Smart Garden Features
PlantHub combines real-time sensor monitoring with weather-aware AI to automate watering, detect plant stress, and explain every decision in plain English. Here's everything your smart garden can do.
Capabilities
Skips watering when rain is on the way. Pre-empts heatwaves. Tells you why before acting.
Ask "Why are my basil leaves drooping?" and get answers backed by your sensor history, weather data, and learned patterns — in plain English.
Analyzes your sensor data and plant profile, then recommends threshold-based automation rules you can approve with one tap. No manual config needed.
Scores each actuator's effectiveness, climbs from LOW to EXPERT confidence, and factors your feedback into the next run.
Compiled from 30 seeded templates and AI recommendations. Your wifi can drop all afternoon — the garden keeps going.
Conservative, Balanced, Aggressive, Power-Saver, Manual. Override at tenant, zone, or device — inheritance sorts the rest.
Checks more often when your plant is stressed, less when it's thriving. Adapts its own interval to save energy and API costs.
Decisions, threshold warnings, and device-offline events streamed over SSE — plus push notifications straight to your phone via SNS.
Moisture, temp, light, pH ranges for tomatoes, herbs, succulents and more. Custom profiles too.
AI learning loop
Most "smart" planters guess once and repeat forever. PlantHub watches what happened, remembers similar situations, discovers what works together, and tightens its own rules.
Every action captures the sensor delta it produced — before vs. after, per actuator.
Effectiveness is scored. Confidence climbs LOW → MEDIUM → HIGH → EXPERT as samples accumulate.
Your feedback — "too early", "too long" — is batched hourly and turned into learned context.
Each evaluation is vectorized and stored. Before the next check, the AI recalls similar past situations to inform its decision.
Detects multi-actuator patterns — like watering + ventilation at 6 AM producing 2× better results than either alone.
All learned patterns — single, multi-actuator, and your preferences — are injected into the next prompt.
Safety you can prove
These aren't prompt instructions — they're enforced inside the code path that delivers every actuator command. The LLM physically can't exceed them.
Max water duration
Hard cap per command — AI cannot override.
Per-actuator cooldown
Same actuator can’t fire again until the timer expires.
Daily evaluation cap
Prevents runaway LLM spend and rapid decision churn.
Early access
We're launching soon. Join the waitlist for early-bird pricing, first access to hardware, and input on features we build next.