Sakina10/search-quality-intelligence-platform ? reverse-engineered prompt

Build me a search quality monitoring platform that can spot when search results get worse even if everything still looks healthy on the surface.

I want it to generate realistic search event data, validate it, load it into a PostgreSQL warehouse, transform it into clean analytics tables, and then calculate a search quality score from both system signals and user behavior like click through rate, dwell time, pogo sticking, and query reformulation. It should also train a model to predict that score, explain what is driving changes, and flag unusual drops or spikes.

Please include a simple API for real time scoring and a dashboard where I can see trends, anomalies, and sample predictions. Make it work in Docker with a clear local setup, good defaults, and enough tests so I can run it confidently. If you need to check current docs for dbt, Feast, FastAPI, or Streamlit while building it, go ahead and look them up.