🚀 Just Built a Custom AI Chatbot in n8n Using RAG! (Inspired by Nate’s Tutorials)

Hey everyone! Super pumped to share a project I’ve been working on—a fully automated AI chatbot in n8n that retrieves answers from Nate Herk’s tutorials! 🎯

I absolutely love AI automation projects like this, and I’m always looking to connect with others who are building cool stuff! Would love to hear your thoughts on this.

What I Built & How It Works

I wanted a chatbot that could accurately answer questions based on Nate’s n8n tutorials—without relying on traditional APIs. So, here’s what I did:

Step 1: Selecting Videos & Structuring Data

First, I picked some of Nate’s best videos on AI agents & n8n automation:

📌 The Ultimate AI Team in n8n – No-Code Personal Assistant

📌 OpenAI Fires Back at Deep Seek with a New Reasoning Model: o3-mini

📌 How to Locally Host Deep Seek R1 for FREE in Under 10 Minutes in n8n

📌 How to Actually Build Agents with Deep Seek R1 in n8n

📌 How I’d Teach a 10-Year-Old to Build AI Agents

📌 How I Wish Someone Explained AI Agents to Me (as a Beginner)

📌 Best Model for RAG📌 AI Agent Prompting Masterclass

📌 5 n8n Tips You NEED to Know

Then, I added these as sources in NotebookLM and used a custom prompt to generate structured text files for each. The output was good but needed refining, so I ran it through ChatGPT to make the structure optimized for embedding.

Step 2: Storing Data in Pinecone Vector Database

To make the chatbot retrieve answers efficiently:

✅ Used Hugging Face’s "sentence-transformers/all-MiniLM-L6-v2" for embedding.

✅ Created a 384-dimension index in Pinecone.

✅ Stored all processed video content in the vector database.

Step 3: Building the AI Agent in n8n

For the chatbot workflow:

✅ Integrated Google Drive in n8n to fetch the structured text files.

✅ Used Claude 3.5 Haiku (20241022) as the AI model (since I had some Anthropic credits).

✅ Implemented Window Buffer Memory for better context retention.

✅ Added “Answer with Vector Store” as a tool to fetch relevant responses.

Final Results 💡

After testing, the chatbot accurately retrieves information from Nate’s tutorials, answering in a structured way. It works fully offline, running on my local n8n instance, so I can’t share access (plus, my Anthropic credits are limited 😅).

I've attached images of both workflows and some Q&A examples with the chatbot. Let me know what you think! Would love to connect with others building AI automation in n8n—let’s share ideas and grow together! 🚀

Drop your thoughts below! ⬇️

0 comments

🚀 Just Built a Custom AI Chatbot in n8n Using RAG! (Inspired by Nate’s Tutorials)