I need an end-to-end time-series forecasting solution that predicts my restaurant’s monthly revenue. You will work with a trusted CSV dataset that I will supply at project start. The stack is already chosen: Streamlit for the interactive UI, FastAPI as the service layer, a SQLite store for both raw data and model artefacts, Scikit-learn for data preparation, and a Temporal Fusion Transformer (TFT) as the core model. Visual insights should be delivered through Matplotlib or seaborn charts embedded directly in the Streamlit app. Key points to hit • Cleanly ingest the CSV, write it into SQLite, and expose CRUD endpoints through FastAPI. • Build the full forecasting pipeline—feature engineering, training/validation splits, hyper-parameter tuning, and model persistence. • Serve monthly forecasts via a REST endpoint and display them in Streamlit alongside historical trends, confidence intervals, and error metrics (MAE, RMSE, MAPE). • Provide clear, commented code plus a brief README so I can retrain or extend the model later.