# AVA - AI-native Visual Analytics

> AVA is an AntV TypeScript framework for loading data, answering natural-language questions, and generating visualizations through an SDK, CLI, or Agent Skill.

This documentation covers AVA v4 alpha. Install `@antv/ava@alpha` with Node.js 22.13.0 or later. Use `@antv/ava` in Node.js and `@antv/ava/browser` in browsers.

The core workflow is `source()` → `analyze()` → `visualize()`. `source()` replaces the current dataset and returns no value; use `schema()` to inspect it and `profile()` for statistics. Use `suggest()` to discover questions, `translate()` and `query()` to control query execution, or `recommend()` and `viz()` to control chart generation. Always await `dispose()` when finished.

DuckDB is the default Node.js engine. JavaScript supports browsers and Node.js; Python, Supabase, and ClickHouse provide additional Node.js execution options. Follow the linked guides for source support, configuration, and complete API signatures.

## Get started

- [Introduction](https://ava.antv.vision/documentation/): Core workflow and ways to use AVA.
- [Installation and quickstart](https://ava.antv.vision/documentation/quickstart/): Install the v4 alpha, configure a model, and run a complete Node.js example.
- [Agent Skill](https://ava.antv.vision/documentation/agent-skill/): Install AVA's agent skill and analyze data with an AI coding agent.
- [Command-line interface](https://ava.antv.vision/documentation/reference/cli/): Inspect datasets, run queries, and render charts using persistent local sessions.

## Data, analysis, and visualization

- [Data sources](https://ava.antv.vision/documentation/data/sources/): Source types and parameters for inline data, files, and databases.
- [Schema and profiling](https://ava.antv.vision/documentation/data/schema/): Inspect tables and relationships and compute deterministic profiling metrics.
- [Analyze data](https://ava.antv.vision/documentation/analysis/overview/): Ask questions, get suggestions, and translate or execute queries separately.
- [Analysis strategies](https://ava.antv.vision/documentation/analysis/strategies/): Choose direct analysis, iterative reasoning, or selected profile context.
- [Generate visualizations](https://ava.antv.vision/documentation/visualization/overview/): Recommend charts and generate GPT-Vis syntax and HTML.

## Configuration and execution engines

- [Model configuration](https://ava.antv.vision/documentation/configuration/): Configure an OpenAI-compatible provider, AI Gateway, or an AI SDK model.
- [Browser usage](https://ava.antv.vision/documentation/browser/): Use the browser entry point and JavaScript engine with inline data.
- [Engine overview](https://ava.antv.vision/documentation/engines/overview/): Compare execution environments, query languages, and supported sources.
- [DuckDB](https://ava.antv.vision/documentation/engines/duckdb/): Run SQL against inline data, files, and attached databases.
- [JavaScript](https://ava.antv.vision/documentation/engines/javascript/): Execute JavaScript against an in-memory dataset.
- [Python](https://ava.antv.vision/documentation/engines/python/): Analyze with pandas using local Python or an application-provided executor.
- [Supabase and ClickHouse](https://ava.antv.vision/documentation/engines/remote/): Discover schemas and execute queries in remote databases.
- [Custom engines](https://ava.antv.vision/documentation/engines/custom/): Implement and register a custom analysis engine.

## API reference

- [AVA API](https://ava.antv.vision/documentation/reference/api/): Constructor options, method signatures, parameters, defaults, and return values.
- [Events and lifecycle](https://ava.antv.vision/documentation/reference/events/): Observe source loading, model work, query execution, and cleanup.

## Optional

- [Architecture and source map](https://ava.antv.vision/documentation/reference/architecture/): Map public capabilities to implementation modules.
- [Benchmarks](https://ava.antv.vision/documentation/reference/benchmarks/): Evaluation results and instructions for reproducing them.
- [AI Playground](https://ava.antv.vision/ai-playground/): Try browser-based visual analytics.
- [GitHub repository](https://github.com/antvis/AVA): Source code and contribution information.
- [Issues](https://github.com/antvis/AVA/issues): Bug reports and feature requests.