VMind is an intelligent data visualization system that leverages Large Language Models (LLMs) to enable users to generate and edit charts through natural language conversations. It targets developers and data analysts seeking a more intuitive and efficient way to create data narratives, offering automatic chart generation from data and descriptions, as well as conversational editing capabilities.

How It Works

VMind integrates with LLM services (defaulting to OpenAI GPT) to interpret natural language prompts and translate them into VChart specifications. It processes data in CSV or JSON formats, extracting field information to understand the data structure. Users can then prompt VMind with descriptions of desired visualizations, and the system generates the corresponding chart configuration. It also supports data aggregation via LLM-generated SQL queries for common operations like grouping, filtering, and sorting.

Quick Start & Requirements

• Install dependencies: rush update
• Start demo: rush docs
• Start development: rush vmind
• Requires LLM service URL and API key (configurable via .env.local).
• Supported LLM models include various GPT versions, Doubao, Deepseek, and custom models.
• Data input formats: CSV, JSON.
• Official Docs: https://github.com/visactor/vmind

Highlighted Details

• Natural language interface for chart creation and editing.
• Supports dynamic charts and various chart types (line, bar, pie, etc.).
• Intelligent data aggregation using LLM-generated SQL (SELECT, GROUP BY, WHERE, HAVING, ORDER BY, LIMIT).
• One-click export of chart narratives as video or GIF.
• Extensible and customizable component architecture.

Maintenance & Community

• Active development indicated by CI/CD badges and npm download stats.
• Contribution guide available.
• Links to community resources are not explicitly provided in the README.

Licensing & Compatibility

• License: MIT License.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

• Dialog-based editing functionality is currently under development.
• LLM-generated SQL operations are limited to basic statements and do not support subqueries, JOINs, or complex conditional logic.