CodexGraph

Introduction

we introduce CodexGraph, a system that integrates LLM agents with graph database interfaces extracted from code repositories. By leveraging the structural properties of graph databases and the flexibility of the graph query language, CodexGraph enables the LLM agent to construct and execute multi-hop queries, allowing for accurate, code structure-aware context retrieval and code navigation.

Results

Real-world Applications

we develop five real-world coding applications based on CODEXGRAPH. With a unified graph database schema, CODEXGRAPH demonstrates competitive performance across diverse academic tasks and highlights its potential application value in real production environments, underscoring its versatility and efficacy in various software engineering applications.

Citation

@misc{liu2024codexgraphbridginglargelanguage,
                  title={CodexGraph: Bridging Large Language Models and Code Repositories via Code Graph Databases}, 
                  author={Xiangyan Liu and Bo Lan and Zhiyuan Hu and Yang Liu and Zhicheng Zhang and Fei Wang and Michael Shieh and Wenmeng Zhou},
                  year={2024},
                  eprint={2408.03910},
                  archivePrefix={arXiv},
                  primaryClass={cs.SE},
                  url={https://arxiv.org/abs/2408.03910}, 
            }