Skip to content
AVADocumentation
Browse documentation
Get started

Introduction to AVA

AI-native visual analytics, from question to insight.

AVA is a TypeScript framework for asking questions about data and turning the answers into visualizations. Use it through an Agent Skill, the JavaScript API, or the command line, with an execution engine that fits your environment.

Use with your agentInstall the Agent Skill and try a complete data analysis demo.Start buildingInstall AVA and run your first analysis in Node.js.Build for the browserBring analytics into your frontend with the JavaScript engine.Use the command lineInspect datasets, execute queries, and render charts from your terminal.

The core workflow

Connect a data source, inspect its structure, ask a question, and optionally generate a chart.

SourceAnalyzeVisualize
  • Discover the structure. schema() reads tables and fields automatically; optionally use profile() for statistics and semantic() to add your business definitions. See Inspect your data.
  • Ask better questions. suggest() recommends questions using the schema and available profile.
  • Analyze in natural language. analyze() translates, executes, and summarizes a question using the selected strategy.
  • Control individual steps. Use translate() and query() separately, or recommend() and viz() for chart generation.

Choose an execution engine

EngineQuery languageEnvironment
DuckDBSQLNode.js; default
JavaScriptJavaScriptBrowser and Node.js
PythonPython / pandasNode.js with a Python executor
SupabasePostgreSQL SQLNode.js; remote execution
ClickHouseClickHouse SQLNode.js; remote execution

The engine guide explains supported sources and configuration.

Explore the documentation

Follow the workflow from your first dataset to a visual answer, then explore more control when you need it.

  1. Install and run AVA

    Set up your model and complete your first analysis with the quickstart.

  2. Connect and understand your data

    Load files or databases, then inspect schema and profiling results.

  3. Ask questions and analyze

    Discover useful questions, get answers, and choose an analysis strategy.

  4. Turn answers into charts

    Generate a visualization or control chart recommendation and rendering separately.

  5. Configure and extend

    Adjust your model setup, choose an execution engine, or implement a custom engine.

Keep the API Reference handy for parameters. The architecture guide maps public capabilities to their source modules.

Resources

Edit this page on GitHub