Preface

The DeepSeek Harness ecosystem design emphasizes that everything is a plugin. When handling data-intensive responses, traditional text output can make it difficult to intuitively present trends and structure. dsh-antv-ava is designed to address this presentation challenge. It takes over model output and automatically transforms it into streamed narrative text, AntV charts, or a natural combination of both.

Plugin Introduction

This is a model inference plugin maintained by HellowVirgil. It leverages AntV AVA and GPT-Vis syntax to render model-generated data-intensive responses in real time as interactive charts and semantic text. The plugin is released under the MIT license.

Core Features

The plugin provides the following capabilities:

  • Automatic presentation selection: The model automatically decides whether to return plain text (NTV), a chart, or a combination of both based on the question.
  • Streamed data explanation: Metrics, changes, shares, and other data are rendered word by word.
  • Streamed AntV charts: Using GPT-Vis syntax, charts are updated in real time as data arrives.
  • Conversation-native design: It inherits DSH typography and styling, blending naturally into the conversation flow.
  • Real-time theme adaptation: Supports light/dark theme switching, with charts updating accordingly.
  • Semantic data expression: Highlights metrics, directional arrows, proportions, and more.
  • Lazy loading: Loads the rendering engine only when the response contains a chart.
  • Secure parsing: Limits source size and blocks executable markup and risky URLs.
  • AVA insights: Checks tabular results after generation to discover trends and anomalies.

Installation and Enablement

Install using the official npm package:

dsh plugin --profile web add dsh-antv-ava

After installation, restart dsh web and refresh the page. The plugin automatically injects the output contract, so general users do not need additional configuration.

Typical Usage

After installation, enter natural language requests directly in the conversation.

Explain data

Explain the following business data and provide concise, scannable conclusions (for example: total monthly orders reached 198 million, up 7.05% year-over-year…)

Explain and visualize

Analyze revenue changes from Q1 to Q4 and choose an appropriate visualization (for example: Q1 1.28M, Q2 1.64M, Q3 1.91M, Q4 2.26M)

Generate a chart only

Plot the following data as a trend chart and return only the chart

Use Cases and Notes

This plugin is suitable for evidence-backed KPIs, business analysis, trends, shares, rankings, and similar scenarios. It does not fabricate visualized values, nor does it execute arbitrary HTML, SVG, or scripts generated by the model.

This plugin is not suitable for flows, hierarchies, or roadmaps that lack a meaningful numeric structure.

Summary

Through streamed rendering and native design, this plugin lowers the barrier to data visualization. For more details, refer to the GitHub repository.