September 2025 (2025-09) highlights: Delivered runtime-configurable prediction_type for FlowStateForPrediction in granite-tsfm, enabling user-driven selection among quantile, mean, and median predictions to increase forecasting flexibility and control. Refactored modeling_flowstate.py to source prediction_type from user input at runtime rather than from static config, reducing misconfiguration risk and aligning with real-world usage patterns. The change is tracked under commit d726f2203fd8c9cb549392910b6f37f712469f48. No major bugs were reported this month; the focus was on robust feature delivery, code quality, and preparing for release. Overall, the work improves adaptability for forecasting workflows, accelerates experimentation, and supports more informed business decisions through flexible prediction options.