LogLLM: Real-Time Windows Log Analysis System Using Vector Embeddings and LLMs
Year: 2025
Type: article
Abstract: We present LogLLM, an advanced real-time system that synergistically combines Windows log ingestion, semantic vector embeddings, and large language models (LLMs) for intelligent log analysis. LogLLM transcends traditional log analysis approaches by enabling natural language querying, sophisticated semantic search, and highly interpretable results through our novel integration of streaming log collection, vector database technology (ChromaDB), and LLM-based retrieval-augmented generation (RAG). Our architecture is informed by recent advances in LLM-based log analysis [3], embedding-driven anomaly detection [2], and domain-adaptive LLMs [4]. Extensive experiments demonstrate that LogLLM achieves exceptional accuracy and responsiveness for operational log analysis tasks, outperforming existing solutions by a significant margin in both precision and query response time.
Author: Yuva T, Taneesha S M, Sushrut R Nayak, Yadamreddy Navaneeth, Prof. Anitha Sandeeep
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6893178151
FWCI:
Cites:
Cited by:
Related to: