About this talk
This talk, presented by Matthew Khouzam from Ericsson, delves into AI-powered trace analysis by incorporating the Model Context Protocol (MCP) with the Trace Server Machine Learning Library (TMLL). The session details how to enhance trace analysis workflows using AI through a structured architecture that exposes trace analysis functions as callable tools, facilitating natural language interactions with extensive trace datasets. It addresses the challenges in traditional trace data analysis and emphasizes combining deterministic analysis methods with AI orchestration to achieve scalable and cost-effective workflows. The talk also covers various implementation details, including debugging techniques with Eclipse Theia and tools for ensuring cross-agent interoperability.
