3D Generation Multi-Agent System

A sophisticated multi-agent system for generating 3D CAD models using AI agents. This project implements an iterative workflow that combines image generation, metadata analysis, and evaluation to create high-quality 3D models.

Features

• 🤖 Multi-Agent Architecture: Uses specialized agents for generation and evaluation

• 🎨 Image Generation: Creates multi-view images using DALL-E

• 📊 Metadata Generation: Produces detailed metadata for 3D CAD reconstruction

• 🔄 Iterative Improvement: Continuously refines results based on evaluation feedback

• ⚡ Multi-API Support: Test different AI providers (OpenAI, Claude, DeepSeek, Qwen)

Architecture

Agents

1. Generation Agent: Creates multi-view images and metadata for 3D CAD reconstruction
2. Evaluation Agent: Assesses quality and provides improvement suggestions
3. Mesh Generation Agent: Converts results into 3D mesh data

Workflow

1. User provides a query describing the desired 3D object
2. Generation agent creates multi-view images and metadata
3. Evaluation agent assesses quality and provides feedback
4. System iteratively improves results until quality threshold is met
5. Final 3D mesh is generated

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd openai-agents-python

2. Set up virtual environment:

   python -m venv env
   source env/bin/activate  # On Windows: env\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure API keys: Create a .env file in the root directory and add your API keys:

   OPENAI_API_KEY=your_openai_key
   CLAUDE_API_KEY=your_claude_key
   DEEPSEEK_API_KEY=your_deepseek_key
   QWEN_API_KEY=your_qwen_key

Usage

Command Line Interface

Run the web application:

cd webapp
python app.py

The system will:

• Start a web server for the 3D generation interface
• Allow you to input your desired 3D object
• Generate and iteratively improve results
• Save outputs to organized directories

API Testing

The system supports multiple AI providers for speed and quality testing:

• OpenAI GPT-4o: Fast and reliable
• Claude 3 Sonnet: High-quality reasoning
• DeepSeek: Cost-effective alternative
• Qwen: Alibaba's AI model

Switch between models by changing the model configuration in the web application or use the interactive selection.

Output Structure

project/
├── renders/                    # Generated images
├── evaluation_reports_*/       # Evaluation reports per iteration
├── mesh_outputs/              # Final 3D mesh data
└── webapp/                    # Web interface

Configuration

Model Selection

The web application allows you to select different AI models through the interface. Available options include OpenAI GPT-4o, Claude 3 Sonnet, DeepSeek, and Qwen.

Quality Thresholds

# In evaluation agent
if all scores > 6.5:
    suggestions = "well done"

Development

Adding New AI Providers

1. Add API key to the configuration
2. Create client with appropriate base URL
3. Add to MODEL_CONFIGS dictionary
4. Test with switch_model() function

Customizing Prompts

Edit the prompts in the web application:

• Generation agent prompts in webapp/app.py
• Evaluation agent prompts in the application logic
• Image generation prompts in the DALL-E integration

Troubleshooting

• API Errors: Check your API keys and quotas
• Import Errors: Ensure all dependencies are installed
• Memory Issues: Reduce image resolution in generation
• Slow Performance: Try different AI providers

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built on OpenAI's agents framework
• Uses DALL-E for image generation
• Inspired by modern multi-agent architectures