METHOD OF AUTOMATED INTEGRATION TEST GENERATION AND CONTINUOUS EXECUTION IN LARGE-SCALE CODE REPOSITORIES

Abstract

Maintaining high-quality tests in large-scale code repositories remains a critical challenge due to frequent code changes and the high cost of manual test creation and maintenance. This paper proposes a method of automated test generation and continuous execution integrated directly into a .NET Azure DevOps pipeline. The method leverages large language models to automatically generate and update integration tests in response to every code commit. Once generated, the tests are compiled and executed within the same pipeline, while a self-healing mechanism attempts to automatically recover failing AI-generated tests. If recovery fails, the pipeline halts to prevent defective deployments. The approach ensures that the test suite evolves continuously alongside the codebase, enabling real-time validation of new functionality. Experimental evaluation on a large-scale project demonstrated improved test coverage, reduced manual testing effort, and enhanced defect detection during continuous integration. The research highlights the synergy between AI-driven code analysis and continuous testing, showing how automated test generation can strengthen DevOps practices. Future work will focus on fine-tuning language models for domain-specific test generation and improving self-healing accuracy to further reduce developer intervention.