Abstract: In collaborative software development, contributors frequently make code changes to a common code base. To avoid conflicts and contribute effectively, developers must keep up with the changes of others. Reading through code changes, pull request descriptions or issues can be time-consuming, particularly in large projects. While there is an abundance of research on generating commit messages and release notes, little attention has been devoted to generating developer-oriented summaries of contributions at the repository level. To address the research gap and help developers understand code changes more efficiently, this thesis proposes an approach to generate highly configurable, personalized summaries of code contributions from GitHub and GitLab repositories on-demand. The approach aggregates data from commits, pull requests and issues, analyzes the data using defined rules, and summarizes pull requests using Large Language Models (LLMs). The resulting summaries make it easier for readers to spot important contributions and allow them to understand the changes and their broader context faster.
Keywords: Code Summarization, SWE Automation, Rule-Based Analysis, LLM-Based Summaries, AI-Powered Developer Tools
PDF: Master Thesis
Reference: Luca Bretting. Automatic Summary of Code Contributions. Master Thesis. Friedrich-Alexander-Universität Erlangen-Nürnberg: 2026.
Discover more from Professorship for Open-Source Software
Subscribe to get the latest posts sent to your email.