Get Started
Back to Templates

Multi-Agent Website Chatbot with OpenAI GPT and RAG Knowledge Base

Created by

Created by: Abdul Mir || abdulmir

Abdul Mir

Last update

Last update 6 days ago

Categories

Share

Company Website Chatbot Agent

Overview

This workflow implements a modular Website AI Chatbot Assistant capable of handling multiple types of customer interactions autonomously. Instead of relying on a single large agent to handle all logic and tools, this system routes user queries to specialized sub-agents—each dedicated to a specific function.

By using a manager-style orchestration layer, this approach prevents overloading a single AI model with excessive context, leading to cleaner routing, faster execution, and easier scaling as your automation needs grow.

How It Works

1. Chat Trigger

  • The flow is initiated when a chat message is received via the website widget.

2. Manager Agent (Ultimate Website AI Assistant)

  • The central LLM-based agent is responsible for parsing the message and deciding which specialized sub-agent to route it to.
  • It uses an OpenAI GPT model for natural language understanding and a lightweight memory system to preserve recent context.

3. Sub-Agent Routing

  • calendarAgent: Handles availability checks and books meetings on connected calendars.
  • RAGAgent: Searches company documentation or FAQs to provide accurate responses from your internal knowledge base.
  • ticketAgent: Forwards requests to human support by generating and sending support tickets to a designated email.

Setup Instructions

  1. Embed the Chatbot

    • Use a custom HTML widget or script to embed the chatbot interface on your website.
    • Connect the frontend to the webhook that triggers the When chat message received node.

  2. Configure Your OpenAI Key

    • Insert your API key in the OpenAI Chat Model node.
    • Adjust the model parameters for temperature, max tokens, etc., based on how formal or creative you want the bot to be.

  3. Customize Sub-Agents

    • calendarAgent: Connect to your Google or Outlook calendar.
    • RAGAgent: Link to a vector store or document database via API or native integration.
    • ticketAgent: Set the destination email and format for ticket generation (e.g. via SendGrid or SMTP).

  4. Deploy in Production

    • Host on n8n Cloud or your self-hosted instance.
    • Monitor usage through the Executions tab and refine prompts based on user behavior.

Benefits

  • Modular system with dedicated logic per function
  • Reduces token bloat by offloading complexity to sub-agents
  • Easy to scale by adding more tools (e.g. CRM, analytics)
  • Fast and responsive user experience for customers on your site
  • Cleaner code structure and easier debugging