Log Analysis System Based on Large Language Models
Xueliang Gao, Shangru Zhao, Mo Pang, Haoxing Zhang, Yuqing Zhang · 2025
Logs record the running status of the system and important information. Developers can analyze logs to understand the system performance and detect anomalies. Therefore, log analysis plays a crucial role in ensuring the stability of software. However, with the increasing complexity of modern software systems, traditional log analysis methods are difficult to handle the heavy log analysis tasks. To address this challenge, this paper designs a log analysis system with large language models (LLMs) as its core. The system automatically collects log data using logstash and performs in-depth analysis by leveraging the powerful natural language understanding capabilities of LLMs. The implementation follows three steps: first, introduce the research status in the field of log analysis and related knowledge; second, describe the design of the system from the overall architecture to the specific modules; and finally, evaluate the system. Through modular design, the system achieves high efficiency and scalability in log data processing and can meet real-time system demands.