An AI-Powered Framework for Intelligent Log File Analysis

Automated log file analysis plays a critical role in ensuring the reliability and maintainability of modern software systems. This paper presents an enhanced AI-based tool designed to interpret system logs through a structured, prompt-oriented approach. Building upon foundational work, the proposed framework integrates advanced preprocessing techniques, targeted prompt design, and a refined evaluation pipeline. System logs are classified as either passed or failed based on extracted error patterns, followed by a three-stage querying mechanism aimed at identifying error presence, extracting primary error messages, and classifying error types. A fine-tuned transformer-based model, built on the LLaMA 2 architecture, is employed to generate accurate, context-aware responses. The system further includes an automated reporting module with detailed visualization components—such as accuracy heatmaps, loss distributions, and confidence analysis—to support interpretability and performance tracking. Experimental results demonstrate an overall accuracy exceeding 90%, with consistent performance across diverse log scenarios. The tool is lightweight, explainable, and adaptable, making it suitable for integration into real-time monitoring and DevOps pipelines.