Give shoppers instant answers and product help in chat. A smart agent understands what the customer wants, searches your store, or replies with store info from documents. Great for online retail teams that want faster replies and more sales.

Incoming chat messages start a guided flow. The text is checked for shopping intent and key details like keywords, price range, and SKU. If it is a product request, the agent pulls items from WooCommerce. If it is a general question, the agent uses a knowledge base made from your Google Drive files and stored in Qdrant. The agent keeps short term memory by session so replies feel natural.

A manual path lets you refresh the knowledge base. It clears the Qdrant collection, downloads files from Google Drive, converts them to text, splits them into chunks, and creates embeddings with OpenAI. This setup reduces response time, keeps answers accurate, and helps customers pick the right product. Set your Qdrant URL and collection, connect Google Drive and WooCommerce, add your OpenAI key, and run tests.

• Chat trigger that listens for new messages and starts the agent flow
• Information extractor that detects shopping intent and pulls keywords, price range, and SKU from the text
• WooCommerce personal shopper tool that searches products using extracted filters like keyword and price
• RAG answers for store info using Google Drive documents embedded into a Qdrant vector store
• OpenAI models for both understanding the request and generating clear replies
• Session memory buffer so the agent remembers the last messages in each chat
• Google Drive import pipeline that lists files, downloads them, converts Docs to text, and splits text into chunks
• OpenAI embeddings that index each chunk and load them into a Qdrant collection
• Manual refresh path that clears the Qdrant collection via HTTP and rebuilds the index

• Reduce manual product lookup from 30 minutes to 2 minutes by routing chat requests to WooCommerce automatically
• Automate up to 80 percent of common store questions using a document based knowledge base
• Improve answer accuracy by 30 percent with up to date Google Drive docs embedded in Qdrant
• Support more concurrent shoppers without adding staff by letting the agent handle routine chats
• Unify product data and store info in one chat experience for a smoother buyer journey

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with OpenAI, Qdrant, WooCommerce and Google Drive. See the Tools Required section above for links to create accounts with these services.
3. OpenAI credentials: Double click each OpenAI node (Chat Model and Embeddings). In the Credential to connect with dropdown, click Create new credential. Paste your API key from your OpenAI account page and save.
4. Qdrant vector store: Open the Qdrant Vector Store nodes. Set your Qdrant URL and collection name. Create a new Qdrant credential when prompted and save. Update the collection name to match your instance.
5. Qdrant delete endpoint: Open the HTTP Request node. Replace the URL with your Qdrant endpoint and collection. In the credentials dropdown, create a new HTTP Header Auth credential and add the required header for your Qdrant API key. Keep Content Type as application/json.
6. Google Drive: Open both Google Drive nodes. In the credentials dropdown, click Create new credential and follow the OAuth steps. Select the drive and folder that hold your store docs. In the download node, keep the conversion option so Google Docs convert to text.
7. WooCommerce: Open the personal shopper node. Create new WooCommerce credentials with your store URL, consumer key, and consumer secret. Test the connection and save.
8. Knowledge base build: Click the When clicking Test workflow node and run a test. Confirm the HTTP Request deletes the collection, then watch files list and download, chunks create, and embeddings load into Qdrant without errors.
9. Chat path: Open the chat trigger node and copy any needed endpoint for your chat surface. Check the Set node so sessionId maps correctly. Ensure the Window Buffer Memory node uses that session key.
10. Extractor tuning: Open the Information Extractor node. Review the schema fields like keyword, min price, max price, and SKU. Adjust the prompt to match your catalog attributes if needed.
11. Functional test: Send a product query like bags under 100 and confirm WooCommerce results appear. Ask a store question like opening hours and confirm a RAG answer comes from your documents.
12. Troubleshooting: If no RAG results, verify the Qdrant collection name and that embeddings were created. If product results are empty, check WooCommerce credentials and the extractor output fields. Review n8n execution logs for errors.