Introduction

DeepSeek Harness (DSH) follows the “everything is a plugin” design philosophy. When developing agents with visual perception capabilities, human pose estimation is a common requirement. The xby-pose plugin provides this capability, allowing developers to invoke pose detection services directly in conversations or scripts without setting up their own inference services.

Plugin Overview

The plugin is maintained by xby-skill and classified under the model inference category. The current version is 1.0.0, and the license is MIT.

Its core functionality is to detect people in an image and output bounding boxes and 17 keypoint coordinates. These 17 keypoints represent, in order: nose, left eye, right eye, left ear, right ear, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left hip, right hip, left knee, right knee, left ankle, and right ankle.

Core Capabilities

The plugin provides the following interfaces:

  • set_xby_apikey
    Set the API key in a chat. This setting is automatically persisted and remains valid after restarting the DSH process.

  • yolo26_pose_estimate
    Detect people in an image and output bounding boxes and keypoint coordinates.

    • Input requirement: A link to the image file.

  • yolo26_pose_estimate_for_data_base64
    Detect people in an image and output bounding boxes and keypoint coordinates.

    • Input requirement: The BASE64 encoding of the image file.

  • yolo26_pose_estimate_for_data_file
    Detect people in an image and output bounding boxes and keypoint coordinates.

    • Input requirement: The file path to the image file.

Installation and Configuration

Run the following command in the terminal to install the plugin:

dsh plugin --profile web add github:xby_skill/xby-pose

After installation, before setting the key with set_xby_apikey, you must first register on the Xiao Ben Yang official website and obtain an API key.

Usage Examples

  1. In a conversation, enter the following command to set the key:
    set_xby_apikey your_key
  2. Use yolo26_pose_estimate to detect an image:
    yolo26_pose_estimate image_url
  3. Or use yolo26_pose_estimate_for_data_base64 to detect Base64 data:
    yolo26_pose_estimate_for_data_base64 Base64_string
  4. Or use yolo26_pose_estimate_for_data_file to detect a local file:
    yolo26_pose_estimate_for_data_file /path/to/image.jpg

Notes

  • The plugin depends on the API service provided by Xiao Ben Yang.
  • Before use, ensure that the network environment allows access to the target image resources (for URL input mode).
  • The plugin runs with the permissions of the current DSH process; it is recommended to review the source code and license before installation.