AI Agent: Business WhatsApp RAG Chatbot with OpenAI & Qdrant

Unlock Enhanced Customer Engagement and Support with this AI Agent

This AI Agent transforms your WhatsApp Business account into an intelligent, 24/7 customer assistant. It leverages Retrieval Augmented Generation (RAG) to tap into your company's knowledge base—sourced from documents in Google Drive and vectorized into a Qdrant database. The agent, powered by OpenAI's advanced language models (like gpt-4o-mini), can understand customer queries in natural language and provide accurate, context-aware responses directly within WhatsApp.

Imagine your customers getting instant answers about your electronics store's products, troubleshooting help, or support information, all without human intervention for common queries. This not only boosts customer satisfaction but also frees up your team to handle more complex issues.

Key Features & Benefits

• AI-Powered WhatsApp Chatbot: Automates customer interactions on a widely used platform, providing immediate responses.
• Retrieval Augmented Generation (RAG): Delivers answers grounded in your specific business documents, ensuring accuracy and relevance. Say goodbye to generic chatbot replies!
• Custom Knowledge Base: Easily connects to your Google Drive folder, automatically processing documents (e.g., product manuals, FAQs, policies) into a searchable Qdrant vector store.
• Advanced AI Model: Utilizes OpenAI's chat models (e.g., gpt-4o-mini) for nuanced understanding and human-like conversation.
• Configurable AI Persona: The agent's system prompt is fully customizable, allowing you to define its role, tone, and specific guidelines (e.g., acting as an assistant for an electronics store as per the example).
• Conversation Memory: Remembers previous parts of the conversation for more coherent and contextual interactions using Window Buffer Memory.
• Scalable Support: Effortlessly handle a growing volume of customer inquiries, improving efficiency and reducing response times.
• Streamlined Setup: The n8n workflow includes a setup process for creating your Qdrant collection and vectorizing your knowledge base.

• An n8n instance (Cloud or self-hosted).
• WhatsApp Business API access and credentials (Phone Number ID, Meta App Access Token).
• OpenAI API Key with access to a suitable model (e.g., gpt-4o-mini).
• Qdrant instance (URL and API Key if required for authentication in HTTP Request nodes).
• Google Drive account with a folder containing your knowledge base documents (PDFs, TXTs, etc.).
• A Meta for Developers App configured for WhatsApp.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Initial Configuration (Node Credentials & Placeholders): a. Update placeholder values like QDRANTURL and COLLECTION name in the 'Create collection', 'Refresh collection', and 'Qdrant Vector Store' nodes. b. Set credentials for: 'OpenAI Chat Model' & 'Embeddings OpenAI' (OpenAI API Key), 'Qdrant Vector Store' (if Qdrant requires API key for its REST API via HTTP nodes), 'Get folder' & 'Download Files' (Google Drive), 'Send' & 'Only message' (WhatsApp).
4. Step 1: Create Qdrant Collection (One-time manual run): a. Configure the 'Create collection' HTTP Request node with your Qdrant URL and desired collection name. b. Execute the 'When clicking ‘Test workflow’' manual trigger. This should run the 'Create collection' node.
5. Step 2: Vectorize Knowledge Base (Manual run after setup): a. Ensure the 'Refresh collection' HTTP Request node points to your QDRANTURL and COLLECTION. b. Configure the 'Get folder' Google Drive node to point to the folder containing your knowledge documents. c. The 'Download Files', 'Default Data Loader', 'Token Splitter', and 'Embeddings OpenAI' nodes will process these documents. d. The 'Qdrant Vector Store' node (in 'insert' mode) will store the vectorized documents. e. Execute the 'When clicking ‘Test workflow’' manual trigger again to run the 'Refresh collection' path, which populates your Qdrant collection.
6. Step 3: Configure Meta Webhook for WhatsApp: a. In the 'Verify' Webhook node, copy its Test URL (Method: GET). b. Go to your Meta for Developers App -> WhatsApp -> Configuration. Paste the URL as the Callback URL. Add a Verify Token (any string, e.g., 'n8n-whatsapp-bot'). Enter the same Verify Token in the 'Verify' Webhook node's settings in n8n (if not already configured to echo hub.challenge, ensure it does). c. Meta will send a verification request. The 'Respond to Webhook' node (connected to 'Verify') is set up to echo back hub.challenge, completing verification. d. Subscribe your webhook to message events in the Meta App.
7. Step 4: Configure AI Agent & Live Operation: a. The 'Respond' Webhook node (Method: POST, same path as 'Verify') receives incoming WhatsApp messages. b. Review the 'AI Agent' node's systemMessage. Customize it to fit your business persona and specific instructions for the AI. c. Ensure the 'OpenAI Chat Model' is set to your preferred model and 'Window Buffer Memory' is configured if you need to adjust memory settings. d. The 'Send' WhatsApp node is configured to send the AI's response back to the user.
8. Activate the workflow to start receiving and responding to WhatsApp messages.