Tuya Smart Irrigation System

Evidence-based irrigation with Tuya Cloud sensors, multi-plant conflict resolution, and self-learning efficiency analysis.

Cluster Types

• With irrigator — automated irrigation + learning alerts
• Without irrigator — moisture monitoring only (manual watering reminders)

check --all auto-detects cluster type and runs the appropriate pipeline. Output is always structured ACTION: / ALERT: lines (exit code 2 = alerts to forward via Telegram).

Setup

Environment Variables (~/.openclaw/.env)

TUYA_CLIENT_ID=your_client_id
TUYA_CLIENT_SECRET=your_client_secret
TUYA_DEVICE_ID=your_irrigator_device_id
TUYA_REGION=eu          # eu | us | cn | in

Initialize Cluster

# 1. Create cluster
tuya-irrigation cluster add "My Plants" --location "Indoor" --environment indoor

# 2. Add plants
tuya-irrigation plant add --cluster 1 "Monstera deliciosa" --category tropical --water-needs medium

# 3. Add irrigator
tuya-irrigation irrigator add --cluster 1 --device-id YOUR_DEVICE_ID --name "Rainpoint"

# 4. Add sensors
tuya-irrigation sensor add --cluster 1 --device-id SENSOR_ID --name "Monstera" --type soil_moisture --plant-id 1

# 5. Set config
tuya-irrigation config set --cluster 1 --mode smart --minutes 2 --interval 12

CLI Reference

Primary (cron / automated)

tuya-irrigation check --all           # Unified check: all clusters, all alert types
tuya-irrigation check 1               # Check single cluster

Output protocol (for agent parsing):

ACTION: irrigated|skipped|monitored|error <cluster_name> [— notes]
ALERT: <cluster_name> <type>          # needs_water | learning
ALERT_ITEM: <emoji> <message>
ALERT_END

Exit codes: 0 = silent (ok/skip) · 2 = alerts to forward · 1 = error

Operations (interactive / manual)

tuya-irrigation status 1              # Full overview: sensors, config, events, smart analysis
tuya-irrigation irrigate 1            # Force irrigation pipeline: sync → weather → decide → execute
tuya-irrigation irrigate 1 --dry-run  # Analysis only, no execution
tuya-irrigation irrigate 1 --temp 22  # Override temperature (skips weather fetch)
tuya-irrigation irrigate 1 --no-sync  # Skip sensor sync (use DB data)
tuya-irrigation monitor 1             # Raw moisture check for sensor-only cluster
tuya-irrigation monitor 1 --no-sync   # Skip sync
tuya-irrigation sync --hours 6        # Cloud → DB sensor sync
tuya-irrigation learn 1               # Learning report + efficiency alerts
tuya-irrigation history 1 --hours 24  # Readings + events combined timeline
tuya-irrigation stats 1 --days 7      # Statistics + CSV export (--export file.csv)

Setup (CRUD)

tuya-irrigation cluster list
tuya-irrigation cluster add "Garden" --location "Backyard" --environment outdoor
tuya-irrigation plant list --cluster 1
tuya-irrigation plant add --cluster 1 "Ficus elastica" --category tropical --water-needs medium
tuya-irrigation plant sync                   # Sync all plants with evidence-based data
tuya-irrigation plant sync --plant-id 1      # Sync specific plant
tuya-irrigation plant sync --cluster 1       # Sync plants in cluster
tuya-irrigation irrigator list --cluster 1
tuya-irrigation irrigator start 1 --minutes 3
tuya-irrigation irrigator stop 1
tuya-irrigation irrigator log-manual 1 --minutes 5 --notes "Watered by hand"
tuya-irrigation sensor list --cluster 1
tuya-irrigation sensor add --cluster 1 --device-id XXXX --name "Nespolo" --type soil_moisture --plant-id 4
tuya-irrigation config get --cluster 1
tuya-irrigation config set --cluster 1 --mode smart --minutes 2 --interval 12

Cron Setup

Cron	Schedule	Command	Purpose
Sensor Sync	every 30min	sync --hours 24	Cloud → DB data freshness
Irrigation & Plant Check	07:00 + 20:00 (Rome)	check --all	Irrigate + monitor + alert

The Irrigation & Plant Check cron runs as an OpenClaw agent job: it executes check --all, parses ALERT_ITEM: lines (exit code 2), and forwards a single batched Telegram message to Kez.

Adding a new cluster (with or without irrigator) is automatically picked up by check --all — no cron changes needed.

Check Command: Internal Logic

Cluster with irrigator

1. Sync sensors from Tuya Cloud (last 6h)
2. Determine temperature (indoor → sensor primary; outdoor → Open-Meteo primary)
3. Run IrrigationLogic.decide_for_cluster():
   • 6h global cooldown check
   • Stress detection (water stress, heat, over-watering)
   • Multi-sensor conflict resolution (driest plant wins)
   • 48h trend analysis
   • Plant-specific targets from scientific literature
4. Execute on irrigator if action == "irrigate"
5. Collect learning alerts (drainage, efficiency)
6. Output ACTION: + any ALERT: blocks

Cluster without irrigator

1. Sync sensors from Tuya Cloud (last 2h)
2. Compare latest soil moisture vs plant targets
3. Flag sensors below threshold as needs_water
4. Output ACTION: monitored + ALERT: needs_water block if dry

Multi-Sensor Logic

With one irrigator serving multiple plants:

Scenario	Action	Duration	Confidence
All adequate (40-65%)	Skip	—	70%
One dry, none wet	Normal irrigation	2-3 min	80-90%
Conflict: one dry + one wet	Short burst	1 min	65%
All wet	Skip	—	80%

Decision uses min_soil_moisture (driest sensor), not average.

Learning Engine

After ≥3 irrigation cycles with sensor data, the system learns:

• Absorption rate: +X%/min of irrigation per plant
• Drainage rate: -X%/hr natural moisture loss
• Efficiency score: How consistently irrigation increases moisture

Alert Types

Alert	Severity	Trigger
Blocked drip	Critical	<0.5%/min absorption, <30% efficiency
Rapid drainage	Warning	>5%/hr moisture loss
Chronic underwatering	Warning	Peak moisture never reaches target (7d)
Unresolvable conflict	Critical	Irrigating dry plant would bring wet plant >85%

Confidence Scoring

Level	Source	Score
Critical stress override	Sensor + trends	95%
Sensor-driven (adequate data)	Sensor	70-90%
Temperature fallback	Open-Meteo	60%
Minimal data	Config defaults	20-30%

References

• Plant Database — Evidence-based plant care data and sources
• Database Schema — SQLite table definitions