Tuya Smart Home Device Control Skill

Basic Information

Official Website: https://www.tuya.com/
Source Code: https://github.com/tuya/tuya-openclaw-skills
Authentication: Via Header Authorization: Bearer {Api-key}
Credentials: Read from environment variable TUYA_API_KEY. Base URL is auto-detected from API key prefix. See references/api-conventions.md for the prefix-to-region mapping table. You can override by setting TUYA_BASE_URL.
API Reference: See individual files under references/
Python SDK: See scripts/tuya_api.py
Device Message Client: See scripts/tuya_device_mq_client.py (real-time WebSocket subscription)

Environment Variable Configuration

Set the following environment variable before use:

export TUYA_API_KEY="your-tuya-api-key"
# TUYA_BASE_URL is optional — auto-detected from API key prefix
# Override only if needed: export TUYA_BASE_URL="https://openapi.tuyaus.com"

The same TUYA_API_KEY is used for both the REST API and WebSocket message subscription. The WebSocket URI is auto-detected from the API key prefix (same 7 data centers as the REST API). See references/device-message.md for the full mapping table.

The skill will not load if the TUYA_API_KEY environment variable is missing.

Usage

Always prefer Method 1 (Command Line) — single command, no boilerplate code. It handles authentication, URL resolution, JSON serialization, and error handling automatically.

Method 1: Via Command Line (Recommended)

python3 {baseDir}/scripts/tuya_api.py <command> [params...]
# Examples:
python3 {baseDir}/scripts/tuya_api.py homes
python3 {baseDir}/scripts/tuya_api.py devices
python3 {baseDir}/scripts/tuya_api.py devices --home 5053559
python3 {baseDir}/scripts/tuya_api.py devices --room 123456
python3 {baseDir}/scripts/tuya_api.py device_detail <device_id>
python3 {baseDir}/scripts/tuya_api.py model <device_id>
python3 {baseDir}/scripts/tuya_api.py control <device_id> '{"switch_led":true}'
python3 {baseDir}/scripts/tuya_api.py rename <device_id> "New Name"
python3 {baseDir}/scripts/tuya_api.py weather 39.90 116.40
python3 {baseDir}/scripts/tuya_api.py sms "Your message"
python3 {baseDir}/scripts/tuya_api.py voice "Your message"
python3 {baseDir}/scripts/tuya_api.py mail "Subject" "Content"
python3 {baseDir}/scripts/tuya_api.py push "Subject" "Content"
python3 {baseDir}/scripts/tuya_api.py stats_config
python3 {baseDir}/scripts/tuya_api.py stats_data <dev_id> <dp_code> <type> <start> <end>
python3 {baseDir}/scripts/tuya_api.py ipc_pic_fetch <device_id> <consent> [pic_count] [home_id]
python3 {baseDir}/scripts/tuya_api.py ipc_video_fetch <device_id> <duration> <consent> [home_id]

CLI validation rules:

devices supports only one scope flag at a time: --home <id> or --room <id>
control requires properties_json to be a valid JSON object (not array/string)
weather validates coordinate range: latitude [-90, 90], longitude [-180, 180]
stats_data validates start/end format yyyyMMddHH and max 24-hour window
ipc_pic_fetch args: <device_id> <consent> [pic_count] [home_id] — consent 1 = decrypted URL
ipc_video_fetch args: <device_id> <duration> <consent> [home_id] — duration in seconds (1-60)
Use python3 {baseDir}/scripts/tuya_api.py --help for command help and examples

Method 2: Via Python SDK

Use when you need to chain multiple API calls or do complex logic in a single script:

import sys
sys.path.insert(0, "{baseDir}/scripts")
from tuya_api import TuyaAPI

api = TuyaAPI()

homes = api.get_homes()
devices = api.get_all_devices()
detail = api.get_device_detail("device_id_here")
result = api.issue_properties("device_id_here", {"switch_led": True, "bright_value": 500})
weather = api.get_weather(lat="39.90", lon="116.40")
# IPC cloud capture — take a snapshot and get decrypted URL
capture = api.ipc_ai_capture_pic_allocate_and_fetch("device_id_here")

Method 3: Device Message Subscription (WebSocket)

