Abstract
This disclosure describes techniques to improve automated root cause analysis (RCA) for integration test failures, especially in iterative debugging. A multi-step artificial intelligence (AI) pipeline is employed to dynamically select the appropriate baseline for log comparison—either the last passing run (LPR) for initial failures or the previous failing run (PFR) for subsequent failures—to reduce semantic noise. A large language model (LLM) performs phase-aware semantic differencing, segmenting logs into execution phases (e.g., setup, data seeding, service under test execution, assertion, teardown) and comparing corresponding phases. This targeted approach allows the LLM to identify critical differences and new errors more accurately by focusing on specific phases where failures manifest. This can enable efficient and precise root cause identification and solution proposals through retrieval-augmented generation (RAG). The techniques enhance developer productivity, improve development velocity, and provide automated feedback loops in CI/CD environments.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Khurana, Shubham and Elumalai, Babu Prasad, "Automated Root Cause Analysis for Integration Test Failures Using Consecutive Failing Run Comparisons and Phase-Aware Differencing Using a Large Language Model", Technical Disclosure Commons, (October 07, 2025)
https://www.tdcommons.org/dpubs_series/8686