rasel1510/RAG-For-Physical-Activity-Guidline — reverse-engineered prompt
Reverse engineered prompt
Build me a polished little web app called ActiveGuide AI that lets someone ask questions about the Physical Activity Guidelines for Americans and get clear, trustworthy answers backed by the actual source text.
I want it to feel like a friendly chat experience with a nice welcome screen, a few suggested starter questions, responsive design, typing feedback, and clean message bubbles. The important part is that it should only answer from the official guideline content, not make things up, and every answer should include page citations like Page 24 plus a way to expand and see the matching source passages and how relevant they were.
Please use the included guideline chunks and set up the full RAG flow, embedding search, top result retrieval, and answer generation through OpenRouter. Add the one time ingestion step to cache embeddings locally so the app stays fast after setup. Make sure it runs locally with simple env setup for the API key. If anything is unclear, check current docs online and make sensible choices.
Want more depth? Deep Reverse