Use when you need real-time device event monitoring (property changes, online/offline status):

import asyncio
import os
import sys
sys.path.insert(0, "{baseDir}/scripts")
from tuya_device_mq_client import TuyaDeviceMQClient

async def main():
    # Uses TUYA_API_KEY for auth; WebSocket URI auto-detected from key prefix
    client = TuyaDeviceMQClient(
        api_key=os.environ["TUYA_API_KEY"],
        device_ids=None,  # None = all devices; or pass a list of device IDs
    )

    @client.on_property_change
    async def on_prop(device_id, properties):
        for prop in properties:
            t = TuyaDeviceMQClient.format_timestamp(prop["time"])
            print(f"[{t}] Device {device_id}: {prop['code']} = {prop['value']}")

    @client.on_online_status
    async def on_status(device_id, status, timestamp_ms):
        t = TuyaDeviceMQClient.format_timestamp(timestamp_ms)
        print(f"[{t}] Device {device_id} is now {status}")

    await client.connect()

asyncio.run(main())

Important: The WebSocket client runs server-side only. It reuses the same TUYA_API_KEY — no separate credentials needed. The WebSocket URI is auto-detected from the key prefix (same 7 data centers as the REST API). Notification throttling (minimum 30-minute cooldown) is mandatory when triggering notifications from device events. See references/device-message.md for message format details and more examples.

Feature Overview

| Module | Capabilities | Reference | |--------|-------------|-----------| | Home Management | List all homes, list rooms in a home | references/home-and-space.md | | Device Query | All devices, devices by home/room, single device detail (including current property states) | references/device-query.md | | Device Control | Query device Thing Model, issue property commands | references/device-control.md | | Device Management | Rename device | references/device-management.md | | Weather Service | Current and forecast weather | references/weather.md | | Notifications | SMS, voice call, email, App push | references/notifications.md | | Data Statistics | Hourly statistics config query, statistics value query | references/statistics.md | | IPC Cloud Capture | Cloud snapshot and short video capture for IPC cameras | references/ipc-cloud-capture.md | | Device Message Subscription | Real-time WebSocket subscription for device property changes and online/offline events | references/device-message.md | | Error Handling | Error codes and recovery strategies | references/error-handling.md | | API Conventions | Request/response format, data center mapping | references/api-conventions.md |

Core Workflows

Workflow 1: Device Control

When the user says things like "turn on the living room light" or "set the AC temperature to 26 degrees":

Locate the device — Find the target device based on the device name or location mentioned by the user. Follow this priority:
- Priority 1 — Room + category match: If the user mentions a room (e.g. "living room AC"), first query the home list → room list to match the room, then list devices in that room and match by category_name or device name
- Priority 2 — Device name match: If the user only mentions a device name (e.g. "AC"), call "List All Devices" API and match by category_name first, then by device name fuzzy match
- Priority 3 — Disambiguation: If multiple devices match, list all candidates with their room information and ask the user to choose
Get current state — Call the "Get Single Device Detail" API
- If result is null: the device does not exist or you have no permission — inform the user and stop
- If online is false: the device is offline — tell the user "Device XX is currently offline, please check its power and network connection" and do not proceed further
- Only continue when result is valid and online is true
- The properties field contains current values of each functional property (e.g. switch state, brightness, temperature)
Query capabilities — Call the "Query Device Thing Model" API to get the device's supported property list
- Important: The result.model field is a JSON string that must be parsed again (e.g. json.loads(result["model"])) to obtain the property definitions
- Check each property's accessMode:
  - ro (read-only): cannot be controlled, only queried — inform the user "this property is read-only"
  - wr (write-only): can be controlled but current value cannot be read
  - rw (read-write): can be both controlled and queried
Map the command — Map the user's intent to Thing Model properties:
- Turn on/off → find a bool-type switch property (e.g. switch_led, switch)
- Adjust brightness → find a value-type brightness property
- Adjust temperature → find a value-type temp property
- If the device does not support the requested function, inform the user and list supported functions
- Relative adjustments — When the user says "a bit brighter", "lower the temperature by 2 degrees", etc.:
  1. Read the current value from properties in the device detail (Step 2)
  2. Read min, max, step from the Thing Model typeSpec (Step 3)
  3. Calculate the target value:
    - Vague ("a bit", "a little") → current value ± (max - min) × 10%
    - Specific ("by 2 degrees", "by 100") → current value ± the specified amount
  4. Clamp the target value within [min, max] and round to the nearest step
- Validate value range: Before issuing, confirm the target value is within the typeSpec min/max range
Issue the command — Call the "Issue Properties" API using the Python SDK: api.issue_properties(device_id, {property_code: value})
- The SDK handles properties JSON string serialization automatically
- If not using the SDK: the properties field must be a JSON string, not a JSON object. You must double-serialize: {"properties": "{\"switch_led\":true}"}
Verify and return result — After issuing the command:
- Wait 1-2 seconds, then call "Get Device Detail" again to read the updated properties
- Compare the target value with the actual value to confirm execution
- If values match: inform the user the operation succeeded
- If values differ: tell the user "command sent, but the device state has not updated yet — there may be a delay"

