The scenario
SkyHelp is a clean introduction to RAG on real-world policy pages: scrape and chunk them, embed them, retrieve, and answer with citations. Then add conversation memory and a simple feedback button to find the questions it gets wrong.
System architecture
01 Policy pages Scraped
02 Chunks + embeddings FAISS
03 Chatbot Retrieve + answer
04 Feedback Find weak answers
High-level flow of SkyHelp, from input to output.
What you will learn
- Scraping and chunking web content
- Embeddings and vector search
- Answer citations and feedback loops
What you will build
- Policy scraper
- RAG chatbot
- Feedback logging
Build plan
- Scrape
- Index
- Chat
- Citations
- Feedback