How often the full sense → capture → describe → reason → act pipeline runs.
Every hour (60) is the default. Shorter intervals mean more photos and more AI calls.
Who the AI is looking after. The care notes are injected into every reasoning prompt, so this is where you shape its preferences.
Keep it factual: light preference, watering frequency, temperature range, overwatering sensitivity.
Guardrails that protect the plant if the AI gets it wrong. The AI sees these values and reasons within them; they don't dictate when to water. To shape the AI's preferences, edit the care notes in Plant.
Consider watering when moisture drops under this.
Watering overridden OFF above this — prevents root rot.
Light forced on for warmth below this.
Watering suppressed above this to avoid heat-stress shock.
Hard cap on valve open time per watering. The Water button uses this.
Minimum time before the AI can water again.
The light won't stay on longer than this in a 24 h window.
These prompts are sent to the AI models every cycle. Changes take effect on the next cycle. Use Reset to default to restore the built-in text.
Controls the reasoning model's creativity. Lower = consistent JSON; higher = more expressive decisions. Above 1.5 may produce invalid output.
How long to wait for the vision or reasoning model before giving up and falling back to rule-based decisions.
Edit this text in the Plant tab.
Live readings straight from the hardware, refreshed every second while this tab is open. Use them to check wiring and calibrate the probes.
Zoom in, then move the box over the plant (or tap the picture where the plant is). Every screen showing the dashboard uses this framing; the AI still sees the whole photo.