GN5-2 WP6 Incubator Project Closure Report: LogsLLM: Automated Log Analysis and Management based on Large Language Models (LLMs)

LogsLLM delivered an explainable, LLM-based log-intelligence layer for nmaas operations, with two key achievements: a standardisation pipeline that normalises heterogeneous logs (Icinga, NetBox, Uptime Kuma, firewall logs) into a unified schema, and an analysis component that correlates alerts across systems to produce root-cause-oriented summaries. The solution was validated in four realistic operational scenarios using multiple LLMs (ChatGPT-4.1 mini, Llama 3.1 8B, DeepSeek-R1 14B), yielding accurate and consistent incident summaries suitable for NOC workflows, and supporting faster triage and reduced alert fatigue.