Aryan41211/ml-pipeline-monitor — reverse-engineered prompt

Build me a local MLOps monitoring app for small data science teams. I want a clean Streamlit dashboard where I can run training pipelines on built in sklearn datasets, pick a classification or regression algorithm, set a few hyperparameters, and watch each pipeline stage update with logs and progress in real time.

It should save experiment runs to a local SQLite database so I can compare metrics, filter runs, and see charts of results. I also want a simple model registry where I can move models from development to staging to production, plus a data drift page that checks baseline versus current data with KS test and PSI and shows distribution plots. Please include system resource monitoring so I can see CPU and memory while jobs run.

Also add a small FastAPI service with health and predict endpoints that can load the current production model. Make it easy to run locally, preferably with a launcher script and Docker compose, and keep tests working. If anything is unclear, look up the current docs online and make sensible choices.