Voice AI customer support for WooCommerce using VAPI, GPT-4o & Gemini with RAG

This workflow integrates a Retrieval-Augmented Generation (RAG) system with a post-sales AI agent for WooCommerce. It combines vector-based search (Qdrant + OpenAI embeddings) with LLMs (Google Gemini and GPT-4o-mini) to provide accurate and contextual responses.

Both systems are connected to VAPI webhooks, making the workflow usable in a voice AI assistant via Twilio phone numbers.

The workflow receives JSON payloads from VAPI via webhooks, processes the request through the appropriate chain (Agent or RAG), and sends a structured response back to VAPI to be read out to the user.

---

Advantages

✅ Unified AI Support System: Combines knowledge retrieval (RAG) with transactional support (WooCommerce). ✅ Data Privacy & Security: Enforces strict email/order verification before sharing information. ✅ Multi-Model Power: Leverages both Google Gemini and OpenAI GPT-4o-mini for optimal responses. ✅ Scalable Knowledge Base: Qdrant vector database ensures fast and accurate context retrieval. ✅ Customer Satisfaction: Provides real-time answers about orders, tracking, and store policies. ✅ Flexible Integration: Easily connects with VAPI for voice assistants and phone-based customer support. ✅ Reusable Components: The RAG part can be extended for FAQs, while the post-sales agent can scale with more WooCommerce tools.

---

How it Works