n8n WhatsApp WAHA Bilingual AI Chatbot with GPT-4o-mini Memory

Turn WhatsApp messages into intelligent, bilingual replies—with per-contact memory and smart human handoff

This n8n workflow connects to WAHA, listens for incoming WhatsApp messages, and uses an OpenAI-powered assistant (GPT-4o-mini) to respond in the same language as the customer—Arabic or English. If a request is sensitive or unknown, it automatically escalates to a human agent with full conversation context.

What this workflow does

• Receives WhatsApp events via webhook: WAHA sends message events to the workflow’s POST webhook endpoint.
• Normalizes incoming data: The workflow standardizes fields like chat ID, sender name, and message text, plus required WAHA configuration values.
• Filters non-customer noise: It ignores group chats, the bot’s own messages, and events with empty text.
• Responds with bilingual AI + memory: It sends the customer’s text to an OpenAI assistant that replies in the same language and uses recent per-contact conversation history for context.
• Escalates when needed: If the assistant output begins with “ESCALATE”, the workflow forwards the conversation context to a designated human agent WhatsApp chat ID and sends the customer a holding message.
• Sends replies through WAHA: For normal requests, it returns the assistant’s response using WAHA’s sendText API.

Use cases

• Customer support for Arabic/English inquiries with continuity per contact.
• Sales/lead qualification where uncertain requests trigger escalation to a human.
• Multi-agent escalation for compliance-sensitive or complex questions.
• SaaS operators who want WhatsApp automation without losing conversation context.

Technical details

• n8n nodes: webhook, if, set, code, http request (plus configuration helpers like sticky notes).
• Integrations: WAHA webhook + WAHA sendText API, and OpenAI (GPT-4o-mini).
• Security: HTTP Header Auth using WAHA’s X-Api-Key.
• Configuration: Update WAHA base URL, WAHA session name, business name, and the human agent WhatsApp chat ID in workflow values.