*LIVE BUILD* n8n AI Chatbot + RAG from Youtube Transcripts (Supabase, Google API & GPT 4.1)
🔥 Join The AI Forge Community 👇
https://www.skool.com/the-ai-forge
In this stream, I evaluate and walk through the complete process of building an AI-powered chatbot designed to interact with my YouTube video content. I demonstrate how to make your entire video library searchable, allowing the chatbot to retrieve specific information and, crucially, provide the exact source link with a precise timestamp from the video where that information was mentioned.
I cover the entire pipeline, starting from fetching video transcripts using the YouTube API within an n8n workflow, processing and structuring this data (including a deep dive into using a code node and getting assistance from LLMs like Gemini and Grock for data transformation), and then embedding these transcript chunks into a Superbase vector database. I also show how to incorporate metadata like video titles and thumbnails.
The second major part involves setting up an AI agent, equipped with a Superbase vector store tool and guided by a carefully crafted system prompt, to understand user queries, fetch relevant segments from the transcripts, and respond with the answer along with the timestamped YouTube URL and thumbnail for easy reference.
Timestamps:
00:00 Project Introduction
01:33 Core Chatbot Architecture
03:52 Fetching Video Data and Transcripts
18:17 Structuring Transcript Data with Custom Code & LLM Assistance
48:51 Building the AI Retrieval Agent in n8n with Superbase Tooling
01:06:48 Live Testing: Querying the Chatbot and Verifying Timestamped Source Links
01:11:11 Discussion: Production Considerations, Edge Cases, and Future Enhancements
✨n8n Cloud (Affiliate Link—Using it supports me!): https://n8n.partnerlinks.io/c2uubbc6vgax
The Recording Software I use (Canvid): https://aibuilders.short.gy/canvid
📱 X: https://x.com/leonardogrig
💼 LinkedIn: / leonardogrig
