Healing Agent 🩺

Healer Agent is an intelligent code assistant that catches with detailed context and fixes errors in your Python code. It leverages the power of AI to provide smart suggestions and corrections, helping you write more robust and "self-healing" code. Your program will be able to fix itself, it will have regenerative healing abilities like Wolverine.

⚠️ Not intended for production use. Be extra careful with the optional AUTO_FIX function, as although it makes backups of your code, it actually changes and runs your code. ⚠️

Goal: first actually usable autonomous coding agent in production

Video demo on Youtube

Features ✨

• 🚨 Automatic error detection and handling of diverse exception types
• 💡 Smart error analysis and solution suggestions (auto-generated fixing hints and code)
• 🔍 Comprehensive error analysis including exception details, stack traces, local and globalvariables and root cause identification
• 🧠 Advanced AI-powered code healing using LLMs of different providers
• 🔧 Zero-config integration with Python projects (just import and decorate)
• 💾 Robust error tracking and debugging:
  • Exception context saved to JSON (code, error details, function info and args)
  • Automatic code backups before fixes
  • Detailed analysis results and fix history
  • Quick test of fixes
• 🤖 (Optionally) Fully automated operation with minimal human intervention

How it works 🧠

graph TD
    A[Import healing_agent] --> B[Configuration: AI access etc.]
    B --> C[Decorate functions with healing_agent]
    C --> D[Run Code / Execute Functions]
    D -->|No problem| L[Success]
    D -->|Exception?| F[Get and Save Detailed Context]
    F --> G[Auto-generate Fixing Hints and Code with AI]
    G --> H[Test Generated Code]
    H --> I[Create backup]
    I --> J[Apply Code Fixes]
    J --> D

Loading

Installation 💻

To install Healing Agent, follow these steps:

PIP package from GitHub:

pip install git+https://github.com/matebenyovszky/healing-agent

OR from source:

1. Clone the repository:

   git clone https://github.com/matebenyovszky/healing-agent.git

2. Navigate to the project directory:

   cd healing-agent

3. Install:

   pip install -e .

   OR run overall test to install and test functionality:

   python scripts/overall_test.py

Usage 🔧

To use Healing Agent in your project, follow these steps:

1. Import the healing_agent decorator in your Python file:

   import healing_agent

2. Decorate the function you want to monitor with @healing_agent:

   @healing_agent
   def your_function():
       # Your code here

   You can also pass parameters to the decorator to change the behavior set in the config file:

   @healing_agent(AUTO_FIX=False)
   def your_function():
       # Your code here

3. Run your Python script as usual. Healing Agent will automatically detect, save context and attempt to fix any errors that occur within the decorated function.

Context (and code file backup in case of auto-fix) is saved to a JSON/Python file in the same directory as your script with actual timestamp in the filename.

Configuration ⚙️

Healing Agent provides extensive configuration options through the healing_agent_config.py file, which defines essential settings such as the AI provider and API credentials. The configuration system follows these principles:

1. Automatic Configuration Loading: On startup, Healing Agent attempts to load settings from healing_agent_config.py
2. Fallback Mechanism: If the configuration file is not found, the system falls back to pre-defined default settings
3. Auto-Configuration: When no configuration file exists, Healing Agent automatically creates one in the default user directory

Supported AI Providers

Healing Agent integrates with multiple AI providers - list could be extended:

• OpenAI
• Azure OpenAI
• LiteLLM
• Anthropic
• Ollama

Note: While multiple providers are supported, Azure OpenAI has been extensively tested. Support for other providers is under active development (feedback welcome).

Testing 🧪

To test Healing Agent, you can use the scripts/test_file_generator.py script to generate test files in the tests directory. overall_test.py will run all tests and provide a report on the functionality of Healing Agent.

Use Cases 💡

• Development: Use Healing Agent during development to catch and fix errors early, and let AI generate fixes for your code. This is what you would do anyways, but now it's automated. 😁
• Educational Tool: Use Healing Agent as a learning tool to understand AI coding capabilities and limitations.

Cooking open source 🍳

Healing Agent is distributed under the MIT License. See LICENSE for more information. Feedback and contributions are welcome!