Build a self-updating RAG system with OpenAI, Google Gemini, Qdrant and Google Drive

This workflow implements a Retrieval-Augmented Generation (RAG) system that integrates Google Drive and Qdrant.

This setup creates a powerful, self-updating knowledge base that provides accurate, context-aware answers to user queries.

Key Advantages

• Automated Knowledge Base Updates No manual intervention is required—documents in Google Drive are automatically synchronized with Qdrant.

• Efficient Search and Retrieval Vector embeddings enable fast and precise retrieval of relevant information.

• Scalable and Flexible Works with multiple documents and supports continuous growth of your dataset.

• Seamless AI Integration Combines OpenAI embeddings for vectorization and Google Gemini for high-quality natural language answers.

• Metadata-Enhanced Storage Each document stores metadata (file ID and name), making it easy to manage and track document versions.

• End-to-End RAG Pipeline From document ingestion to AI-powered Q&A, everything is handled inside one n8n workflow.

How It Works

This workflow implements a Retrieval-Augmented Generation (RAG) system that automatically processes, stores, and retrieves document information for AI-powered question answering. Here’s how it functions:

1. Document Processing & Vectorization:

   • The system monitors a specified Google Drive folder for new or updated files.
   • When a file is added or modified, it is downloaded and split into manageable chunks using a Recursive Character Text Splitter.
   • Each chunk is converted into vector embeddings using OpenAI's embedding model.
   • These vectors, along with metadata (file ID, file name), are stored in a Qdrant vector database.

2. Automatic Updates:

   • The workflow includes a mechanism to delete old vectors associated with an updated file before inserting the new ones, ensuring the knowledge base remains current.

3. Query Handling & Response Generation:

   • When a user sends a chat message (via a chat trigger), the system:
     • Retrieves the most relevant document chunks from Qdrant based on the query's semantic similarity.
     • Uses a Google Gemini language model to generate a context-aware answer grounded in the retrieved documents.
   • This provides accurate, source-based responses instead of relying solely on the AI's internal knowledge.

4. Initial Setup & Maintenance:

   • The workflow can be triggered manually to create the Qdrant collection or clear all existing data.
   • It processes all existing files in the Drive folder during initial setup, populating the vector store.

Set Up Steps

To configure this workflow, follow these steps:

STEP 1: Create Qdrant Collection

• Replace QDRANTURL in the "Create collection" and "Clear collection" nodes with your Qdrant instance URL (e.g., http://your-qdrant-host:6333).
• Replace COLLECTION with your desired collection name.
• Ensure the Qdrant API credentials are correctly set in the respective HTTP Request nodes.

STEP 2: Configure Google Drive Access

• Set up OAuth credentials for Google Drive to allow the workflow to:
  • Read files from a specific folder .
  • Download files for processing.
• Update the Folder ID in the "Search files" and "Update?" trigger nodes to point to your target Google Drive folder.

STEP 3: Set Up AI Models

• Configure the OpenAI API credentials in the Embeddings nodes for generating text embeddings.
• Configure the Google Gemini (PaLM) API credentials in the Google Gemini Chat Model node for generating answers.

STEP 4: Configure Metadata

• The system automatically attaches metadata (file_id, file_name) to each document chunk. This is set in the Default Data Loader nodes.
• This metadata is crucial for identifying the source of information and for the update mechanism.

STEP 5: Test the RAG System

• The workflow includes a chat trigger ("When chat message received") for testing.
• Send a query to test the retrieval and answer generation process.

Need help customizing?

Contact me for consulting and support or add me on Linkedin.