DeepSeek Harness (DSH) adopts a plugin-based architecture. Developers can build agents by composing different plugins. When developing agents that involve visual interaction, the model needs to be able to recognize gestures in images. The following section introduces a model inference plugin that can convert gestures in an image into structured bounding box information.

What Is This

xby-gesture-detection is a DSH plugin maintained by xby-skill. It belongs to the model inference category, and its core function is to take an image as input, detect gestures within it, and output the bounding boxes, confidence scores, and labels for all detected objects in the image.

Core Features

The plugin provides the following main features:

  1. set_xby_apikey: Set the API key in the chat window. This action is automatically persisted and remains valid after DSH is restarted.
  2. gesture_detection: Input an image URL, detect the gestures in it, and return the bounding boxes, confidence scores, and labels.
  3. gesture_detection_for_data_base64: Input the Base64-encoded image data and perform the same gesture detection.
  4. gesture_detection_for_data_file: Input the local image file path and perform gesture detection.

Installation and Enablement

Add the plugin to the specified profile using the official installation command:

dsh plugin --profile web add github:xby_skill/xby-gesture-detection

Typical Usage

  1. Set the API key: First, register on the official xby website and obtain an API key. Then call the set_xby_apikey command in the DSH chat window to set the key.
  2. Detect an image URL: Call gesture_detection and provide the image URL.
  3. Detect a Base64 image: Call gesture_detection_for_data_base64 and provide the Base64 string.
  4. Detect a local file: Call gesture_detection_for_data_file and provide the local file path.

Notes

  • The plugin depends on an external API service. Before use, register on the official xby website and obtain an API key.
  • The plugin runs with the permissions of the current DSH process. Please check the source code and license (MIT) before installing.

Ecosystem Context

The DSH philosophy is “everything is a plugin”. This plugin is available in the community plugin directory and has no official affiliation with DeepSeek / High-Flyer.