LEDGE : Leveraging Dependency Graphs for Enhanced Context Aware Documentation Generation

In software engineering, effective documentation is crucial for understanding complex codebases, yet it often remains incomplete or outdated, hindering developer productivity. This paper introduces LEDGE (Leveraging Dependency Graphs for Enhanced Context Aware Documentation Generation), a novel framework that integrates dependency graphs with large language models (LLMs) to automate the generation of structured, context aware software documentation. By leveraging GraphRAG, LEDGE captures semantic and structural relationships within codebases, enabling precise documentation that highlights architectural insights and module dependencies. Our methodology employs a parser based approach to construct dependency graphs, stored in MemGraph using Cypher queries, and utilizes vector embeddings for similarity based retrieval. Evaluated on diverse open source repositories, LEDGE demonstrates superior clarity and relevance compared to traditional documentation, as evidenced by qualitative and quantitative analyses. The framework enhances software maintainability, developer onboarding, and knowledge transfer, offering a scalable solution for modern software development. Our code and data are available at https://github.com/MihirRajeshPanchal/LEDGE.