This project creates a real-time voice AI pipeline using WebSockets and edge computing to connect over 100 speech-to-text, LLM, and text-to-speech models. It runs on Arduino ESP32 hardware with Cloudflare Workers AI and integrates services like OpenAI Realtime API, Deepgram, ElevenLabs, Gemini, and Grok. The system is designed for AI toys, companions, and edge devices, enabling them to have conversational voice interactions with low latency.