Office Assistant
LLMDockerHomeAssistantPythonPostgreSQL

Office Assistant

The Problem

The office space lacked any automated management - security monitoring required manual camera checking, temperature control meant physically adjusting the heater, and there was no intelligent access control. Staff had to remember to arm/disarm monitoring, turn off lights and heating when leaving, and had no way to receive alerts about unauthorized access while away.

Our Approach

Built a comprehensive automation system on Home Assistant running on Debian/Docker:

- Security layer: Camera with human detection triggering snapshots, CompreFace integration for face recognition, RFID card reader for access control

- AI integration: Conversation agent for natural language control, AI-enhanced Mattermost notifications with context-aware messages, personalized greetings using face recognition

- Climate automation: Temperature-triggered heating control (activates below 19°C when office is open), automatic shutdown when office closes

- Control interfaces: Zigbee buttons, Aqara cube for lighting control, NFC tags, voice commands, Nest Hub dashboard casting

- Polish TTS: Piper voice synthesis for all audio announcements

The Outcome

The office now operates autonomously - detecting and identifying visitors, adjusting temperature based on occupancy, and sending AI-crafted alerts with camera snapshots to team chat. One button press or RFID scan switches between "open" and "closed" modes, automatically handling monitoring, heating, and lighting. The system greets recognized users by name and announces state changes audibly.

Key Metrics

98%
Face Recognition Accuracy
<3s
Alert Response Time
30+
Active Automations

Cookie Policy

We use cookies to improve your experience on our website. You can customize your preferences.