Observability & Evaluation · Software component
Behavioral Baseline Builder
Software componentObservability & EvaluationObservability & Evaluationarc:BehavioralBaselineBuilder
An observability component that derives and continuously refreshes behavioral baselines from recent normal operational history, accounting for growth, seasonality and legitimate operational change.
Responsibility. Builds and keeps current the baselines used to judge agent behaviour as normal or anomalous.
Also known as: Baseline refresh, Rolling-window baseline
Relationships
reads dependency
produces lifecycle
Design guidance
- MUST refresh baselines periodically rather than establishing them once, so legitimate new normals (e.g., post-scaling load) are not flagged.
- MAY weight recent behaviour exponentially to stay current without overreacting to temporary fluctuations.
Quantitative guidance
As stated by the sources; verify before use.
- Worked example uses rolling 24-hour windows (Ch10.4).
- Example new normal: 500 -> 800 requests/hour after infrastructure scaling is not an anomaly (Ch10.4).
Classification
- Patterns
- Rolling-window statisticsExponentially weighted baselineTime-series forecasting with seasonality
- Quality attributes
- Flexibility (ISO/IEC 25010)Maintainability (ISO/IEC 25010)
- Risks mitigated
- Baseline drift producing continuous false positivesBlindness to genuine drift after operators tune out alerts
Sources
- Ch10.4: T. Nguyen, "Human-in-the-Loop," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 10.4. ISBN: 9798244538229.