Muse IoT

Muse-IoT is an LLM-driven IoT platform that replaces hardcoded sensor thresholds with real judgment. Sensor nodes (ESP32 over MQTT) stream temperature, soil, and humidity to an orchestrator, which assembles live readings, 48-hour history, and past reflections into one prompt. Meta Muse Spark reasons over it like a farmer — then answers in one JSON command that drives real pump, fan, and light relays. A safety engine validates every decision against hard interlocks (duration limits, cooldowns), and every action is logged with its plain-English reason, model, and latency. Problems it solves: 1. Dumb automation. Traditional IoT is if temp > 32: pump.on() — a state machine that can't see a storm coming, learn soil behavior, or explain itself. Farmers trust their eyes over it, so adoption stalls. Muse-IoT reasons instead of reacting: it held the pump off at 30.8°C to avoid damping-off and distinguished ambient heat from equipment failure on its own. 2. Unexplainable machines. Every decision carries a "why" you can interrogate in chat — "why did you water at noon?" — answered from memory with 24-hour stats. 3. Unsafe autonomy. The model proposes, but a safety engine disposes