face data stays on the local device and is not sent to the cloud, with up to 20 people enrolled. Dialogue after the wake word is handled by the online Xiaozhi service.
One Home Assistant host runs the automations for the whole home. The host has no radio: add Connect ZBT-2 for Zigbee / Thread devices and Connect ZWA-2 for Z-Wave; HVAC controllers with RS-485, RS-232 or CAN interfaces need an extra reComputer R11 series gateway.
No. It targets central air-conditioning controllers with OPC UA, Modbus or BACnet/IP interfaces and does not apply to split units. It starts in observe-only mode and writes no points until an operator enables control.
Yes, standard 1080p 30fps H.265 or H.264 RTSP streams from existing cameras can be decoded and processed locally on Jetson Orin, RK3588, or Hailo-8 hosts. Detection alerts publish over MQTT, and raw video stays on the local network.
Yes. The Seeed devices used in this solution can be customized with your own logo, enclosure, packaging and pre-loaded firmware, and existing models can be adapted to add or drop interfaces and features. See Customization Service for scope and process, or .