Our paper: "Repo2Run: An LLM-based Build Agent System for Containerized Environments" has been accepted by NeurIPS 2025 main track as a spotlight!

An LLM-based build agent system that helps manage and automate build processes in containerized environments. This project provides tools for handling dependencies, resolving conflicts, and managing build configurations.

• Docker-based sandbox environment for isolated builds
• Automated dependency management and conflict resolution
• Support for Python version management
• Waiting list and conflict list management for package dependencies
• Error format handling and output collection

• Python 3.x
• Docker
• Git

1. Clone the repository:

git clone https://github.com/bytedance/repo2run.git
cd repo2run

2. Install the required dependencies:

pip install -r requirements.txt

The main entry point is through the build agent's main script. You can run it with the following arguments:

python build_agent/main.py <repository_full_name> <sha> <root_path>

Where:

• repository_full_name: The full name of the repository (e.g., user/repo)
• sha: The commit SHA
• root_path: The root path for the build process

• build_agent/ - Main package directory
  • agents/ - Agent implementations for build configuration
  • utils/ - Utility functions and helper classes
  • docker/ - Docker-related configurations
  • main.py - Main entry point
  • multi_main.py - Multi-process support

The project uses Docker containers to create isolated build environments, ensuring clean and reproducible builds.

• Waiting List: Manages package installation queue
• Conflict Resolution: Handles version conflicts between packages
• Error Handling: Formats and processes build errors

Utilizes GPT models to assist in build configuration and problem resolution.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

@article{hu2025llm,
  title={An LLM-based Agent for Reliable Docker Environment Configuration},
  author={Hu, Ruida and Peng, Chao and Wang, Xinchen and Gao, Cuiyun},
  journal={arXiv preprint arXiv:2502.13681},
  year={2025}
}

Apache-2.0

https://github.com/Aider-AI/aider

pengchao.x@bytedance.com