KGCompass

KGCompass is a novel approach for repository-level software repair that accurately links code structure with repository metadata using a knowledge graph, enabling more precise bug localization and patch generation.

Paper link: https://arxiv.org/abs/2503.21710

Fully Containerized Workflow with GPU Support

This project uses Docker and Docker Compose to provide a fully reproducible environment. The setup includes:

• A base image with CUDA and Python pre-installed.
• A service for the Neo4j database with necessary plugins.
• An application service with all Python dependencies and access to the host's GPU.

Prerequisites

1. NVIDIA GPU & Drivers: A compatible NVIDIA GPU with recent drivers installed on your host machine.
2. NVIDIA Container Toolkit: You must install this on your host to allow Docker to use the GPU. For Debian/Ubuntu-based systems, you can do so by running the following command block in your terminal:

   curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
   && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
     sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
     sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list \
   && sudo apt-get update \
   && sudo apt-get install -y nvidia-container-toolkit \
   && sudo nvidia-ctk runtime configure --runtime=docker \
   && sudo systemctl restart docker

3. Docker & Docker Compose: This project uses Docker Compose V1.
   • Ensure Docker is installed on your system.
   • Install Docker Compose V1 (if not already present) by running:

   LATEST_COMPOSE_V1="1.29.2"
   sudo curl -L "https://github.com/docker/compose/releases/download/${LATEST_COMPOSE_V1}/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
   sudo chmod +x /usr/local/bin/docker-compose

4. API Keys: Create a .env file in the project root by copying the example:

   cp .env.example .env

   Then, edit the .env file and fill in your GITHUB_TOKEN, and BAILIAN_API_KEY.

Step 1: Build and Start All Services

This single command will build the base CUDA image, the Neo4j image, and the final application image, then start all services in the background.

docker-compose up -d --build

Step 2: Run the Repair Pipeline

Execute the repair script inside the application container. The container will have access to the GPU.

docker-compose exec app bash run_repair.sh <instance_id>

# Example:
docker-compose exec app bash run_repair.sh astropy__astropy-12907

Step 3: Stopping the Environment

docker-compose down -v

Citation

If you use KGCompass in your research, please cite the following paper:

@article{yang2025enhancing,
  title={Enhancing Repository-Level Software Repair via Repository-Aware Knowledge Graphs},
  author={Yang, Boyang and Tian, Haoye and Ren, Jiadong and Jin, Shunfu and Liu, Yang and Liu, Feng and Le, Bach},
  journal={arXiv preprint arXiv:2503.21710},
  year={2025}
}