Mazhar26/model-drift-detective ? reverse-engineered prompt

Build me a model drift monitoring app for a telco churn dataset that lets me see when my model starts behaving differently on new data. I want a clean Streamlit dashboard for overview, drift detection, drift explanation, impact on model accuracy, feature importance changes, a drift timeline, and practical recommendations like when to monitor closely or retrain. Back it with a FastAPI service that exposes the drift analysis and returns clear JSON, and make sure the dashboard talks to the API instead of doing everything in one place.

Use the existing dataset in the repo, include simple charts and status summaries, and make the app feel polished and easy to understand for someone who is not a machine learning expert. If you need to check current docs for any library details, go ahead and look them up online. Also make sure it runs locally with Docker and is set up in a way that could later be deployed to Kubernetes.