Preface

When building agents based on DeepSeek Harness (DSH), processing unstructured data is a common requirement. When input sources contain handwritten text (such as handwritten notes, signatures, or forms), general-purpose OCR models often struggle to handle them. Through the DSH plugin system, handwriting recognition capabilities can be integrated directly into workflows, avoiding reinventing the wheel.

Plugin Overview

xby-ocr-handwriting is a model inference plugin that runs in the DSH environment. It is maintained by xby-skill and provides the ability to process images containing handwritten text by automatically detecting text lines and recognizing their content. The plugin is open source under the MIT license.

Features

The plugin provides the following core capabilities:

  • set_xby_apikey: Sets the API key in the conversation. The key is automatically persisted and remains valid after a process restart.
  • ocr_handwriting: Recognizes handwritten text. Requires a link to the image file as input.
  • ocr_handwriting_for_data_base64: Recognizes handwritten text. Requires a BASE64 encoding of the image file as input.
  • ocr_handwriting_for_data_file: Recognizes handwritten text. Requires a file path to the image file as input.

Installation and Configuration

Installation

Add the plugin to DSH using the following command:

dsh plugin --profile web add github:xby_skill/xby-ocr-handwriting

API Key Configuration

Before using the plugin, you must register on the official XiaoBenYang website and obtain an API key. Once obtained, configure it using the set_xby_apikey command.

Typical Usage

Choose the corresponding interface based on the form of the input data:

  1. Network image link: Call ocr_handwriting and pass the image URL.
  2. BASE64 encoding: Call ocr_handwriting_for_data_base64 and pass the encoded string.
  3. Local file path: Call ocr_handwriting_for_data_file and pass the local file path.

Notes

  • API dependency: This plugin depends on the API service provided by XiaoBenYang. You must register and obtain an API key before use.
  • Permission scope: The plugin runs with the permissions of the current DSH process. When calling the file-path interface, ensure that the process has read permissions.
  • Security check: Before installing and using it, it is recommended to review the GitHub repository source code and license to confirm that they meet project requirements.

Summary

xby-ocr-handwriting provides DSH users with a standardized handwriting recognition solution, supporting three input methods: link, Base64, and file path. For more technical details, refer to the plugin directory or the GitHub repository.