n8n WhatsApp AI Sales Agent: OpenAI + Gemini RAG

Turn WhatsApp chats into AI-powered, catalogue-grounded sales conversations

This n8n WhatsApp AI Sales Agent workflow listens for new customer messages on WhatsApp, understands them (text, voice, images, and videos), and replies with retrieval-augmented answers grounded in your indexed product brochure PDF—using OpenAI and Google Gemini.

What this workflow does

• Builds a knowledge base (manual run): downloads a product brochure PDF, extracts its text, splits it into chunks, generates OpenAI embeddings, and indexes everything into an in-memory vector store.
• Triggers on incoming WhatsApp messages: loads workflow configuration including models, prompts, WhatsApp phone number ID, and the vector store key.
• Understands different message types:
  • Audio/voice: transcribes audio using OpenAI.
  • Images: describes images using OpenAI Vision.
  • Video: describes videos using Google Gemini.
  • Text: uses the message text directly.
• Answers with RAG + per-customer chat memory: uses an OpenAI chat model with per-customer memory and a vector-store retrieval tool to respond based on the indexed product catalogue.
• Sends WhatsApp replies: returns the agent’s answer via WhatsApp, or a predefined message for unsupported WhatsApp message types.

Use cases

• Automate product questions on WhatsApp (pricing, specs, availability) with answers pulled from your brochure.
• Qualify leads by maintaining per-customer conversation context and guiding customers through the catalogue.
• Accept voice notes, product photos, or short video inquiries and respond in a consistent sales tone.

Technical details

• Integrations/Credentials: WhatsApp Business Cloud (trigger + sending/access), OpenAI (chat, embeddings, transcription, vision), Google Gemini (video analysis).
• n8n flow components (nodes): set, switch, WhatsApp, sticky note, HTTP request, manual trigger.
• Knowledge base setup: configure a public direct URL to your product brochure PDF in the knowledge base settings.