Building RAG agent, need ideas? · AI Automation Society

Building RAG agent, need ideas?

Hello Friends,

I am building an RAG agent and need some ideas about which LLM and Vector space to use.

In short, I will upload reports (finance and business-related news, developments, and opinions published daily) of 3 to 10 pages each in PDF format. The content will be mostly text with some charts and tables. I will upload 40 to 60 PDFs daily and will upload 1000PDF for the start.

My agent will basically answer simple questions based on the content in PDFs (the most recent content is most important). For example, "What are the recent opinions on the effects of tariffs on corporate taxes?"

Which vector space should I use? Supabase? pinecone? Chrome? or any other one?

Which LLM will be performing best (with the best cost efficiency)?

Thank you in advance.

1 comment