Context-Enriched Logging for Intelligent Code Fix Suggestions Using Large Language Models

Abstract

Software systems increasingly depend on runtime logs for problem diagnosis, yet traditional logging lacks depth for automated analysis. Introducing a breakthrough framework that transforms source code to generate context-rich logs containing precise metadata - file locations, function signatures, line positions, and failure details. This metadata creates direct linkages between operational failures and their code origins. Built on language-neutral architecture using intermediate representation techniques, the system works across Java, Python, C#, and other languages within heterogeneous environments. At its core, the framework employs specialized machine learning that analyzes integrated failure contexts to pinpoint root causes and suggest tailored code fixes. Testing activities conducted across web services, embedded systems, including IoT and RIoT devices, mobile applications, and distributed platforms exhibited significant improvements for complicated defects that formerly required considerable manual analysis. And while these solutions primarily address the immediate challenge of diagnosis, they also provide a foundation to develop next-generation systems that perform self-repair while requiring minimal human intervention.