0

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?

Improve this question

1 Answer 1

Reset to default
1

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.

Improve this answer
Sign up to request clarification or add additional context in comments.

3 Comments

Thank you for your comment. I understand that the pre-vector chunked data, which includes the original text passages or documents, is usually stored in a separate database or data source. Is it correct to assume that this database is typically a relational (RDB) or NoSQL-based database?
Yes, that's correct. The pre-vector chunked data, which contains the original text passages or documents, is often stored in a separate database or data source. This database can be either relational (RDB) or NoSQL-based, depending on the specific requirements and preferences of the system architect. Both relational and NoSQL databases can effectively store and retrieve text data for use in retrieval-based systems like RAG.
Thank you for your very clear explanation! I really appreciate it.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.