Create RAG Vector Database from Google Drive Documents using Gemini & Supabase

This n8n workflow automates the process of extracting text from Google Drive documents, generating embeddings using Google Gemini, and storing them in a Supabase vector database. This setup is ideal for building Retrieval-Augmented Generation (RAG) systems, allowing you to query your private documents with AI models.

What it does

1. Triggers on Workflow Execution: The workflow is designed to be executed by another workflow, allowing for flexible integration into larger automation processes.
2. Lists Google Drive Files: It retrieves a list of files from a specified Google Drive folder.
3. Loops Through Files: Each file is processed individually in a loop.
4. Loads Document Data: For each file, it loads the document content using the Default Data Loader.
5. Generates Embeddings with Google Gemini: The document content is then passed to the Google Gemini Embeddings model to generate vector representations.
6. Stores Vectors in Supabase: Finally, these embeddings are stored in a Supabase vector database, ready for use in RAG applications.

Prerequisites/Requirements

• n8n Instance: A running n8n instance.
• Google Drive Account: With access to the documents you wish to process.
• Google Gemini API Key: For generating document embeddings.
• Supabase Project: With a configured pgvector extension and a table to store the embeddings.
• Postgres Database: (Optional, but present in the workflow) A Postgres database connection for potential additional data storage or retrieval.

Setup/Usage

1. Import the Workflow: Import the provided JSON into your n8n instance.
2. Configure Credentials:
   • Google Drive: Set up your Google Drive OAuth2 credentials.
   • Google Gemini: Configure your Google Gemini API Key credential.
   • Supabase Vector Store: Set up your Supabase credentials, including the Supabase URL, API Key, and specify the table and embedding column names.
   • Postgres: (If used) Configure your Postgres database credentials.
3. Customize Google Drive Node (ID: 58):
   • Specify the Folder ID from which you want to retrieve documents.
   • Adjust the Limit if you only want to process a subset of files.
4. Customize Supabase Vector Store Node (ID: 1231):
   • Ensure the Table Name and Embedding Column match your Supabase configuration.
5. Activate the Workflow: Once configured, activate the workflow. It will then be ready to be triggered by another workflow.