Introduction

DeepSeek Harness (DSH) adopts a plugin-based architecture. When it is necessary to process text in images directly within the conversational context, users typically need to manually copy text or rely on external tools. The xby-ocr-pro plugin integrates an external OCR service and completes image content extraction and recognition directly within the chat flow.

Plugin Positioning

This is a high-accuracy text recognition plugin under the MIT license. It accepts image inputs containing text, automatically detects and recognizes the content, and is suitable for scenarios such as documents, billboards, and screenshots.

Core Features

The plugin provides the following main capabilities:

  • set_xby_apikey: Sets the API key in the chat. Once set, it is persisted automatically and remains valid after restarting the DeepSeek Harness process.
  • ocr_pro: High-accuracy text recognition. The recognition target is an image file link.
  • ocr_pro_for_data_base64: High-accuracy text recognition. The recognition target is the BASE64 encoding of an image file.
  • ocr_pro_for_data_file: High-accuracy text recognition. The recognition target is the file path of an image file.

Installation and Enable

Install the plugin via the command line:

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

After installation, the features above can be invoked from the DSH chat interface.

Usage Instructions

An API key is required before use. Users can register on the Xiaobenyang official website to obtain a key, then configure it with the set_xby_apikey command.

When invoking the recognition feature, select the corresponding command based on the input data type:
1. When using an image file link, invoke ocr_pro.
2. When using BASE64 encoding, invoke ocr_pro_for_data_base64.
3. When using a local file path, invoke ocr_pro_for_data_file.

Notes

  • Permission Dependency: The plugin runs with the permissions of the current DSH process. Ensure that the process has permission to read the input file path.
  • API Key: The API key must be configured before recognition can be performed.
  • Input Format: Different recognition commands have different format requirements for input data and must be matched strictly.

Summary

xby-ocr-pro addresses the need to directly read image content within the DSH context. For developers who need to extract structured text from visual information, this is a foundational engineering tool.

GitHub repository: https://github.com/xby_skill/xby-ocr-pro