KGCompass combines knowledge graphs with large language models to significantly improve repository-level software repair by creating meaningful connections between issues and code.

• Addresses key challenges in software repair by bridging semantic gaps between issue descriptions and necessary code patches
• Uses a repository-aware knowledge graph that links issues, pull requests, and code artifacts for better context understanding
• Enhances LLMs' reasoning capabilities through structured representation of repository knowledge
• Demonstrates improved bug localization and repair accuracy compared to existing approaches

This innovation matters for engineering teams by reducing debugging time, improving fix accuracy, and enhancing code quality maintenance at scale across complex repositories.

Enhancing Repository-Level Software Repair via Repository-Aware Knowledge Graphs