CodeCraft Agents

Transform natural language task descriptions into fully functional, tested, and dockerized code solutions

🌐 Other Languages

• README in Ukrainian

🚀 Overview

CodeCraft Agents is an advanced multi-agent AI system that automatically converts natural language descriptions into production-ready code. The project demonstrates the power of specialized AI agents working in collaboration to handle the entire software development lifecycle - from task analysis to deployment.

Note: This is a prototype version currently optimized for single-file applications.

✨ Key Features

• Fully Automated Code Generation - From concept to working code with minimal human intervention
• Multi-Agent Architecture - Specialized agents for each development stage
• Comprehensive Verification - Built-in testing, validation, and feedback loops
• Docker Integration - Automatic containerization of generated applications
• Knowledge Extraction - Continuous learning from generated solutions

🤖 Agent System Architecture

Our system employs a specialized agent for each stage of the development process:

1. DecomposerAgent - Breaks down the task description into structured modules and interfaces
2. ValidatorAgent - Verifies the completeness and correctness of the plan
3. ConsistencyAgent - Checks for data type and logic consistency across modules
4. CodeGeneratorAgent - Creates the actual code based on the approved plan
5. CodeExtractorAgent - Saves the code to appropriate files
6. DockerRunnerAgent - Generates Dockerfile and docker-compose configurations
7. TesterAgent - Creates test cases and validates application functionality
8. DocumentationAgent - Generates comprehensive usage documentation
9. MonitorAgent & CoordinatorAgent - Oversee the process and manage workflow transitions

🔄 How It Works

When you provide a task description to CodeCraft Agents, the system follows this workflow:

1. Task Analysis - The DecomposerAgent breaks down the task into manageable components
2. Planning & Validation - The system creates and verifies a comprehensive development plan
3. Code Generation - Clean, efficient code is created based on the approved plan
4. Testing & Containerization - The code is thoroughly tested and packaged in a Docker container
5. Documentation - Comprehensive usage documentation is automatically generated

The entire process is governed by our adaptive feedback loop, which ensures high-quality output through multiple verification layers.

📋 Example Use Case

Given a simple input:

Create an API server with route /sum that takes two GET parameters a and b (numbers) and returns their sum. Use aiohttp.

CodeCraft Agents will produce:

1. A working Python application with error handling and proper structure
2. Dockerized environment for easy deployment
3. Test suite to verify functionality
4. Comprehensive documentation for usage and deployment

🛠️ Technical Implementation

The system is built on a modular architecture that includes:

• Advanced prompt engineering for specialized agent roles
• Verification mechanisms for quality control
• Adaptive feedback loops for error correction
• Isolated execution environments for testing
• Vector knowledge base for context enhancement

🔍 Current Limitations

As a prototype, CodeCraft Agents currently has some limitations:

• Optimized for single-file Python applications
• Works best with web API and basic utility applications
• Limited complexity in database relationships and authentication
• Does not yet support full-scale multi-file projects

🚀 Getting Started

Prerequisites

• Python 3.9 or higher
• Docker and docker-compose
• Access to OpenRouter API
• Qdrant for vector storage

Installation

1. Clone the repository:

git clone https://github.com/yourusername/codecraft-agents.git
cd codecraft-agents

2. Install dependencies:

pip install -r requirements.txt

3. Start Qdrant using docker-compose:

docker-compose up -d

4. Set your OpenRouter API key:

export OPENROUTER_API_KEY=your_key_here

Usage

Run the system with a task description:

python main.py

By default, the system will generate code for a sample task. To customize, edit the task variable in main.py.

Output files will be created in the project/ directory.

📚 Documentation

For more detailed information on the system's architecture and components, see:

• Agent Architecture
• Prompt Engineering Details
• Verification System
• Feedback Loop Mechanism

📈 Future Roadmap

• Multi-file project support
• Additional programming languages
• Frontend code generation
• More complex database integrations
• CI/CD pipeline generation
• Enhanced knowledge extraction and reuse
• React and other frontend framework support

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE.md file for details.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

---

Built with ❤️ by the CodeCraft Agents Team