Auto Coder Project

This project implements an AI-powered code generation system using the DeepSeek v3 model through OpenRouter's API. It consists of two implementations: a single-iteration version and an iterative version with code improvement capabilities.

⚠️ IMPORTANT SECURITY WARNING

This project executes AI-generated code on your machine. This can be potentially dangerous!

• AI-generated code could contain harmful operations
• You can consider adding a code review step before code execution
• Use in an isolated environment (virtual machine) for maximum safety possibly

Use at your own risk. The authors are not responsible for any damage caused by AI-generated code.

🤖 Model Flexibility

While this project defaults to using DeepSeek v3 through OpenRouter, it's designed to work with any model available on OpenRouter! You can easily switch models by changing the MODEL constant in either script:

MODEL = "deepseek/deepseek-chat"  # Default

• Any other model supported by OpenRouter

The code generation, error fixing, and improvement capabilities will adapt to whatever model you choose!

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Set environment variable:

export OPENROUTER_API_KEY="your_api_key_here"  # Linux/Mac
set OPENROUTER_API_KEY="your_api_key_here"     # Windows

❤️ Support & Get 400+ AI Projects

This is one of 400+ fascinating projects in my collection! Support me on Patreon to get:

• 🎯 Access to 400+ AI projects (and growing daily!)
  • Including advanced projects like 2 Agent Real-time voice template with turn taking
• 📥 Full source code & detailed explanations
• 📚 1000x Cursor Course
• 🎓 Live coding sessions & AMAs
• 💬 1-on-1 consultations (higher tiers)
• 🎁 Exclusive discounts on AI tools & platforms (up to $180 value)

1. Single Iteration Auto Coder (1_auto_coder_single_iteration.py)

A basic implementation that generates code once based on a user prompt.

Features:

• Takes user input as a constant variable (USER_PROMPT)
• Uses DeepSeek model for code generation
• Streams model responses in real-time
• Extracts code from between <code> and </code> tags
• Saves generated code to files in the generated_code directory

Usage:

1. Set your desired prompt in the USER_PROMPT constant
2. Run the script:

python 1_auto_coder_single_iteration.py

2. Iterative Auto Coder (2_auto_coder_iterative.py)

An advanced implementation that iteratively improves generated code through execution feedback.

Features:

• All features from the single iteration version, plus:
• Executes generated code with a 5-second timeout
• Provides execution feedback to the model
• Uses separate API calls for:
  • Initial code generation
  • Error fixing
  • Code improvements
• Saves each iteration as numbered files (e.g., generated_code_1.py, generated_code_2.py)
• Maximum 5 iterations for improvements
• Handles timeouts gracefully (not treated as errors)

Execution Flow:

1. Generate initial code
2. Execute the code
3. Based on execution result:
   • If timeout: Attempt to improve efficiency
   • If errors: Make fresh API call to fix errors
   • If success: Make fresh API call to add features
4. Save new iteration
5. Repeat until:
   • Max iterations reached
   • No more improvements needed
   • Error can't be fixed

Usage:

1. Set your desired prompt in the USER_PROMPT constant
2. Run the script:

python 2_auto_coder_iterative.py

System Prompts

The iterative version uses three distinct system prompts:

1. SYSTEM_PROMPT_GENERATE: For initial code generation
2. SYSTEM_PROMPT_FIX: For fixing errors (fresh context)
3. SYSTEM_PROMPT_IMPROVE: For adding features (fresh context)

Output Directory Structure

generated_code/
├── generated_code_1.py  # Initial generation
├── generated_code_2.py  # First improvement/fix
├── generated_code_3.py  # Second improvement/fix
└── ...

Error Handling

Both versions include comprehensive error handling:

• API call errors
• Code extraction errors
• File saving errors
• Code execution errors (iterative version)
• Timeout handling (iterative version)

Terminal Output

Both versions provide colored terminal output using termcolor:

• Cyan: Informational messages
• Yellow: Processing status
• Green: Success messages
• Red: Error messages

Limitations

1. 5-second execution timeout
2. Maximum 5 improvement iterations
3. No persistent context between API calls
4. Code must be complete and self-contained
5. Generated code must be valid Python

Dependencies

• openai: For API communication
• termcolor: For colored terminal output
• Python standard library modules:
  • os
  • subprocess (iterative version)
  • time (iterative version)