AVA ChartGenie LogoAVA, AI-Native Visual Analytics
AVA Logo

AVA

AI-Native Visual Analytics

A technology framework designed for more convenient visual analytics, powered by AI

💬

Natural Language

Ask questions about your data in plain English

🤖

LLM-Powered

Leverages large language models for intelligent analysis

⚡

Smart Processing

Auto-switches between in-memory and SQLite based on data size

Contents

📦Installation

Install AVA using your preferred package manager:

# npm
npm install @antv/ava
# yarn
yarn add @antv/ava

🚀Quick Start

Get started with AVA in just a few lines of code:

import { AVA } from '@antv/ava';

// Initialize with LLM config
const ava = new AVA({
  llm: {
    model: 'ling-1t',
    apiKey: 'YOUR_API_KEY',
    baseURL: 'LLM_BASE_URL',
  },
  sqlThreshold: 1024 * 1024 * 2, // 2MB threshold
});

// Load data
await ava.loadObject([
  { city: 'Hangzhou', gdp: 18753 },
  { city: 'Shanghai', gdp: 43214 }
]);

// Ask questions
const result = await ava.analysis(
  'What is the average GDP?'
);
console.log(result.text);
// Access: result.code, result.sql, result.data

// Generate visualization from analysis result
const viz = await ava.visualize(result);
console.log(viz.syntax); // chart syntax for GPT-Vis
// or use viz.html for standalone HTML

// Clean up
ava.dispose();

🏗️Architecture

AVA uses a modular pipeline architecture:

1
Data Module
Load from CSV, JSON, URL, or text
2
Metadata Extract
Type inference and statistics
3
Size Check
<10MB: JavaScript | ≥10MB: SQLite
4
Analysis Module
Generate & execute code/SQL
5
LLM Summary
Natural language response
6
Visualization
Chart advisor + chart generation

📚API Reference

Constructor

new AVA(options: AVAOptions)
llm— LLM configuration (model, apiKey, baseURL)
sqlThreshold— Size threshold for SQLite (default: 10MB)

Data Loading

loadCSV(content: string)

Load data from CSV string

loadObject(data: object[])

Load data from array of objects

loadURL(url: string, transform?: Function)

Load data from URL

loadText(text: string)

Extract data from unstructured text

Analysis

analysis(query: string)

Analyze data with natural language query. Returns text summary + structured data + code/SQL.

Returns object with:
• text — Natural language summary of the analysis
• data — Structured analysis result (array or object)
• code — JavaScript code used (if applicable)
• sql — SQL query used (if applicable)

Visualization

visualize(analysisResult: AnalysisResponse)

Generate chart from analysis result. Must be called after analysis().

Returns object with (or null if no visualization needed):
• chartType — Recommended chart type
• syntax — GPT-Vis chart syntax for rendering
• html — Standalone HTML that renders the chart

Query Suggestions

suggest(count?: number)

Get AI-recommended analysis queries based on dataset characteristics (default: 3)

Returns array of objects with:
• query — Suggested analysis question
• score — Meaningfulness score (0-1)
• reason — Explanation for the suggestion

💡Usage Examples

Query Suggestions

await ava.loadObject([
  { city: 'Hangzhou', gdp: 18753 },
  { city: 'Shanghai', gdp: 43214 }
]);

// Get 5 suggested queries
const suggestions = await ava.suggest(5);
console.log(suggestions[0]);
// { query: "What is the average GDP?",
//   score: 0.95,
//   reason: "Reveals economic patterns" }

// Use suggested query for analysis
const result = await ava.analysis(suggestions[0].query);

Browser File Upload

const fileInput = document.querySelector('input[type="file"]');
fileInput.addEventListener('change', async (e) => {
  const file = e.target.files[0];
  const content = await file.text();
  await ava.loadCSV(content);
  const result = await ava.analysis('Show trends');
});

API Data Analysis

await ava.loadURL(
  'https://api.example.com/sales',
  (response) => response.data.items
);
const result = await ava.analysis('Compare by region');

⚙️Configuration

AVA supports multiple LLM providers:

OpenAI
model: 'gpt-4'
baseURL: 'api.openai.com/v1'
Custom Provider
model: 'ling-1t'
baseURL: 'your-llm-api.com'

⭐Best Practices

✓Always call ava.dispose() when done to free resources
✓Wrap async calls in try-catch blocks for error handling
✓Validate data format before loading
✓Set sqlThreshold based on expected data sizes
✓Never expose API keys in client-side code

Resources

→ GitHub Repository→ Report Issues→ AntV Community