Algroveon Dashboard – Family Hub with Voice Assistant
10-inch touchscreen dashboard on Raspberry Pi 5: local voice assistant, widget interface for the hallway, fully integrated with the local AI home server.
A permanently installed touchscreen dashboard for the hallway with a local voice assistant – everyday information for the whole household, without browser or cloud.
PyQt6 application on Raspberry Pi 5 with Picovoice Porcupine for wake-word detection and direct integration with Algroveon-AI and Algroveon-Agent.
The project is a fixed touchscreen dashboard for the hallway that brings important everyday information for the entire household directly to the wall. At the same time, it serves as a local voice assistant with a direct connection to my AI home server – without a browser, without cloud services, and without external dependencies.
Hardware
| Component | Model |
|---|---|
| Computer | Raspberry Pi 5 (8 GB) |
| Display | Waveshare 10.1" Capacitive Touch, 1280×800 |
| Audio | XFV-3800 + Speaker |
| Motion Sensor | LD2450 Radar Sensor |
The radar sensor detects movement and activates the display from sleep mode.
Architecture
The UI is built using QML, a language used to describe user interfaces within the Qt environment—a widely used framework for graphical programming interfaces. The logic runs in Python based on PyQt6. The connection is established via the Qt Meta-Object System (Signals/Slots) through a central DashboardBridge. QML is solely responsible for the presentation.
QML Frontend (Widgets, Theme, State)
↕ Qt Signals / Slots
Python Backend (Services, Voice-Pipeline)
↕ HTTP (LAN)
Local AI Home Server (STT · LLM · TTS)
Widgets and Content
The fixed content of the dashboard includes time, weather, and a calendar with current and upcoming appointments. Additionally, it features real-time public transport connections from the home location, smart home information from Home Assistant, Spotify playback control, and a shared household shopping list.
The day/night theme automatically switches between a light display from 06:00 to 20:00 and a dark mode during the evening and night hours. Depending on detected activity, the system also switches between the states Active, Idle, and Sleep.
Voice Assistant
Wake-word "Hey Algroveon" via Picovoice Porcupine – running locally on the Pi. The pipeline then works as follows: Audio → STT on the home server (Faster-Whisper) → LLM (Ollama/Gemma-4) → TTS back to the Pi (Piper). Voice data never leaves the home network.
Speaker Recognition: The assignment of family members via voice vector using resemblyzer is still in testing and runs entirely locally on the Pi, without the cloud. In the future, a different context and memory area should be loaded depending on the recognized person.
Memory
A local memory system for facts, preferences, and notes per family member has already been established. However, the content design and the subsequent retrieval context are still being revised, as the voice assistant as a whole is not yet reliable enough.