ESP32 + AI

ESP32 + AI: intelligent
plant care that learns.

PlantHub doesn't just read sensors — it understands what the data means. ESP32-C3 sensor nodes feed real-time soil, climate, and light data to an AI engine that knows 50+ plant species, checks the weather forecast, and explains every decision in plain English. The result: garden automation that thinks, learns, and runs offline.

The AI + ESP32 stack explained

Most garden automation systems are just timers with a sensor attached. PlantHub is fundamentally different: it pairs ESP32 sensor hardware with a cloud-based AI engine that makes contextual decisions about plant care.

Here's how the stack works: ESP32-C3 sensor nodes collect environmental data (soil moisture, temperature, humidity, light) and publish it over MQTT with mutual TLS encryption to AWS IoT Core. PlantHub's AI engine — built on Spring AI — receives this data in real time and evaluates it against three knowledge sources:

  • Plant species profiles — 50+ built-in profiles with optimal moisture ranges, temperature tolerances, light requirements, and seasonal adjustments
  • Weather forecasts — 28-hour predictions including rain probability, temperature extremes, and humidity trends
  • Historical sensor data — Stored in TimescaleDB for time-series analysis, enabling trend detection and anomaly identification

The AI produces care decisions with full reasoning: "Skipping watering because rain is expected tomorrow at 70% probability. Soil moisture at 42% is sufficient for basil until then." These decisions are logged for audit and compiled into lightweight rules that execute directly on the ESP32 node.

AI Modes

Five AI modes for every growing style

Conservative

Waters only when soil moisture drops well below the plant's threshold. Ideal for drought-tolerant species like succulents, cacti, and Mediterranean herbs.

Balanced

The default mode. Maintains optimal moisture levels based on species profiles, weather data, and time of day. Best for most home gardens.

Aggressive

Proactive care for water-hungry plants like tomatoes, tropical ferns, and seedlings. Waters earlier and more frequently, checks sensors more often.

Power-Saver

Extends battery life by reducing sensor check frequency and using simplified rules. Perfect for solar-powered or remote sensor node deployments.

Manual

Full human control. AI provides recommendations and reasoning but takes no automated action. You decide when and how much to water.

AI modes can be set at the tenant level, overridden per zone, or customized per individual device. A greenhouse operator might run Aggressive mode on seedling trays and Power-Saver on established perennials — all managed from a single dashboard.

The AI learning loop

PlantHub's AI doesn't just follow static rules — it continuously improves through a six-step learning loop:

1. ObserveCollect sensor readings — soil moisture, temperature, humidity, light — at configurable intervals.
2. RateCross-reference readings against the plant's species profile, weather forecast, and historical baselines.
3. ListenFactor in your feedback — if you override an AI decision, the system notes your preference.
4. RememberBuild a history of conditions, decisions, and outcomes for each plant and zone.
5. CombineMerge sensor data, weather, species knowledge, and your preferences into a unified care model.
6. AdaptRefine watering schedules, alert thresholds, and check frequencies based on accumulated experience.

This loop runs continuously for each device. Over time, PlantHub develops a nuanced understanding of your specific growing conditions — microclimate, soil type, watering patterns — and tailors its recommendations accordingly.

Offline AI: compiled rules on ESP32

Cloud-dependent garden automation has a fatal flaw: when your internet drops, your plants stop getting care. PlantHub solves this by compiling AI decisions into lightweight rules that run directly on the ESP32-C3.

When the AI engine generates a care schedule — "water Zone 2 for 8 seconds at 6 AM, skip if soil moisture is above 55%" — it compiles this into a compact rule that's pushed to the ESP32 node over MQTT. The rule executes locally on the microcontroller's 160 MHz RISC-V processor, no cloud connection required.

This means your ESP32 sensor nodes are not just data collectors — they're autonomous care agents that keep your garden alive during internet outages, router reboots, or cloud service interruptions. When connectivity returns, the node syncs its actions back to PlantHub for logging and AI refinement.

AI safety: limits the AI cannot override

PlantHub enforces hard safety limits in code — not in AI prompts. These constraints protect your plants and property regardless of what the AI recommends:

  • Maximum watering duration — Caps the longest single watering session to prevent flooding
  • Cooldown periods — Enforces minimum time between watering events
  • Daily water caps — Limits total water usage per 24-hour period per device
  • Threshold guards — AI cannot set thresholds outside species-safe ranges

Every AI decision is logged with its full reasoning chain. You can review exactly why the AI chose to water (or not water), what data it considered, and which safety limits were active. Full transparency, full control.

Chat with your garden AI

PlantHub includes a natural-language AI chat that lets you ask questions about your plants and get answers backed by real sensor data:

You: "Why are my basil leaves drooping?"

PlantHub AI: "Your basil's soil moisture has been below 30% for 14 hours. The current reading is 22% at 28°C. I recommend watering for 10 seconds now — the forecast shows no rain for the next 48 hours, so the soil won't recover on its own."

The AI chat isn't a generic chatbot — it has access to your live sensor data, historical trends, active rules, and plant species profiles. Every answer is grounded in real measurements from your ESP32 sensor nodes.

PlantHub brings AI intelligence to ESP32 garden hardware — making smart plant care accessible to everyone. Explore the full system:

Early access

Be first
in line.

We're launching soon. Join the waitlist for early-bird pricing, first access to hardware, and input on features we build next.

30% early-bird discount First access to hardware Shape the roadmap