n8nDemo Build

RAG-Based AI Agent

A chatbot that stays in sync with your Google Drive

Supabase vector storeOpenAI embeddingsAuto-sync on create, update, delete

The Problem

Chatbots trained on company documents go out of date the moment a file changes. Someone has to re-upload everything, or the bot gives old answers.

How It Works

01Drive file changes
02Update vector store
03User asks question
04Agent searches docs
05Grounded answer

What I Built

Chat agent on OpenAI that searches a Supabase vector store as a tool
File Created: downloads and embeds new documents
File Updated: deletes the old vectors, then re-embeds the new version
File Deleted: removes that file's vectors so the bot stops quoting it

Tools Used

n8n

OpenAI

Supabase

Google Drive

What Was Tricky

Most RAG demos only add documents. The hard part is keeping the knowledge base accurate, so updated files replace their old chunks instead of duplicating them, and deleted files disappear from the bot's answers.

Where This Fits

Internal help desksCustomer support teamsCompanies with policies or SOPs in Drive