WhatsApp personal AI assistant with voice, image, and PDF support

Description

📺 Full walkthrough video: https://youtu.be/-kpt0BwjKls

Who it's for

This workflow is for professionals and power users who want a personal AI assistant accessible directly via WhatsApp, capable of managing their emails, calendar, files, and online searches through natural conversation — including voice messages.

How it works

A WhatsApp Trigger receives incoming messages and routes them by input type: text, audio, image, or PDF document.
Audio messages are downloaded and transcribed via OpenAI Whisper. Images are downloaded and analyzed by GPT-4o mini. PDF files are validated, downloaded, and text-extracted.
All input types are normalized into a unified text field before being passed to the AI agent.
A personal assistant agent powered by Claude Sonnet processes the request using a full tool suite: Gmail (send/search), Google Calendar (create/read events), Google Drive (search), Airtable (contact email database), SerpAPI (web search), Discord (direct messages), and a calculator.
If the original input was a voice message, the agent's response is converted to audio via OpenAI TTS and sent back as a WhatsApp audio message. Otherwise, a text reply is sent.

How to set up

[ ] Connect your WhatsApp Business API credentials to the trigger and all send nodes
[ ] Add OpenAI API credentials to the transcription, image analysis, and TTS nodes
[ ] Add Anthropic API credentials to the Claude Sonnet model node
[ ] Connect Gmail OAuth2 credentials to the send and search email tool nodes
[ ] Connect Google Calendar OAuth2 credentials to the create and get events tool nodes
[ ] Connect Google Drive OAuth2 credentials to the search tool node
[ ] Connect Airtable credentials and configure the base/table IDs for the email database
[ ] Add SerpAPI credentials to the web search tool node
[ ] Add Discord bot credentials and configure the target user ID for direct messages
[ ] Set the correct phone number ID and recipient number in all WhatsApp send nodes

Requirements

WhatsApp Business API account
OpenAI API account
Anthropic API account
Google account (Gmail, Calendar, Drive) with OAuth2
Airtable account with a contacts/email base
SerpAPI account
Discord bot with direct message permissions

How to customize

Add or remove tools from the Personal Assistant Agent node to expand or restrict capabilities (e.g. add a Notion or Slack tool).
Adjust the memory window size in the Conversation Memory Buffer node to control how much conversation history the agent retains.
Edit the agent's system prompt to change its persona, language, or restrict it to a specific scope (e.g. a customer support assistant).