Build a Business WhatsApp AI RAG Chatbot with OpenAI & Qdrant

How it works

This n8n workflow creates a sophisticated AI-powered chatbot for Business WhatsApp. It leverages a Retrieval-Augmented Generation (RAG) approach to provide intelligent and context-aware responses.

The workflow begins by setting up a Qdrant collection and clearing existing data, preparing it for new information.

It then accesses your Google Drive to retrieve documents from a specified folder ('test-whatsapp'). These documents are downloaded, converted to plain text, and then processed.

The core RAG process involves:

1. Document Loading: Loading files from Google Drive.
2. Text Splitting: Tokenizing the documents into manageable chunks using Token Splitter.
3. Embeddings: Generating vector embeddings for these chunks using Embeddings OpenAI.
4. Vector Storage: Storing these embeddings in Qdrant using Qdrant Vector Store.

When a message is received via a WhatsApp webhook (Respond node), the workflow checks if it's a user message. If it is, the message is sent to the AI Agent node.

The AI Agent node, configured with OpenAI Chat Model (GPT-4o-mini) and a detailed system prompt, retrieves relevant information from the Qdrant vector store (acting as the knowledge base) and generates a response. Window Buffer Memory helps maintain conversation context.

Finally, the generated response is sent back to the user via WhatsApp using the Send node.