WhatsApp AI Assistant — Google Docs-Powered Customer Support

Overview

This project is a customer-support chatbot for WhatsApp Business that answers questions using a company's own documentation. The knowledge base is simply a Google Doc: there is no model training or re-indexing step, so editing the document updates what the bot knows immediately.

How it works

The n8n workflow starts with a WhatsApp Business Cloud API trigger. For each incoming message it fetches the current content of the knowledge-base Google Doc, builds a prompt combining the current date, the document text and the user's message, and passes it to an AI agent backed by Google Gemini with per-user buffer memory for multi-turn context.

Every exchange is logged to Google Sheets for later review. A code node then cleans the model output — stripping markdown and AI preambles — so replies read naturally in WhatsApp. Finally, the workflow checks whether the user is still inside WhatsApp's 24-hour customer-service window: if so it sends the answer directly, otherwise it sends an approved template message to reopen the conversation.

The design keeps setup simple for non-technical teams while respecting WhatsApp platform rules, and the README documents swapping Gemini for OpenAI or replacing simple memory with PostgreSQL-backed memory for persistence.