docs/GETTING_STARTED.md

Getting Started with GAIA Agent Development

This guide will help you get started with developing the GAIA Agent using your existing virtual environment.

Prerequisites

• Python 3.8+
• Virtual environment (already in .venv)
• Hugging Face account (for deployment)

Setup and Installation

1. Activate your existing virtual environment:

   source .venv/bin/activate
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Install additional packages for the agent:

   pip install gpt4all beautifulsoup4 pandas pillow python-dotenv searchapi
   

Development Workflow

1. Local Testing:

   python app_local.py
   

   This will run a local version of the agent with a limited question set for testing.

2. Running the full agent:

   python app2.py
   

   Note: This requires Hugging Face authentication when running locally.

3. Evaluating the agent:

   python utilities/evaluate_local.py
   

   This will evaluate your agent against the common questions dataset.

Project Structure

• app2.py - The main GAIA agent implementation
• app_local.py - Modified version for local testing without requiring login
• devplan.md - Development plan and architecture design
• question_set/ - Contains question datasets for testing
• utilities/ - Helper scripts for evaluating and testing
• docs/ - Documentation about the API and submission process

Next Steps

See the NEXT_STEPS.md file for a checklist of planned improvements.

Troubleshooting

• Authentication Issues: For local testing, use app_local.py which doesn't require HF login
• Missing Dependencies: Make sure to install all requirements with pip install -r requirements.txt
• File Not Found Errors: Create a dataset directory for downloaded files