Get Started
Back to Templates

AI Personal Assistant with GPT-4o, RAG & Voice for WhatsApp using Supabase

Created by

Created by: Amanda Benks || amanda

Amanda Benks

Last update

Last update a month ago

Categories

Share

🧠 Intelligent AI Assistant with RAG & Voice for WhatsApp – Built with GPT-4o & Supabase

📌 About this workflow and its creator

Hi! I’m Amanda, a creator of intelligent automations using n8n and Make. I’ve been building AI-powered workflows for over 2 years, always focused on usability and innovation. This one here is very special to me – a truly advanced AI assistant that reads, listens, interprets and responds like a real human 🤖✨

This ready-to-use workflow acts as a powerful AI personal assistant capable of understanding messages via voice, text, documents, or even images. It supports full multi-channel operation (WhatsApp via Evolution API, Instagram, Facebook, and more), and includes advanced RAG capabilities using Supabase + GPT-4o. It’s designed to be highly extensible, with memory, prompt update tools, and knowledge base management.

⚙️ What this workflow does

  • 💬 Understands user input via text, document, audio or image (voice, OCR, PDF)
  • 🎤 Transcribes and interprets voice messages using OpenAI Whisper
  • 🧠 Understands prompts and user commands using GPT-4o via LangChain agent
  • 🗂️ Searches knowledge base using RAG + Supabase vector DB
  • 📄 Accepts documents and automatically indexes them for future questions
  • 🧾 Summarizes documents and stores metadata in Supabase
  • 🗃️ Offers memory support (PostgreSQL chat memory per user session)
  • 📧 Sends replies through WhatsApp (Evolution API), Instagram, Facebook, etc.
  • 📅 Manages schedules (via tool integration with Google Calendar)
  • 📬 Sends and searches emails (with support tools)
  • 🛠 Modular and expandable structure (tools for saving knowledge, deleting, updating prompt)

🔧 Setup Instructions

  1. n8n Hosting
    This workflow requires n8n self-hosted (or n8n Cloud with custom credentials + community nodes enabled).

  2. Create required databases
    Use the provided SQL queries inside the setar_supabase_tabelas_vectoriais, criar_cerebro, and criar_rag_controle nodes to initialize:

    • documents table for RAG
    • cerebro table for prompt
    • memoria_chat for session memory
    • rag_controle for summaries and indexing

  3. Credentials needed

    • OpenAI API (for chat, embeddings and Whisper transcription)
    • Redis (for managing message buffer)
    • Supabase (for vector store + metadata)
    • Postgres (for memory and prompts)
    • Evolution API (or other messaging platforms)

  4. Webhook
    Set the webhook path to receive messages from your Evolution or WhatsApp API provider.

  5. Configure ‘Set’ node
    In the config node, adjust:

    • adminNumero: your personal WhatsApp or admin number
    • evolutionApiKey: your private API key
    • utilizacaoApenasViaAdmin: toggle if this should only respond to admin numbers

  6. Tool connections
    Ensure the supporting workflows are also imported and connected for:

    • Emails
    • Knowledge management
    • Calendar events

📎 Notes

  • This workflow uses LangChain agents, OpenAI GPT-4o, Supabase, Redis, and PostgreSQL.
  • It includes multiple “sticky notes” inside the workflow with explanations.
  • Ideal for businesses, consultants, and developers looking to offer an intelligent and extendable AI chatbot experience.

🛍 Want to use this on your system?

❤️ Buy workflows: https://iloveflows.com
☁️ Use n8n Cloud with my partner link: https://n8n.partnerlinks.io/amanda