Log Root Cause Analysis Agent for System Incidents

Can a small local language model find an incident in system logs and point at the line that shows it? This project builds a read-only CrewAI crew: an investigator with Drain3 log templates, grep and timeline tools, and a reviewer that reads the cited line and keeps or clears the finding. The model (qwen2.5:3b) runs in local Ollama on a laptop with RTX 2050 4 GB (Ryzen 7, 8 GB RAM).

NOTE: Research and teaching demo on a stored public log. Read-only analysis: the agent reads stored public logs and suggests. It runs no command on any system and no remediation; a person decides. The score uses BGL alert tags as a proxy for root cause: nobody recorded the true root cause of these lines, and BGL_2k is a sample of the full log, so a window is not a real incident timeline.

Category: Department of Computer Science and Engineering / Agent Development

Agent Development Project AG22 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.