Workflow 2: Rename Device

Locate the device using Workflow 1 Step 1 to obtain the device_id
Call the "Rename Device" API with the new name
Return the result

Workflow 3: Notifications

Identify the message type: SMS / Voice / Email / App Push
Extract required parameters (message content; email and push also need a subject)
Call the corresponding API (all notification APIs are self-send — messages can only be sent to the current logged-in user)
Return the send result

Workflow 4: Weather Query

Obtain coordinates:
- First call Home List API and check the latitude / longitude fields
- Note: the coordinate format is {"Value": "30.3"} — you must extract the .Value field (e.g. home["latitude"]["Value"])
- If the home has no location set, ask the user for their city and convert to coordinates (see common city coordinates in references/weather.md)
Determine which weather attributes to query (default: temperature, humidity, weather condition)
Call the weather query API
Translate the returned data into a human-readable description

Workflow 5: Data Statistics

Locate the device (same as Workflow 1 Step 1)
Call the "Statistics Config Query" API to confirm whether the device has the corresponding statistics capability
If available, call the "Statistics Value Query" API
- Time inference: Convert the user's natural language to yyyyMMddHH format:
  - "today" → start = today 00:00, end = current hour
  - "yesterday" → start = yesterday 00:00, end = yesterday 23:00
  - The time range cannot exceed 24 hours per request — for longer ranges, make multiple requests and aggregate
  - Format example: 2024010100 = January 1, 2024 00:00
Aggregate and return the results

Workflow 6: Device Status Query

When the user asks "Is the living room light on?" or "What's the AC set to?":

Locate the device and get current state (same as Workflow 1 Steps 1-2; stop if device not found or offline)
Read properties values, cross-reference with the Thing Model property names/descriptions, and translate to natural language (e.g. "switch_led": true → "the light is currently on")

Workflow 7: Multi-Device Batch Control

When the user says "Turn off all lights" or "Set all ACs to 26 degrees":

Call "List All Devices" API and filter matching devices by category_name or device name keyword
For each matching device: check online status (skip offline devices and note them), then execute Workflow 1 Steps 3-6
Aggregate results: report how many devices succeeded, which ones failed or were offline
Add a brief delay (0.5-1s) between requests to avoid rate limiting

Workflow 8: IPC Cloud Capture

When the user asks to "take a photo with the camera" or "record a short video from the camera":

Locate the IPC device — same as Workflow 1 Step 1, filter by camera category
Determine capture type:
- Snapshot → PIC (optional pic_count, 1-5)
- Short video → VIDEO (optional video_duration_seconds, 1-60, default 10)
Execute capture — Use the all-in-one helper methods:
- For PIC: api.ipc_ai_capture_pic_allocate_and_fetch(device_id, pic_count=1)
- For VIDEO: api.ipc_ai_capture_video_allocate_and_fetch(device_id, video_duration_seconds=5)
- These methods handle the full allocate → wait → poll → retry flow automatically
Return the result — Extract the URL from the resolve result:
- PIC with consent: resolve["decrypt_image_url"]
- VIDEO with consent: resolve["decrypt_video_url"] (cover image may be null if still uploading)
- If status is still NOT_READY after all retries, inform the user that the device may be slow to upload and suggest trying again later

Workflow 9: IPC Visual Recognition

When the user asks "What's in front of my camera?", "Is there anyone at the door?", or "Describe what the camera sees":

