MQTT discovery allows a device or service to describe its entities to Home Assistant via MQTT. Instead of manual YAML, a configuration message is published and Home Assistant creates a sensor, binary_sensor, switch, or other entity. It is convenient for plants: GrowerHub, ESP32 or other controller can announce soil moisture, temperature, leakage and pump status.
But discovery does not replace careful design. If you do not set unique_id, device metadata, state topic, and availability, entities will be confused, duplicated, or remain “available” under an outdated retained message. For automatic watering, this is not a trifle, but the risk of making the wrong decision.
What should be in the configuration
The entity must have a stable unique_id. It should not change when the zone is renamed. We need state_topic, from where Home Assistant reads the value. The device block is useful for the device: identifiers, name, model, manufacturer or other fields. This is how several sensors of one controller are combined into one device.
For plant sensors, enter device_class and unit_of_measurement when appropriate: temperature, humidity, percentage, liters. If the value comes in JSON, use template, but keep the format stable.
Availability and expire_after
Home Assistant can show the unavailable entity if there is no fresh data or a separate availability payload has arrived. This is important for irrigation sensors. The old retained message should not look like actual humidity. The MQTT sensor has expire_after, which helps to consider the state as stale after lack of updates.
In GrowerHub the same principle should be used in the rules: if the sensor is not available, automatic watering should not start at the old value. Availability is not a decorative status, but a part of security.
Discovery and retained
Configuration discovery messages are often retained so that Home Assistant sees them after a restart. But sensor states with retain require caution: Home Assistant can get the old value instantly. For soil moisture, it is better to rely on fresh update or expire_after rather than consider the retained state to be alive.
If the device is restarted, it must re-publish discovery or respond to the birth message Home Assistant. This reduces the chance that entities will be left without configuration after a restart.
GrowerHub and Home Assistant
GrowerHub can publish discovery for its zones or read existing entities through MQTT. In both cases, the correspondence map is important: zone GrowerHub, MQTT topic, Home Assistant entity, physical device. Without this map, integration becomes a collection of similar names.
A broader scenario is described in the article GrowerHub and Home Assistant via MQTT.
Conclusion
MQTT discovery simplifies adding plant sensors to Home Assistant, but requires stable identifiers, correct topics and availability. For GrowerHub, this is a good way of integration, if we do not forget that automatic watering should make decisions only based on fresh and understandable data.