Multi-Source Time Series Forecasting Pipeline with Uncertainty

Design a comprehensive Python forecasting pipeline using pandas and Prophet that can integrate multiple heterogeneous data sources (economic indicators, seasonal trends, external variables) and generate probabilistic predictions with confidence intervals. The solution must handle missing data, automatically detect and manage time series anomalies, and produce both point estimates and prediction ranges. Include robust error handling, logging mechanisms, and a modular architecture that allows easy swapping of forecasting algorithms.