A sequential Retrieval-Augmented Generation (RAG) system that allows you to ask questions about any website and get grounded answers using Google Gemini 2.0. It extracts website text, creates embeddings, retrieves relevant context, and generates answers with source citations.
- Extracts text from
p,h1,h2,h3tags while ignoring scripts, styles, headers, footers, and navigation. - Converts webpage text into embeddings using Cohere embeddings.
- Builds a FAISS vector store for fast retrieval of relevant content.
- Generates answers using Google Gemini 2.0 Flash model.
- Provides:
- Grounded answers
- Source citations
- Performance metrics (retrieval & generation time)
- Fully sequential; no Flask or web server required.
- Python 3.10+
- Install dependencies:
pip install requests beautifulsoup4 langchain-community-cohere langchain-text-splitters faiss-cpu google-generativeai