Contextual Anomaly Detection Pipeline for Time Series Data

Design a modular anomaly detection system that can process time series data from multiple sources, implementing advanced statistical and machine learning techniques. The pipeline should support pluggable detection algorithms (Z-score, Isolation Forest, DBSCAN), handle multi-dimensional data, provide configurable sensitivity thresholds, and generate actionable insights. Include robust error handling, support for streaming and batch processing, and a flexible output format for integration with monitoring systems.