Fighting Logs with Log Detective - Automated Log Analysis for Package Build Failures

Learn how to automate log analysis and diagnose package build failures using Log Detective, an innovative system that combines Template Mining and Large Language Models to streamline the debugging process for software maintainers. Discover how this 35-minute conference talk from DevConf.CZ 2025 addresses the time-consuming challenge of identifying root causes in Copr and Koji build failures that traditionally requires extensive expertise. Explore the technical architecture behind Log Detective's automated analysis system that processes log files through https://logdetective.com/explain, requiring only a simple paste of your log file link for comprehensive explanations within minutes. Understand how existing models like Granite, Llama, and Mistral provide valuable feedback while learning about the development of specialized models trained on log annotations under the CDLA-Permissive-2.0 license to better understand specific ecosystem requirements. Gain insights into the technical decisions behind creating a service designed to deliver near-perfect explanations and solutions comparable to guidance from experienced maintainers. Watch demonstrations of the tool in action and discover why this presentation is essential for software builders looking to optimize their debugging workflows and reduce time spent on log analysis.