Need Advice: Moving My GPT Workflow from Google Sheets to RAG?

[amab]

Hey everyone!

I’m working on an automated social media content generation workflow for several brands I manage as part of my business. Right now, I’m using n8n combined with Google Sheets as my database of example posts, categorized by brand.

How it works: When I want to generate a new post (via Telegram), the workflow searches the Google Sheet for matching examples (based on brand, etc.), builds an enriched prompt, and sends it to GPT-4, which generates two new posts in the same style.

Question: I want to improve this system to make it smarter, more scalable, and capable of learning over time.I’m not sure whether I should stick with my standard logic-based workflow in n8n or move to a RAG (Retrieval-Augmented Generation) setup using Supabase + pgvector, which might offer better performance.

Has anyone here implemented something similar?

Thanks in advance

[solomon]

n8n + Supabase for sure.

Way more capable database and super fast.
I’m going to link a sequence of 3 videos from Cole Medin, on YouTube. His RAG setup is the best I’ve seen so far.

• https://www.youtube.com/watch?v=PEI_ePNNfJQ
• https://www.youtube.com/watch?v=T1ZKEmDN8AA
• https://www.youtube.com/watch?v=mQt1hOjBH9o

He’s even storing metadata and, on the third version of the workflow, he also stores the URL of the file.

Additionally, he has implemented many improvements over normal RAG functionality that makes this RAG solution way above average.

His workflows are available on his GitHub.