Imagine having a personal AI assistant that manages your daily life, all through WhatsApp. Sounds like science fiction, right? But I’ve made it a reality, and I’m excited to share my project with you.
I was tired of switching between multiple apps to stay on top of my day. So, I created a personal AI assistant that I interact with entirely through WhatsApp, using n8n as the backbone.
What Can It Do?
My AI assistant can do some amazing things, like:
* Manage my bills: I forward it a message with my credit card due dates, and it parses the text, totals the bill amounts, and sets reminders in my calendar.
* Keep my schedule: I can ask it to remind me to hit the gym, and it adds it to my Google Calendar and sends me a notification.
* Summarize my inbox: Instead of scrolling through emails, I ask it to check if I have any important mail, and it gives me a clean, bulleted list of important subjects and senders.
* Understand images: I snapped a photo of a delivery address, and it extracted all the text, identified the pincode, state, and other details.
* Act as a music DJ: It can suggest playlists for any mood or task. When I asked for Ilaiyaraaja songs for work, it gave me a curated list and created a YouTube playlist for me on command.
The Tech Setup
The real magic happens in n8n, where I built a workflow that orchestrates everything. Here’s how it works:
* WhatsApp trigger node kicks everything off.
* A primary AI node receives the message and figures out what I want to do.
* Based on the intent, it passes the task to a specific sub-workflow, which includes:
* Calendar/task agents that connect directly to Google Calendar and Tasks APIs.
* Research agent that queries Wikipedia and my self-hosted Perplexity instance for reliable and up-to-the-minute data.
* Image analysis that calls an external API to perform OCR and feeds the extracted text back to the main AI for interpretation.
It’s been an incredibly powerful way to create a single, conversational interface for my digital life. And the best part? I can host the core logic myself with n8n and even the research LLM.
What Do You Think?
I’d love to hear your thoughts on this project. Are there any cool features I should consider adding to the workflow? Happy to answer any questions about the setup.
*Further reading: n8n.io, Perplexica*