WhatsApp & Telegram Chat Routing with GPT-4o + Google Sheets

Route WhatsApp & Telegram chats to a GPT-4o multi-agent assistant—with Google Sheets updates

This n8n workflow captures inbound WhatsApp and Telegram messages, routes them through an OpenAI GPT-4o multi-agent assistant, and uses PostgreSQL plus Google Sheets to log, validate, and (optionally) sync product updates—while alerting admins and responding back to the user.

What this workflow does

• Triggers on inbound messages from WhatsApp or Telegram (or supports a manual sample run) and normalizes them into a common sender/message format.
• Loads your configured business profile, determines whether the sender is a customer, supplier, or admin, and de-duplicates events by claiming the message in PostgreSQL.
• If the message is an authorized Telegram admin command, it applies approve/reject decisions in PostgreSQL and confirms the result back to the admin in Telegram.
• Otherwise, it retrieves approved knowledge and recent conversation history from PostgreSQL, then sends packaged context to a main GPT-4o agent that delegates to up to six specialist agents that can read relevant Google Sheets tabs.
• Applies a strict policy layer to validate and constrain the proposed intent/actions, then writes the approved plan and any resulting records to PostgreSQL.
• If a product row changed, it synchronizes the updated product data to the Products tab in Google Sheets.
• Sends a Telegram admin alert when actions or manual-review notes exist, and—when enabled—replies to the sender on the original channel (Telegram or WhatsApp).

Use cases

• Support teams triaging customer questions from WhatsApp/Telegram with context-aware responses.
• Ops workflows where product decisions require admin approve/reject before changes take effect.
• SaaS operators who want conversation history + action logs stored in PostgreSQL and product updates reflected in Google Sheets.

Technical details

• n8n nodes/tools: if, code, postgres, telegram, whats app, sticky note
• Core systems: PostgreSQL for deduping, approval state, and conversation history; Google Sheets for approved knowledge and product sync (Products tab)
• AI layer: GPT-4o via a main agent with up to six specialist agents, governed by a strict policy layer