Chat with documents via RAG: Google Drive to GPT-5 with Supabase vector database

📄 n8n RAG Ingestion & Query Workflow

Overview This workflow is your all-in-one pipeline to turn any document into a powerful searchable knowledge base using RAG (Retrieval-Augmented Generation). From the moment a file lands in your Google Drive, it’s automatically processed, understood, and made ready for instant AI-powered answers.

If you’re looking to unlock hidden value in your files and get answers in seconds instead of hours, this workflow is the foundation you need.

---

What It Does for You - 📥 Automatic Ingestion – New files in a designated Google Drive folder are instantly picked up. - 🔍 OCR Extraction – Extracts all text, whether it’s plain or inside tables. - 📚 Vector Database Storage – Keeps your documents in Supabase for lightning-fast semantic search. - 🧩 Smart Chunking – Each page becomes a single chunk for better understanding. - 💡 AI-Powered Answers – Ask questions in natural language and get precise, context-aware responses. - 🧠 Persistent Memory – Remembers previous chats for more coherent conversations. - ⚡ GPT-5 Intelligence – Uses OpenAI’s most advanced model for deep, accurate answers.

---

How It Works 1. Detect → Watches your Google Drive folder for new files. 2. Extract → Uses Mistral AI to read all text, including tables. 3. Chunk → Splits content so one page = one chunk for better context. 4. Embed → Generates vector embeddings with OpenAI for semantic search. 5. Store → Inserts processed content into Supabase. 6. Retrieve & Answer → When you ask, the system searches the database and passes the results to GPT-5. 7. Remember → Stores conversation history in Postgres for continuity.

---