What's a cost effective, no code approach to small-scale RAG?

GPT builders
1

I have a data set of 30-50 pdf academic papers on a google drive. I would like to use ChatGPT to work with this data.

I don’t know how to code and don’t need to have dynamic changes to the data set. It is very complex verbally, not numerical data.

Without a lot of coding or spending a lot of money, how can I do RAG over this data set? I have used Google’s NotebookLM and this serves the purpose but with a lot of errors and weakness in the “reasoning.” I tried to do this with Llamacloud and failed to figure it out, using all my tokens.

How can I do this with ChatGPT without spending a fortune or getting a migrane?

2

Hello there, when i first started this journey i started with No-Code AI Agents constructors, they are pretty simple and easy to use, i think that if they are only for you and not production you can give them a try. They offer a free layer and after the free token you can paid, you can choose the model in the configuration and you can load all your files easily.

Cheers!

1 Like
3

Thanks very much, that’s a good lead. I gave it a try, and it basically works as I requested in my post here. However, the output of the chatbot is quite short and truncated. It refuses to “disclose” which LLM it is using; it says OpenAI is an option in settings but that’s not what it’s using.

I don’t yet know much about this world, but it is clear to see that the limitation is the cost of CPU cycles for computing this kind of output, which they are calling tokens. 30 text documents is not huge in terms of size, but this kind of text takes a lot of computing to get sensible “analysis” with - way more than talking about spreadsheets.

The search goes on. You’re right, a local version just for my own use is sufficient.

4

Hi there, I’m the co-founder of a no-code RAG builder, ragdollai.

For your needs, our platform may be a good solution.

We offer both regular vector RAG and LightRAG (a more scalable version of graphRAG) which may help with your challenges with getting comprehensive analyses.

Get in touch with us at info@ragdollai.io and I can get you set up on a staging account to use more extensive features and give you more personalized advice. We are still in early stage so the public version doesn’t have as many options.

Cheers!