How it works
Simulates a live smart city environment by publishing realistic telemetry, including traffic, air quality, parking, pedestrian, and transit metrics for downtown, harbor, university, and suburban districts, to retained MQTT topics every 10 seconds. A broker-native AI agent continuously monitors for natural language queries on the smartcity/ai/ask topic, actively utilizing BROKER_TOOLS at query time to inspect and read the latest retained state across the smartcity/# hierarchy. The agent grounds its response in this real-time snapshot and publishes the formatted markdown answer directly to smartcity/ai/reply in a seamless Ask/Reply flow. The system pipeline includes built-in multi-provider support, allowing rapid swapping between OpenAI, Anthropic Claude, or a locally hosted Ollama instance, while customizable simulator ranges allow operators to validate the AI pipeline before connecting physical urban sensors.