Food Chain Visualizer

An interactive demonstration of Large Language Models (LLMs) running directly in the browser, showcased through a food chain visualization tool. This project demonstrates how AI models can be integrated into web applications without requiring server-side processing.

Overview

This application uses WebLLM to run a quantized LLM directly in your browser. It generates detailed biological information about animals and their food chain relationships in real-time, demonstrating the capabilities of client-side AI processing.

Key Technical Features

• Browser-based LLM execution using WebLLM
• No server requirements - all processing happens client-side
• Efficient model loading and execution
• Structured JSON outputs from natural language queries
• Interactive visualization of AI-generated data

Visualization Features

• Interactive network visualization of food chain relationships
• Color coding based on diet type:
  • Green: Herbivores
  • Red: Carnivores
  • Orange: Omnivores
• Dynamic node sizing based on animal's physical size
• Automatic placeholder completion
• Random animal suggestions
• Detailed animal information display

Setup

1. Install dependencies:

npm install

2. Start the development server:

npm start

Usage

• Enter an animal name and press Enter or click "Add to Food Chain"
• Click "Random Animal" to add a random animal to the network
• Use "Clear Network" to reset the visualization
• Click "Auto Complete" to automatically fill in placeholder nodes
• Single-click nodes to view detailed information
• Double-click placeholder nodes to get AI-generated data

Technical Details

• Uses Llama-3.2-3B-Instruct model (quantized for browser execution)
• Generates structured JSON responses for consistent data formatting
• Implements WebLLM for client-side model execution
• Visualizes data using vis.js network library

Browser Compatibility

The application requires a modern browser with WebAssembly support. Performance may vary depending on your device's capabilities.

Note

This project demonstrates the potential of running AI models directly in the browser, making AI-powered applications more accessible and reducing the need for server-side infrastructure.