Smart lighting control can greatly benefit from AI planning methods that balance energy efficiency and user comfort. We implement a robust controller that integrates a symbolic planner with an edge computing IoT stack comprising a Raspberry Pi, Z-Wave luminaires, occupancy and illuminance sensors, and a containerized pipeline for telemetry using MQTT and InfluxDB alongside visualization through Grafana. The planner reasons over environmental context such as time of day and daylight levels while respecting user agency through manual overrides and a configurable freeze window that temporarily disables automation. Evaluated against a standard baseline, our system achieves consistent energy savings without compromising required illuminance and shows low user intervention rates, indicating strong alignment with user expectations. By combining interpretable planning with real-time sensing and edge execution, the architecture delivers adaptive device control. The modular design supports straightforward extension to multi-room deployments and integration with existing building management systems. All code and configuration are open and containerized, enabling full reproducibility and serving as a foundation for future work in smart building automation.