Projects → Temperature Monitoring

Temperature Monitoring

Active

IoT System — Sweetness & Light, Ferndale CA — ESPHome / Home Assistant

Real-time temperature and humidity monitoring for a chocolate shop where ambient conditions directly affect product quality and storage. Sensors report to Home Assistant, a custom dashboard tracks readings over time, and alerts fire when conditions go out of range. Also the proof-of-concept for a broader commercial SaaS idea.

ESPHome Home Assistant ESP32 Python MQTT

Chocolate is finicky. Temperature swings cause bloom — the whitish haze you get when cocoa butter migrates to the surface. Humidity affects texture and shelf life. The shop has a controlled storage area, but knowing when conditions drift and catching it before product is affected requires monitoring, not guesswork.

Off-the-shelf solutions (commercial loggers, cloud-connected sensors) exist but charge monthly for data you should own. Building it on ESPHome and Home Assistant took an afternoon and costs nothing to run.

ESP32 boards running ESPHome firmware report temperature and humidity over WiFi to Home Assistant via MQTT. HA handles the dashboard, historical charting, and alert automations — push notifications when readings leave defined thresholds.

The live dashboard is hosted separately from tostecode.dev (uptime requirements are different for a monitoring system vs a portfolio site). The home/shop WireGuard VPN connects both locations so the shop sensors report to the home-based HA instance.


The personal version works well enough to imagine building a commercial version for other small businesses that need this kind of monitoring — bakeries, wine shops, restaurants, dispensaries — without paying enterprise prices or surrendering their data to a cloud platform.

The concept (ColdWatch) would use SwitchBot BLE sensors for easier deployment, Raspberry Pi Zero 2 W edge devices to aggregate readings locally, WireGuard tunnels to report to a central dashboard, and a customer web interface for alerts and historical data. The architecture is straightforward. The question is whether the market wants it badly enough.