OmniFlow Agent v1.0

// ARCHITECTURE & CORE

PydanticAI Orchestration.

Built on a modular FastAPI Orchestrator (Port 8000). The system uses the Model Context Protocol (MCP) to route tasks between isolated agents. Your local LLM (Ollama) drives a deterministic PydanticAI Agent for type-safe execution.

{
  "mode": "local",
  "model": "llama-3-8b",
  "telemetry": false,
  "offline": true
}

// IOT & HARDWARE MESH

Zero-Latency Control.

Direct hardware integration via Serial/USB or REST over WiFi. The local Python core creates a near-realtime mesh with your ESP32 nodes. No MQTT broker required—just direct, point-to-point control.