shubhanshuvaibhav/regu_sense ? reverse-engineered prompt

Build me a Python app called ReguSense AI that helps compliance teams keep up with AI regulations without getting buried in irrelevant federal notices.

I want a daily pipeline that pulls notices from the Federal Register, filters out the noisy stuff, and keeps only the ones that are actually about AI policy. Use an AI based relevance check to score each notice and drop the low confidence ones. Store the cleaned records and metadata in a SQL warehouse, then turn the useful documents into embeddings and save them in a vector database for search.

Also build a simple Streamlit interface where someone can ask plain English questions like what their audit requirements are for a specific kind of AI system and get a useful answer from the stored regulatory data. Please organize it so the pipeline, transformations, scripts, and app are all separate and easy to run. If anything is unclear, use current docs online and make reasonable choices so it works end to end.