Step by Step: RAG AI Agents Got Even Better

[@ЕгорЗахаров-д4й]

This is really inspiting! Thank you for keeping updated with the latest tech

[@carlosf3421]

Concise and straightforward. Excellent walkthrough and breakdowns. Thank you.

[@thesaltyone4400]

I've been working on this myself using your KZbin data as a base for Gemini and now here you are delivering.

[@GregPeters1]

Good stuff, Nate. Also, i like the Jazz background music for Sunday builds.

[@jozzzzen]

very nice tutorial, well explained, ty!

[@vladimirrumyantsev7445]

Thanks for such valuable content, you're the best much appreciate you 🤝👍

[@nateherk]

Thank you!

[@ClowBwas]

Great video Nate!

[@nateherk]

Thank you!

[@parthchandak7944]

How do you deal with multiple files uploaded at once?

[@MrDenisJoshua]

I wonder if is possible to make a flow that utilize the PGVector to memorize the important personal thinks during the conversation with AI. I don't need the documents, I need that AI decide if an information is important and then update PGVector with this information in order to take is if I'll ask that information next year :-) For example today in a chat session I tell to IA that I just buy a mobile phone model XYZ... next year I wont to ask the IA "witch model of mobile phone I have and how much time I used". Is this possible (locally better) ? do you have a video that is near to what I ask please ? Thanks a lot

[@fathin7480]

How do you handle large unstructured documents that would require large context window to enable more accurate results. For example, would this workflow work effectively for a large text file that contains exported chats from a group chat in WhatsApp. Would I be able to get an accurate result if I prompt the chat in the workflow you have shown to export all the names of people in JSON format

[@contractorwolf]

after your "limit" node couldnt you just feed back into the original "downloading file" node instead of creating a new one?

[@nateherk]

I had played around with this but for simplicity sake I split them up. Was running into issues trying to reference the file ID to put it into the metadata in Supabase. I was playing around with operators like ?? and || but I wanted to keep this tutorial more straightforward. Great point though thank you!

[@noame]

Is there a way to do the same job using Qdrant Vector Store ? I cant find a node to delete a vector collection ? Did some one went through this ?

[@nateherk]

Great question. I am still exploring different Vector Database providers, hopefully I can make a video about Qdrant soon.

[@noame]

@@nateherk Thank you Nate for the quality of this tutorial. Looking forward next ones !

[@ilias8884]

Can we build multiple RAG ai agent in one workflow.

[@nateherk]

What would be the use case? How would you want to interact with each one?

[@gabrielmn89]

Is this faster than just putting the google drive folder into the agent ai tools so it can search the docs and answer questions based on the files?

[@nateherk]

Yes having data in a vector database will be much more efficient especially as you start to continuously add more information

[@HugoCatarino]

If you would indicate budget 0 instead of “we have no money” it would probably be more accurate.

[@parthchandak7944]

I keep getting this error when I run the SQL query on supabase: ERROR: 42710: extension "vector" already exists

[@nateherk]

This is likely because your table has already been created

[@DIY4Profit]

What is the benefit for n8n over make?

[@plusone.network]

1. Open source 2. Self hostsable 3. Community

[@Nuetzt-ja-nix]

Just tested: 'file updated' also triggers on new files

[@nateherk]

Try splitting them into two different workflows or making sure everything is configured correctly within google drive

[@sleepless-nite]

“Got even better,” I thought N8N got some updates or something. Lol😅. How are the hallucinations when you have a complex data?

[@nateherk]

From my POV, they got much better as I haven’t explored this method yet 😅 Hallucinations always happen but that’s where iterating and refining comes into play, a huge percentage of time should be spent here when building out workflows.

[@francisco.garcia]

Monthly cost for running efficient RAG system like this.

[@nateherk]

It is really going to depend on how much data you have, hard for me to give you a ballpark without knowing more

[@BriansRoar]

It'd be helpful if you outlined the total monthly cost of the basic tech stack needed to run a small business with a manager agent that uses 4+ support agents.

[@lucasstellet5644]

First one!

[@nateherk]

🔥