Masterarbeit
AI-Driven Transparent Repair of Software Dependency Configurations
Research Area
Intelligent Information Management
Advisers
Dr. Sheeba Samuel
Description
In modern software development and scientific computing, correct dependency configurations and import statements are essential for ensuring that code executes reliably and remains reproducible. However, many projects—especially those shared on platforms such as GitHub—suffer from broken or missing dependencies, incorrect import statements, or outdated configuration files, leading to execution failures and reduced usability. These issues are particularly common in research software, where environments are often not fully specified or maintained. The objective of this thesis is to design and implement an AI-driven system that automatically detects and repairs issues related to software dependencies and import statements. Using AI techniques, the system will analyze source code, configuration files (e.g., requirements.txt, environment.yml), and execution errors to identify inconsistencies and propose fixes. A key aspect of the work is transparency and explainability, ensuring that all suggested repairs are accompanied by clear justifications and, where possible, minimal changes to the original code. The thesis will result in a prototype tool that can analyze software repositories, recommend or apply fixes for dependency and import-related issues, and generate human-readable explanations for its actions. The system will be evaluated on a set of real-world repositories to assess its effectiveness in improving executability and reproducibility.
