Provide ecommerce support with Qdrant RAG, WooCommerce, and human backup

This workflow is a complete AI-powered customer support automation for WooCommerce e-commerce websites. It combines conversational AI, Retrieval-Augmented Generation (RAG), vector search, WooCommerce integration, and human escalation into a single intelligent support system.

The workflow allows customers to interact with an AI assistant through a website chat interface, providing real-time support, product discovery, policy assistance, and personalized shopping guidance.

---

Key Advantages

1. ✅ Intelligent Customer Support

The chatbot can answer customer questions instantly, reducing response times and improving customer satisfaction.

Benefits:

• 24/7 automated assistance
• Faster customer response times
• Reduced support workload
• Consistent customer experience

2. ✅ Real-Time WooCommerce Integration

The workflow connects directly to WooCommerce to retrieve live product data.

Capabilities:

• Product search
• Availability checks
• Price retrieval
• Product recommendations
• Category browsing

This ensures customers always receive accurate and up-to-date information.

3. ✅ RAG-Powered Knowledge Base

The workflow uses Qdrant Vector Store and OpenAI embeddings to create a semantic knowledge base from company documents stored in Google Drive.

Advantages:

• AI can answer questions about:

• shipping policies

• returns and refunds

• FAQs

• store information

• sizing guides

• promotions

• Context-aware responses

• Better accuracy than traditional keyword search

• Easy document updates through Google Drive synchronization

4. ✅ AI Guardrails and Safer Conversations

The workflow includes dedicated AI guardrails before the AI agent processes customer messages.

Benefits:

• Safer interactions
• Reduced hallucinations
• Better brand protection
• More controlled AI behavior
• Prevention of off-topic conversations

5. ✅ Personalized Shopping Experience

The AI agent acts like a virtual shopping assistant.

Capabilities:

• Product recommendations
• Outfit suggestions
• Personalized guidance
• Product comparisons
• Contextual conversations using memory

This creates a premium customer experience similar to an in-store assistant.

6. ✅ Conversation Memory

The workflow includes a buffer memory system that maintains context during conversations.

Advantages:

• More natural conversations
• Better understanding of customer preferences
• Improved multi-step interactions
• Personalized responses across the session

7. ✅ Human Escalation System

When the AI cannot fully resolve a request, the workflow automatically escalates the conversation to human support via Gmail.

Benefits:

• Seamless customer handoff
• Better customer satisfaction
• Reduced frustration
• Support continuity
• Full conversation transcript included

---

8. ✅ Scalable and Modular Architecture

The workflow is built using modular n8n nodes and reusable AI tools.

Advantages:

• Easy to customize
• Easy to scale
• Simple maintenance
• Flexible integrations
• Suitable for multiple e-commerce industries

Ideal Use Cases

This workflow is ideal for:

• Fashion e-commerce stores
• Luxury boutiques
• WooCommerce websites
• Online retail businesses
• Customer support automation
• AI shopping assistants

---

How it works

This workflow implements an AI-powered customer support chatbot for an e-commerce store (Fashionart). It integrates with webiste chatbot (chat trigger), Qdrant (vector store), Google Drive (knowledge base), WooCommerce (product data), and Gmail (human escalation).

1. User message received – The workflow starts when a customer sends a message via the chat trigger.
2. Guardrails applied – The input is first passed through a Guardrails node to enforce safety and content policies.
3. AI Agent processes the request – An E-Commerce Customer Support AI Agent (LangChain agent) decides which tool to call based on the user’s intent:

• rag_search – Retrieves company policies, FAQs, and static knowledge from Qdrant (populated from Google Drive documents).
• get_product – Fetches live product details from WooCommerce.
• get_many_products – Searches/multiple products from WooCommerce.
• get_human_support – Escalates to staff by sending a transcript + customer phone/email via Gmail.
• Calculator – Optional tool for simple calculations.

1. Memory context – A Window Buffer Memory node keeps conversation history (last 10 messages) for contextual replies.
2. Response returned – The agent’s output is sent back to the customer via the chat node.

> The workflow also includes a manual ingestion branch (Test workflow → Google Drive → Qdrant) to vectorize documents into the knowledge base.

---

Set up steps

1. Create Qdrant collection

• Ensure Qdrant is running at http://qdrant_jush:6333
• Run the Create collection HTTP request node (or manually create a collection named fashionart with vector size 1536, Cosine distance).

1. Configure Google Drive

• Set the folder ID in the Get folder node (contains all product/policy documents).
• Ensure the Download Files node can access the drive (authenticate with OAuth2).

1. Set up OpenAI embeddings

• Add your OpenAI API key in the Embeddings OpenAI and Embeddings OpenAI2 nodes.
• Same for OpenAI Chat Model1 (or switch to the existing Google Gemini Chat Model with proper credentials).

1. Configure WooCommerce tools

• Add WooCommerce credentials (API key, secret, store URL) in get_product and get_many_products nodes.

1. Set up Gmail for human escalation

• Authenticate the get_human_support Gmail node.
• Update the recipient email ([email protected] to the actual support email).

1. Adjust agent system prompt

• Review the system prompt inside the E-Commerce Customer Support AI Agent node – it contains the full assistant behavior, tone, and tool usage logic.
• Change any store-specific details (name, address, examples).

1. Activate chat trigger

• Deploy the workflow.
• Use the webhook URL from the When chat message received node to connect to javascript chatbot or any chat frontend.

1. Populate the vector store

• Click Test workflow (manual trigger) to run the ingestion branch – this will read documents from Google Drive, split them, embed them, and store them in Qdrant.

1. Set workflow to active – Change "active": false to true in the JSON or via n8n UI.

---

👉 Subscribe to my new YouTube channel. Here I’ll share videos and Shorts with practical tutorials and FREE templates for n8n.

---

Need help customizing?

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