The AI Technologies Powering Aurix

As I promised earlier, I'd like to share the technologies behind Aurix (Adaptive User Responsive Intelligent eXperience).

Before diving into that, I want to briefly highlight the updates I completed last week—updates that significantly enhance Aurix's ability to sense the world, learn from it, and move freely and independently.

Over the past week, I refactored the component architecture and improved how the modules are connected. I also integrated a battery system, making Aurix fully autonomous and no longer dependent on external power.

To improve environmental awareness, I added an ultrasonic sensor for object detection and obstacle avoidance during movement. In addition, I integrated a camera, allowing Aurix to finally "see" the world and recognize objects through visual input.

Now, let's get back to the technologies used. Aurix's intelligence can be divided into four main AI models:

* Wake Word Detection – The trigger phrase that activates Aurix.

* STT (Speech-to-Text) – Converts human speech into text (Faster Whisper).

* LLM (Large Language Model) – Acts as Aurix's brain (LLaMA-3.3-70B-Versatile).

* TTS (Text-to-Speech) – Serves as Aurix's voice (Piper).