I believe that when using an LLM with a Retrieval-Augmented Generation (RAG) approach, the results retrieved from a vector search must ultimately be presented in text form. Otherwise, the prompt would just contain a series of numbers (vectors), which would be meaningless. I assume that the pre-vector chunked data needs to be stored somewhere within the vector database. Is this usually maintained within the vector database itself?

>Solution :

In a RAG system, the vector database stores only numbers that represent the text, not the text itself. The actual text is kept in a different place. When you search for something, the system uses these numbers to find the relevant text and then presents it to you.

In a typical RAG system, the vector database does not maintain the pre-vector chunked data. Instead, the vector database stores only the vector representations of the text data. The pre-vector chunked data, which includes the original text passages or documents, is usually stored separately in another database or data source. When a retrieval is performed using vectors, the system retrieves the corresponding pre-vector chunked data from this separate source based on the vectors retrieved from the vector database.