Tired of smart home devices that need the internet to turn on a light bulb? Fed up with voice assistants that mishear "dim the bedroom" as "order three tons of cat food"? This session explores how to create a truly intelligent home automation system using Ollama for local AI processing and MQTT for reliable device communication. We'll examine how a custom MCP (Model Context Protocol) server translates natural language commands into MQTT messages, letting you control Z-Wave devices, smart switches, and sensors through conversational AI that runs entirely on your local network. Unlike corporate voice assistants that maintain corporate politeness, your local AI can have personality – imagine an assistant that responds to "turn off all the lights" with "Finally going to bed at a reasonable hour? Impressive." or sarcastically comments when you ask it to turn the heat up for the third time today.

We'll explore how this architecture bridges AI understanding and IoT device control, turning "make the living room cozy for movie night" into precise MQTT commands thatcoordinate multiple devices. The session covers integration patterns with popular platforms like Home Assistant and Z-Wave JS, examines Home Assistant's remarkable progress through their Year of Voice initiative (including their recent achievements in local voice processing), and demonstrates how structured prompting creates reliable device control while maintaining the personality you actually want in your home. No cloud dependencies, no monthly subscriptions, and definitely no corporate eavesdropping on your late-night snack requests – just a working example of how local AI can make your smart home both intelligent and entertaining.