A sleepless night in a noisy city prompted a hobbyist to cobble together a DIY monitoring system. Leveraging an existing Home Assistant setup, he added two USB microphones, a Raspberry Pi, and sync’d sleep data from his Garmin watch. In roughly eight hours of AI‑assisted coding he produced a web app that visualises audio events alongside sleep stages.

The Pi records into a rolling buffer, turning on only when Home Assistant detects the user is home, in bed, and within usual sleep hours. When a sound passes a calibrated threshold, a short clip with pre‑ and post‑context is saved, tagged in JSON, and sent to an internal web server. The progressive‑web UI displays tracks like a DAW, marking red spikes where sleep stages shift.

Analyzing weeks of data revealed recurring culprits—slamming doors, high‑pitched dishware clatter, and street traffic from motorbikes and trucks. Armed with concrete evidence, the author installed acoustic panels, sealed the bedroom door and window, and discussed quieter habits with housemates. The experiment demonstrates how AI‑augmented tooling can shrink development time from weeks to a single weekend, making personalized hardware hacks practical for everyday problems.