The core philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” When an agent needs image recognition capabilities (for example, detecting people, vehicles, animals, and so on in an image), you usually need to encapsulate the model invocation logic yourself. The xby-image-detect plugin encapsulates the YOLOv6 object detection model and provides a set of standardized interfaces, making it convenient for developers to call directly.

Core Features

This plugin provides the following features:

  • set_xby_apikey: Sets the API key. This operation is automatically persisted and remains valid after restarting the DSH process.
  • yolo26_detect: Detects 80 COCO object classes in an image. An image file link must be provided as input. The output includes bounding boxes, confidence scores, and class labels.
  • yolo26_detect_for_data_base64: Detects 80 COCO object classes in an image. A Base64-encoded image file must be provided as input.
  • yolo26_detect_for_data_file: Detects 80 COCO object classes in an image. An image file path must be provided as input.

Installation

Install the plugin via the command line:

dsh plugin --profile web add github:xby_skill/xby-image-detect

Usage and Configuration

Before using it, register on the XiaoBenYang official website to obtain an API key, and then call set_xby_apikey in DSH to configure it.

Choose the appropriate detection function based on the different input data:
* If the image already has a URL, use yolo26_detect.
* If the image is Base64-encoded, use yolo26_detect_for_data_base64.
* If the image is in the local file system, use yolo26_detect_for_data_file.

Notes

  • Input requirements: All detection functions require an image file link, a Base64-encoded string, or a file path as input.
  • Development mode: The “install from a local directory” and “develop and debug via cordis.patch.yml” approaches mentioned in the GitHub README are only intended for local development and debugging scenarios.

Summary

xby-image-detect provides lightweight object detection capabilities for the DSH ecosystem, making it suitable for agent scenarios that require integrated image analysis features. The plugin source code and directory page are as follows: