DataChat

Interactive tool for exploring and analyzing datasets through natural language conversations.

Features

• Natural language queries to SQL translation
• Support for multiple data formats:
  • CSV files
  • SQLite databases (.sqlite3, .db)
• Real-time data exploration
• Interactive table previews
• Smart question suggestions
• Pre-built dataset demos:
  • Card Transactions Analysis
  • HR Employee Data
  • Marvel Character Powers
• Dark mode support

Usage

1. Upload your data files (CSV or SQLite)
2. View table schemas and sample data
3. Ask questions in natural language
4. Get SQL queries and visualized results
5. Try suggested sample questions
6. Explore pre-built demo datasets

Setup

Prerequisites

• Modern web browser with ES Modules support
• Web server for local development

Local Setup

1. Clone this repository:

git clone https://github.com/gramener/datachat.git
cd datachat

2. Serve the files using any static web server. For example, using Python:

python -m http.server

3. Open http://localhost:8000 in your web browser

Deployment

On Cloudflare DNS, proxy CNAME datachat.straive.app to gramener.github.io.

On this repository's page settings, set

• Source: Deploy from a branch
• Branch: main
• Folder: /

Technical Details

Architecture

• Frontend: Vanilla JavaScript with lit-html for rendering
• Database: SQLite WASM for client-side data processing
• LLM Integration: GPT-4o Mini through LLM Foundry API
• Styling: Bootstrap 5.3.3 with dark mode support

Dependencies

All dependencies are loaded via CDN:

• SQLite WASM v3.46 - Database operations
• lit-html v3 - Template rendering
• Bootstrap v5.3.3 - UI components
• D3-dsv v3 - CSV parsing
• marked v13 - Markdown parsing
• highlight.js v11 - Code syntax highlighting

License

MIT