SWE-crafter uses your LLM to automatically fix GitHub issues. Emphasizing a Test-First approach, it validates PRs with pre-written tests, providing high-quality feedback and helping the agent correct hallucinations.

SWE-crafter explores the application of good human development practices—like those from Extreme Programming—on autonomous agents. By integrating a Test-First methodology, SWE-crafter aims to enhance code quality and bug resolution in automated software engineering.

Benchmarks for this method are ongoing, showing promising improvements in code quality and bug resolution.

• Python 3.8 or higher
• An API key for your preferred LLM (e.g., OpenAI API key)

Clone the repository:

git clone https://github.com/umans-tech/SWE-crafter.git
cd SWE-crafter

Install the required dependencies:

pip install -r requirements.txt

Set your LLM API key as an environment variable:

export OPENAI_API_KEY='your-api-key-here'

Run SWE-crafter to start resolving GitHub issues:

python swe_crafter.py

(Adjust the command above based on your actual entry point.)

• Automated Issue Resolution: Uses your LLM to automatically address GitHub issues.
• Test-First Approach: Validates pull requests with pre-written tests to ensure code quality.
• High-Quality Feedback: Provides detailed feedback to help correct hallucinations and improve solutions.
• Improved Code Quality: Early benchmarks indicate significant improvements in bug resolution and overall code quality.
• Adopts Best Practices: Implements human development practices from Extreme Programming in autonomous agents.

Detailed documentation is coming soon! Stay tuned for updates.

We are actively benchmarking SWE-crafter to evaluate the effectiveness of the Test-First approach in autonomous agents. Preliminary results are promising, showing improvements in both code quality and bug resolution.

We welcome contributions from the community!

• Questions & Discussions: Feel free to open an issue or start a discussion for any questions or ideas.
• Reporting Bugs: If you encounter any issues, please report them via GitHub Issues.
• Contributing Code: We appreciate your pull requests! Please ensure your updates include appropriate tests and documentation.

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