The core idea of DeepSeek Harness (DSH) is “everything is a plugin.” When building agents with visual perception capabilities, calling third-party model inference services is a common approach. xby-pedestrian is a model inference plugin for DSH, used to perform pedestrian detection on images in a chat context and output bounding boxes, confidence scores, and labels.

Core Features

The plugin provides the following core capabilities:

  • set_xby_apikey: Sets the API key in chat. The key is automatically persisted and remains valid after restarting the DSH process.
  • detect_pedestrian: Detects pedestrians in an image. Requires the URL of an image file. Outputs the bounding boxes, confidence scores, and labels for all detected targets.
  • detect_pedestrian_for_data_base64: Detects pedestrians in an image. Requires the BASE64 encoding of an image file. Outputs the bounding boxes, confidence scores, and labels for all detected targets.
  • detect_pedestrian_for_data_file: Detects pedestrians in an image. Requires the file path of an image file. Outputs the bounding boxes, confidence scores, and labels for all detected targets.

Installation

Use the DSH CLI tool to install from the GitHub repository:

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

Configuration and Usage

Before using the plugin features, you must obtain an API key.

  1. Visit the Xiaobenyang official website to register an account and obtain an API key.
  2. Call the set_xby_apikey command in chat to set the key.

After obtaining the key, call the appropriate detection interface based on the type of input data:

  • Use an image URL:
    Call detect_pedestrian and pass the image file URL.
    detect_pedestrian https://example.com/image.jpg

  • Use Base64 encoding:
    Call detect_pedestrian_for_data_base64 and pass the image file BASE64 encoding.

  • Use a file path:
    Call detect_pedestrian_for_data_file and pass the image file path.

All interfaces return the bounding boxes, confidence scores, and labels for all targets in the image.

Notes

  • The plugin relies on the Xiaobenyang API service.
  • Ensure the API key is correctly configured before use.
  • Choose the correct interface function based on the input data source (URL, Base64, file path).

Reference