Website Q&A Chatbot with RAG, OpenAI GPT-4o-mini & Supabase

This workflow solves the problem of static websites lacking interactive Q&A by automating the creation of a Retrieval-Augmented Generation (RAG) chatbot. It starts with a form trigger accepting a website URL, fetches HTML content via HTTP requests, cleans it by removing scripts and ads, and parses it into smaller, manageable text chunks using function nodes.

Next, it generates vector embeddings for each chunk using OpenAI's GPT-4o-mini model and stores them along with metadata in Supabase's vector database through API calls or the native Supabase node. For user interactions, a chat trigger receives questions, converts them into embeddings, performs similarity searches in Supabase to retrieve relevant content chunks, and feeds this context plus the query to OpenAI for generating accurate, grounded responses.

Benefits include real-time, context-aware answers reducing user drop-off, easy deployment on your site, and scalability for any website. Use cases span customer support, FAQ bots, knowledge bases, e-commerce product queries, and internal docs search, saving hours of manual content curation and boosting engagement.