Capture a snapshot — Follow Workflow 8 Steps 1-3 to take a PIC capture
Get the image URL — Extract resolve["decrypt_image_url"] from the capture result. If the resolve failed or returned NOT_READY, inform the user and stop
Download the image — Fetch the image content from the decrypted URL
Send to AI vision model — Pass the image to the AI large model for visual understanding. Describe the image content in natural language based on the user's question:
- General question ("What's there?") → describe the overall scene, objects, and people
- Specific question ("Is there a package?", "Is anyone at the door?") → focus on answering the specific question
Return the description — Respond to the user with the visual analysis result in conversational language

Workflow 10: Real-Time Device Monitoring

When the user asks to "monitor device changes in real time", "watch for property updates", or "notify me when a device goes offline":

Determine scope — Ask which devices to monitor (all or specific device IDs). If specific, locate devices using Workflow 1 Step 1
Determine event types — Property changes (on_property_change), online/offline status (on_online_status), or both
Write the subscription script — Using TuyaDeviceMQClient from scripts/tuya_device_mq_client.py:
- Import and instantiate with api_key=os.environ["TUYA_API_KEY"] (WebSocket URI auto-detected from key prefix)
- Register appropriate handlers using decorators
- Call await client.connect() to start listening
Apply throttling — If the subscription triggers notifications or device control actions, implement a cooldown mechanism (minimum 30-minute interval for notifications)
Cross-reference with REST API — Property codes from WebSocket events correspond to Thing Model codes. Use api.get_device_model(device_id) to look up property names and value ranges when needed

Workflow 11: Event-Driven Automation

When the user asks to "turn on the hallway light when the door opens" or "send me a notification when the AC turns off":

Identify trigger and action — Parse the trigger device, trigger condition (property code + value), and the action to execute
Locate devices — Use Workflow 1 Step 1 to find both the trigger device and the action device
Write the automation script — Combine TuyaDeviceMQClient for event listening with TuyaAPI for device control:
- Subscribe to the trigger device's property changes
- When the trigger condition is met, call api.issue_properties() to control the action device
- Implement notification throttling (30-minute cooldown) if sending notifications
Verify — Confirm the trigger condition and action mapping with the user before running

Important Notes

Device name matching uses fuzzy matching; when multiple results are found, ask the user to confirm
The statistics API time format is yyyyMMddHH, and the time range cannot exceed 24 hours per request
All four notification APIs are self-send only — messages can only be sent to the currently logged-in user
The weather query requires latitude and longitude; if unavailable from the Home API, ask for the user's city
Base URL is auto-detected from API key prefix. See references/api-conventions.md for details
If you encounter issues, visit https://github.com/tuya/tuya-openclaw-skills for announcements and troubleshooting
Never log or display the TUYA_API_KEY value in output
CLI exits with code 2 for usage/validation errors, and 1 for runtime/API/network errors

Supported and Unsupported Operations

Supported Property Types for Control

Only basic data type properties are currently supported for device control:

| Type | Description | Example | |------|-------------|---------| | bool | Boolean on/off | Turn light on/off, turn AC on/off, turn plug on/off | | enum | Enumeration selection | Switch AC mode (auto/cold/hot), set fan speed (low/mid/high) | | value (Integer) | Numeric value | Adjust brightness (0-1000), set temperature (16-30) | | string | String value | Set device display text |

Unsupported Operations

The following operations involve sensitive actions or complex data types and are NOT supported:

Lock control — Unlock doors, lock/unlock smart locks (security-sensitive)
Live video streaming — Pull real-time video streams or view camera live footage (cloud snapshot/short video capture IS supported — see Workflow 8)
Image operations — Retrieve or push images from/to devices
Complex data type control — Properties with raw, bitmap, struct, or array typeSpec are not supported for issuing commands
Firmware upgrades — OTA firmware update operations
Device pairing/removal — Adding new devices or removing existing devices

If the user requests any of these unsupported operations, clearly inform them that the operation is not available through this skill and suggest using the Tuya App directly.

Data Egress Statement

This skill sends data to the Tuya Open Platform:

| Data Type | Sent To | Purpose | Required | |-----------|---------|---------|----------| | Api-key | User-configured base_url | API authentication | Required | | Device ID | User-configured base_url | Device query and control | Required | | Control commands | User-configured base_url | Device property issuance | Required | | Api-key | Auto-detected WebSocket URI | Real-time event subscription authentication | Required for message subscription |