📄 DocQuery RAG — Automated Drive-to-RAG System

This n8n workflow automates a Retrieval-Augmented Generation (RAG) pipeline that connects Google Drive, Supabase, Slack and OpenAI for document-aware AI chat and intelligent document indexing.

---

⚙️ Key Features

• 🧩 Auto-ingestion pipeline
  Detects new or updated files in Google Drive and indexes them into a Supabase vector database.

• 🧠 RAG chat interface
  Retrieves relevant document embeddings for user queries and generates grounded responses using OpenAI or OpenRouter models.

• 🔎 Semantic search & reranking
  Uses OpenAI embeddings and Cohere reranker to ensure precise document retrieval.

• 🧰 Memory persistence
  Postgres-backed chat memory for contextual continuity across user sessions.

• 🔔 Slack alerts
  Sends success or error notifications to a Slack channel (#ceren-news) for monitoring.

• 🛠 Error recovery
  Automatic error detection and fallback models for continuous reliability.

---

🏗️ Tech Stack

Component	Technology
Workflow Engine	n8n
Vector Database	Supabase
LLMs	OpenAI GPT-4.1, Anthropic Claude via OpenRouter
Embeddings	OpenAI text-embedding-3-large
Reranker	Cohere
File Source	Google Drive API
Notifications	Slack API
Memory	Postgres (Chat Memory Node)

---

🧾 Workflow Overview

🔹 1. File Detection

• Triggers when a new or updated document appears in Google Drive.

🔹 2. Text Extraction & Embedding

• Downloads and converts the document.
• Extracts text and generates vector embeddings via OpenAI.

🔹 3. Vector Storage

• Inserts embeddings into the drive_docs table in Supabase.

🔹 4. Chat Querying

• When a user sends a message, the system retrieves semantically related content from the vector store, reranks, and generates a contextual AI answer.

🔹 5. Monitoring

• Scheduled system health checks and Slack alerts for failures.

---

🚀 How to Use

1. Import the workflow file into your n8n instance.
2. Set up credentials for:
   • OpenAI
   • Supabase
   • Google Drive
   • Slack
   • (Optional) OpenRouter & Postgres
3. Adjust folder IDs and table names in the parameters.
4. Activate the workflow.

---

📈 Example Use Cases

• AI-powered internal knowledge assistant
• Auto-indexing system for Drive-based document archives
• Semantic document search bot for company data

---

📬 Author

Ceren Kaya Akgün
🧠 AI & Automation Engineer | Berlin, Germany
🌐 Portfolio • GitHub • Medium

---

🪪 License

MIT License — free to use and adapt for non-commercial or educational purposes.