This project provides an experimental autonomous GPT agent capable of interacting with a web browser to accomplish user-defined tasks. It is designed for developers and researchers interested in AI-driven automation and agentic workflows, offering a practical demonstration of GPTs controlling browser actions.

How It Works

The assistant leverages GPT-4 to interpret natural language prompts and generate a sequence of actions. These actions are then executed using Puppeteer, a Node.js library that provides a high-level API to control Chrome or Chromium over the DevTools Protocol. This approach allows the agent to navigate websites, extract information, and interact with web elements programmatically.

Quick Start & Requirements

• Install dependencies: npm install
• Setup: Clone the repo, create a .env file from .env.template, and add OPENAI_KEY and DATABASE_URL (Postgres).
• Generate tables: npx prisma migrate dev --name init
• Run: npm run dev
• Prerequisites: Node.js 14+, OpenAI API key (GPT-4 recommended), Postgres database.

Highlighted Details

• Demonstrates autonomous task completion via browser interaction.
• Utilizes GPT-4 for prompt interpretation and action generation.
• Employs Puppeteer for headless browser automation.
• Built with Qwik for the frontend framework.

Maintenance & Community

• Early-stage project with ongoing development.
• Community support available on the Builder.io Discord server in the #gpt-assistant channel.

Licensing & Compatibility

• License: Not explicitly stated in the README.
• Compatibility: Requires Node.js and a Postgres database.

Limitations & Caveats

The project is in its early stages, with sparse documentation and potentially messy code. Contributing guidelines mention that it's a work in progress.