Two independent nodes communicate over encrypted WebSocket channels. An AI coordinator manages routing, retries, and protocol negotiation automatically.
Local LLM handles intent classification, message routing decisions, anomaly detection, and adaptive protocol selection — all offline, no cloud.
Real-time voice synthesis and recognition across nodes. Whisper-based STT, Kokoro TTS, sub-200ms end-to-end latency on local hardware.
Python FastAPI WebSocket Whisper Kokoro TTS Ollama asyncio
┌─────────────────┐ ┌─────────────────┐
│ NODE A │◄───────►│ NODE B │
│ │ WS/TLS │ │
│ AI Coordinator │ │ AI Coordinator │
│ Voice Engine │ │ Voice Engine │
│ FastAPI :8100 │ │ FastAPI :8100 │
└─────────────────┘ └─────────────────┘
▲ ▲
│ │
Client A Client B
(any device) (any device)