Skip to main content

The AI Engine: How the Brain Thinks

Our AI system, HX47, isn't just a chatbot. It's a "thinking engine" that connects your words to your physical smart devices. This guide explains how we keep that thinking safe and accurate.

1. The Two-Part Brain​

To keep you safe, we split the AI's work into two separate areas:

  • The Creative Part (Python): This is where the AI dreams up plans. It can be creative and understand human language, but it's not allowed to touch your devices directly.
  • The Logical Part (TypeScript): This is the strict supervisor. It takes the AI's plans, checks them against our safety rules, and only then carries them out.

2. The Step-by-Step Plan​

When you ask for something, the AI creates a Digital To-Do List (a Graph):

  • Each task has to wait for the previous one to finish.
  • If a task fails (like a light not responding), the AI can stop the whole plan or try an alternative.

3. Double-Checking Reality​

One of our coolest features is Reality Validation.

  • Before the AI starts a plan, it takes a "snapshot" of your home's state.
  • If something changes while the AI is thinking (like you manually turning off a light), the AI will see that its plan is based on old information and will stop immediately to avoid making a mistake.

4. The "One at a Time" Rule​

To prevent confusion, we have a Lease System.

  • Only one part of our cloud can work on your AI plan at a time.
  • It's like a "talking stick" in a meeting—you can't speak unless you're holding it. This prevents two different parts of the AI from trying to control the same light at the same time.

5. Learning & Memory​

The AI has its own Memory Vault.

  • It remembers your preferences (like "I prefer the bedroom lights to be warm at night").
  • It can also spot patterns, like an "anomaly" (something unusual happening), and notify you or the Smart Shield.