---

[English] | 简体中文

---

reCamera_Gimbal-OpenClaw

Use OpenClaw to control the motor, camera, LED, microphone, and speaker of a reCamera Gimbal.

Wiki Link:https://wiki.seeedstudio.com/use_cpenclaw_to_control_the_recamera_gimbal/

What This Is

This project provides an OpenClaw Skill + Node-RED flow for controlling a reCamera Gimbal edge AI camera.

It enables:

  * Motor (yaw/pitch) control via HTTP API   * Image capture and retrieval   * LED control   * Audio recording and playback   * Vision-based interaction via OpenClaw

Role in OpenClaw: Skill (with external Node-RED runtime integration)

---

Prerequisites

Important

You need the following components (derived from project files):

  * A reCamera Gimbal device (RISC-V edge AI camera)   * Node-RED running on the device (port 1880)   * OpenClaw environment with Exec tool enabled   * Network access to the device (local IP like 192.168.31.xxx)   * PowerShell (for Windows-based skill scripts)

---

Quick Start

1. Import Node-RED Flow

Import the provided file into Node-RED:

openclaw_V2.json

This creates two HTTP endpoints:

  * Control gimbal: http://<DEVICE_IP>:1880/api/gimbal?yaw=90&pitch=45   * Capture photo: http://<DEVICE_IP>:1880/api/photo

---

2. Install Skill into OpenClaw

Copy the skill folder recamera-gimbal/ into your OpenClaw workspace:

~/.openclaw/workspace/skills/recamera-gimbal/

---

3. Configure openclaw.json

The openclaw.json file is located in your OpenClaw installation directory. This file contains all the configuration settings for connecting to AI models. You need to add the following configuration for the reCamera Gimbal into openclaw.json:

Note

   * Replace "C:\\Users\\seeed\\.openclaw\\workspace\\skills" with the actual path to your skills folder.    * Replace "192.168.31.198" with the actual IP address of your reCamera Gimbal.    * Replace "recamera.1" with the actual password of your reCamera Gimbal.

"skills": {
  "load": {
    "extraDirs": [
      "C:\\Users\\seeed\\.openclaw\\workspace\\skills"
    ]
  },
  "entries": {
    "recamera-gimbal": {
      "enabled": true,
      "env": {
        "RECAMERA_IP": "192.168.31.198",
        "RECAMERA_PASS": "recamera.1"
      }
    }
  }
}

---

4. Verify

Test APIs manually:

# Move gimbal
http://<DEVICE_IP>:1880/api/gimbal?yaw=120&pitch=90

# Get image
http://<DEVICE_IP>:1880/api/photo

If successful:

  * Gimbal moves   * Image returns as JPEG

---

Configuration

HTTP API Parameters

From Node-RED flow:

Field	Type	Default	Range
yaw	number	180	1 – 345
pitch	number	90	1 – 175

Example:

/api/gimbal?yaw=120&pitch=90

---

Skill Script Paths

From SKILL.md:

# LED control
scripts/control_led.ps1 -Action on|off

# Photo capture (via HTTP)
http://<DEVICE_IP>:1880/api/photo

---

How It Works

graph TD
    Camera[Camera] --> Model[YOLO Detection]
    Model --> SaveImage[Save Image]
    Model --> Preview[Preview Rendering]

    HTTP_Gimbal[/api/gimbal/] --> Parser[Parameter Parsing]
    Parser --> Motor[Motor Control]

    HTTP_Photo[/api/photo/] --> ServePhoto[Return Photo]

    SaveImage --> Global[(Global Cache)]
    Global --> ServePhoto

Loading

Flow Summary

  * Camera captures frames   * YOLO model processes detections   * Latest image stored globally   * HTTP endpoints expose motor control and image retrieval

---

Features

  * Gimbal Control: Control yaw and pitch via HTTP API.   * Live Image Capture: Retrieve latest frame as JPEG.   * Vision Integration: YOLO-based object detection pipeline.   * LED Control: Turn light on/off via PowerShell scripts.   * Audio I/O: Record and play audio via scripts.

---

Onboarding

From SKILL.md:

Capability	Trigger	Action
Vision capture	“look”, “see”, “take photo”	Call /api/photo, analyze image
Gimbal control	directional commands	Call /api/gimbal
LED control	“turn on/off light”	Run PowerShell script
Audio	“record/play”	Run script

---

Policy

From SKILL.md:

Rule	Description
No file inspection	Do not read/edit scripts/
Exec only	Only use predefined commands
Fixed output format	Must return image in strict markdown format

---

Troubleshooting

Gimbal does not move

  * Check Node-RED is running on port 1880   * Verify device IP   * Ensure CAN/motor nodes are connected

No image returned

  * Ensure model node has debug enabled   * Check global variable latest_image is set

PowerShell scripts fail

  * Run with:   -ExecutionPolicy Bypass

HTTP API not reachable

  * Check firewall/network   * Confirm Node-RED flow is deployed

---

Links

  * OpenClaw Skill Spec: https://agentskills.io/specification#allowed-tools-field

---