Introduction

When processing customer-provided CSV, XLSX, or JSON enterprise lists in DeepSeek Harness, you usually need to perform data cleaning first (trimming spaces, normalizing phone numbers, deduplicating) and enrichment. dsh-data-cleaning-agent aims to provide a localized data cleaning and enrichment solution, with support for further enterprise data expansion through Qichacha MCP.

Plugin Overview

This plugin is maintained by dhu2000 and released under the MIT License. It targets enterprise list data and provides complete capabilities from cleaning, enrichment, to profile analysis. It also supports exporting results as CSV/XLSX files.

Core Features

The plugin includes the following core capabilities:

  • Data Cleaning: Supports trimming column contents (removing leading and trailing spaces), normalizing phone numbers, and rule-based filtering (such as missing required values, negative amounts, duplicate rows).
  • Data Enrichment: Supports filling in null values or reporting items that cannot be enriched.
  • Profile Analysis: Provides column overview statistics and amount distribution analysis to help quickly understand data quality.
  • Qichacha MCP Enrichment: Supports integrating the Qichacha data source to enrich enterprise industrial and commercial profiles, equity penetration, and other information. Note: The enrichment feature uses customer-provided connection and account (BYO); no built-in API key is included.
  • Format Parsing: Supports parsing data in CSV, XLSX, and JSON formats.
  • Asynchronous Task Management: Built-in task state machine and persistent storage support long-running task management.
  • Image List Parsing: Supports parsing enterprise lists from images, with native thumbnails and zoomed viewing.
  • Five-Step Workflow: Provides a standardized task flow: “Import → Rules → Matching → Enrichment → Download.”

Installation and Enablement

The installation command is as follows:

dsh plugin --profile web add dsh-data-cleaning-agent@0.9.17

After installation, you must completely restart DeepSeek Harness to ensure the new features take effect.

Typical Usage

After installation and restart, a “Data Cleaning & Enrichment” entry appears at the top of the sidebar.

  1. Enter a command in the conversation, for example: “Help me clean this enterprise list data.”
  2. The plugin loads the embedded Skill and guides the user to upload CSV/XLSX/JSON files.
  3. Select cleaning rules (such as deduplication) and enrichment fields in the workbench or below the input field.
  4. The system runs an asynchronous task. After completion, you can preview and download the cleaned and enriched results (XLSX) and the exception list (XLSX).

Notes

  • Better Sidebar Dependency: Better Sidebar is an optional workbench dependency. If it is missing, basic tools (such as data_clean_rows, data_profile) remain available, but the interactive workbench (table import, mapping confirmation, result preview) is not available.
  • Qichacha Connection: The Qichacha enrichment feature does not include a built-in API key. Users must configure the connection and account in DSH themselves.
  • Image Parsing Configuration: Local image parsing requires separate configuration of the official qcc-document-mcp.
  • Version Upgrade: After upgrading the plugin, you must completely restart DeepSeek Harness. Migration of old sessions is not covered in the acceptance scope of this version.

Summary

By combining a local engine with Qichacha MCP, this plugin provides DeepSeek Harness with a closed-loop enterprise list data processing solution. Developers can leverage its cleaning and enrichment capabilities to quickly process customer data and export standardized result files.