PatrickMurray78/rag-chatbot-case-study

This is a case-study of an embeddable Retrieval-Augmented Generation (RAG) customer-support chatbot with a no-code admin panel. It documents architecture, design decisions, and lessons; source code is not included. Who uses it - Site owners and non-technical admins who upload FAQs, knowledge bases or product catalogs and configure bot behavior/appearance. - End users on host sites who ask questions and receive streamed, knowledge-grounded answers. Key features - Document ingestion with structural semantic chunking. - Vector retrieval (FAISS) with distance-based confidence threshold. - Three-mode prompt routing: high-confidence, cautious fallback, and product-aware fallback. - Optional product catalog and configurable upsell mode. - Token-by-token streaming for real-time replies. - No-code admin panel (content, behavior, appearance) and an embeddable chat widget. Tech highlights: FastAPI backend, OpenAI embeddings + chat, React frontend; storage via JSON + FAISS.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.