Trying to build RAG with Documents and LArge CSV files

n8n
1

Hi,

I am trying to build RAG agent to use with large documents (txt files) and Large CSV files. I followed @ColeMedin video’s and able to get document part working. But somehow when i ask it questions about data in CSV it gives wrong answer.

Lets say i import csv file with information for 500 customers including their first name, last name, phone number and address. When i ask question like “What is Linda’s phone number” it doesnt return anything or returns wrong response. I feel it is storing csv as large data chunks instead of storing it as a tabular data so it cant do those queries.

Any one worked with CSV’s with RAG agent?

2

Hey! Got the same issue, even without large csv files.
What LLM you are using?

3

llama3.1 is one i have used. I hope someone here has done this. I can use local DB or Supabase.

4

I assume you are using a smaller variation of Llama 3.1? Usually smaller LLMs don’t do well with larger files for RAG. And smaller models don’t do the best with RAG in n8n in general unfortunately.

5

Check my post about it with Cole answers.

1 Like