GPT-VisGPT-Vis is an open-source visualization library for AI agents and applications. It renders 26 types of visualization from a small JSON configuration or readable Syntax, including statistical charts, relationship graphs, tables, and text summaries.
Choose a starting point
- Use an agent: install the chart-visualization Skill to choose charts, generate configurations, and create reusable themes.
- Build an application: follow the SDK quickstart to render your first chart in JavaScript or TypeScript.
- Explore the output: browse chart examples or switch themes in the homepage gallery.
One theme across your visualizations
The theme option accepts shared colors, typography, and supported style settings. Use the same object for statistical charts, relationship graphs, tables, and summaries. Explore the theme gallery, download its JSON, or ask your agent to create a theme from your brand references.
Custom theme objects are available from @antv/gpt-vis 1.3.0. Use gallery presets by their built-in names, such as pop or midnight. For a customized style, pass the complete JSON object. See themes and supported names.
Model output, rendered as it arrives
GPT-Vis accepts both JSON and Syntax. For streaming responses, pass the accumulated Syntax to the same instance as more text arrives. See streaming rendering.
What GPT-Vis handles
Your application supplies the data and analysis results. GPT-Vis handles rendering, chart defaults, visual styling, and supported chart interactions. Querying databases, forecasting, and explaining business causes belong in the agent or application that uses it.
Next steps
Install the Agent Skill, start with the SDK, or browse all chart types.