A production incident triggers a spike in error rates and degraded response times-but the cause isn’t obvious. In this session, we walk through a concrete, real-world root-cause analysis scenario using Microsoft Fabric Eventhouse. Live telemetry, application logs, and infrastructure events are ingested into Eventhouse, where embeddings are generated directly and stored alongside time-series data. Vector similarity search is then used to match the current incident to past outages, known failure patterns, and relevant runbooks, while KQL filters narrow results by service, deployment version, and timeframe.

Finally, this retrieved context is fed into a large language model to produce a grounded explanation of the likely root cause-such as a failed deployment, upstream dependency issue, or configuration drift-complete with supporting evidence and suggested remediation steps. You’ll leave with a clear, repeatable pattern for using Eventhouse as the backbone of real-time, AI-assisted root-cause analysis in Fabric.