Unlock Multimodal RAG Agents in n8n (Images, Tables & Text)

👉 Upgrade your n8n AI Agents with our Advanced RAG workflows https://www.theaiautomators.com/?utm_...

RAG Masterclass -    • n8n RAG Masterclass - Build AI Agents + Sy...  
Cache Augmented Generation Video -    • Will CAG replace RAG in N8N? Gemini, OpenA...  
Hybrid Search Video -    • This Hybrid RAG Trick Makes Your AI Agents...  

Chapters:
0:00 - Overview
0:35 - The Multimodal RAG Process
2:43 - Building a Simple Workflow
3:14 - Setting up Mistral OCR
5:08 - Retrieving OCR Results
9:53 - Vectorizing and Uploading Data
14:44 - Chatting with Your Data
17:29 - Uploading image files to Supabase
26:30 - Merging annotations and file URLs

In this video, I'll show you how to build a powerful multimodal RAG agent capable of indexing and analyzing text, images, and tables from complex PDFs at scale.

I'll walk you through the entire process, starting with how I use a powerful OCR API to extract data and annotate media from documents. We'll be using Mistral's OCR for this, which provides information in a markdown format that is LLM-friendly. I'll explain how this process not only extracts images but also uses a vision model to analyze and understand the content of those images, giving us deep context. We will then store this data, including the images and their annotations, in Supabase. I will guide you through the process of chunking this data and using an embedding model to store it in a vector database.

Once our data is indexed, I'll demonstrate how to build an AI agent using n8n to chat with this data. I will show you how to set up the agent to query the Supabase vector store and use a large language model like GPT-4 to generate responses. A key part of this is enabling the agent to render the indexed images directly in its responses, making the output far more effective and informative. I will cover how to set up the necessary HTTP requests, handle API keys, and process the data to get it ready for our agent. I will also walk you through the code needed to integrate the image annotations directly into the markdown. Finally, I will show you how to build a complete workflow that uploads the files to Supabase storage and makes them available to the vector database for retrieval.

I hope this video helps you build your own advanced RAG agents.