WhatsApp Voice and Vision Concierge

A system that works while you are not looking

This concierge lets customers talk to your business however they prefer on WhatsApp. They can type, send a voice note or send a photo of a product. The system transcribes voice, describes images, looks up real product data from a Google Sheets catalog and answers with accurate information, images and next steps. It remembers returning customers through Supabase, so it can greet them by name and recall previous preferences. It is built for ecommerce and service businesses that want fast, accurate WhatsApp support without adding staff.

A WhatsApp trigger receives every inbound message. A Switch detects the type. Text goes directly to the AI agent, voice notes are sent for transcription and then forwarded as text, and images are uploaded and described with a vision model. The agent has two data tools. One reads product details from Sheets, the other reads and writes memory in Supabase. For each question, it fetches the relevant product info, checks memory for past interactions, generates a helpful answer, and sends it back with an optional image URL if the product has one. Conversation history is stored so future messages feel continuous.