Retrieval-Augmented Generation (RAG) combines a large language model (LLM) with a knowledge retriever, improving accuracy and contextual understanding.
To build an image-based RAG, the image is uploaded and stored in Supabase Storage, and a description is generated using Azure OpenAI.
The description is converted into vector embeddings for search functionality using Azure OpenAI, and the embeddings are saved in Supabase Vector.
A user-friendly interface can be added using Streamlit to make the RAG system more accessible.