Advice on building a review-based RAG chatbot

Hello. I’m looking for advice on how to approach a RAG app I want to build.

I want to build a chatbot that allows you to get insights from user reviews that I’ve scraped. I have some metadata like reviewer name, reviewer rating, review date, and review link. And I also have the review text.

Would I only store the review text as an embedding? I want to link the review text to the review link as the source of the review, just like Perplexity shows links to its sources when you search. I want queries like this to work:
“What are the user complaints about ABC feature in the product?”
“What are the 3 most liked features in the product?”
“Return the last 10 reviews with 4 stars”

I’m thinking of using Supabase as a backend with pgvector for the embedding column. And then storing the rest of the data in separate columns.

What do you think?