The master’s thesis addresses the challenges of software configuration with an emphasis on feature flags as a key mechanism in modern development environments. We proposed an extensible system for managing and removing these flags by integrating large language models, which enables language independent detection and elimination of technical debt. The developed solution achieves 90% success rate in automated flag removal and 81% success rate in combined automated flag detection and removal, while also introducing an architecture for dynamic evaluation and more robust testing. The thesis contributes a comprehensive approach to managing complexity in functionality management and offers a basis for further research.
|
