Creating AI-Agents was never this easy!

Hey there, friends! It’s been way too long since my last blog post. After getting Nico-chan’s AI project to work (cue confetti), life threw me a curveball—or a series of them—and I ended up letting some projects lapse. Let me be honest, I felt the sads hard.

But hey, here we are! AI is now everywhere, and AI agents aren’t just for tech wizards anymore. Enter N8N—a game-changer that’s made automation feel like child’s play. And the cherry on top? It’s free to run on your own hardware. Score!

Meet N8N: The Swiss Army Knife of Automation

I stumbled upon N8N after watching a few hype videos, and wow, it exceeded my expectations. This thing is flexible. You can automate pretty much anything, including AI tasks, all for free. And since I’ve got my homelab server running, installing this was as easy as Docker one—total breeze.

Quick shoutout to N8N’s docs: they’re top-notch. No need to stress over complicated setups. Just head over to their GitHub and documentation pages, fire up Docker, and you’re good to go. I’m talking that easy.

Restarting Nico-chan with N8N

Now, why am I telling you all this? Well, I decided it was time to breathe life back into my old Nico-chan AI project. But instead of starting from scratch, I jumped into N8N’s world. It felt like cheating how simple it was.

I found a sweet template for AI-powered social media automation on N8N templates page. The template used OpenAI and Perplexity, which are great, but let’s be real—my budget’s not that flexible. Plus, we’ve already got our local Ollama AI server running. So, I decided to make it my own.

Tweaking the Workflow: My Hacky Way

First up? Simplifying the workflow for Twitter only. Easy enough. Then, I swapped out OpenAI and Perplexity with our local heroes—Ollama and OpenWebUI. Talk about cost-effective!

I also added some custom tweaks:

Category Randomizer: Because who wants all their tweets to be about Gemini AI? Not me! Now it picks random topics from a list I set.
Text Cleaning Code: Just to make sure the content isn’t all over the place.

0:00

/1:05

Pro tip: Version 1 is always rough, but it works—for now. Expect updates as I figure out what’s working (and what’s not).

Results? Eh, Not Bad for a First Try

Three days in and my little AI agent has been posting away on Twitter without skipping a beat. Fingers crossed this experiment sticks around long-term. If it does, I’ll keep you all updated.

Next Up: More N8N Fun!

For now, I’m plotting my next move—maybe something with Telegram or another fun project. Stay tuned for more N8N shenanigans in the coming weeks!

Quick Win: How I Set Up N8N on My Homelab Server

If you’re curious about how to get started, here’s a quick run-through:

Install Docker: Easy peasy via your package manager or script.
Fire it Up: Follow their docs, and you’ll be automating in no time.

Or for you, PH guys that doesn't have the time to read, here is my Docker command for setting up N8N:

docker run -it --rm --name n8n -p 5678:5678 -e N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true -e N8N_RUNNERS_ENABLED=true -e GENERIC_TIMEZONE="Asia/Singapore" -e TZ="Asia/Singapore" -e N8N_SECURE_COOKIE=false -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n

And hey, if you want to see my tweaked workflow for inspiration, here is a GitHub link to my workflow.

I am using the Qwen3 LLM model for the topic selection, the DeepSeek R1 LLM model, and the OpenWebUI for the Twitter post creation. Feel free to tweak them and the AI completion's user and system message because these are what work for me.

Thanks for Hanging With Me!

If you found this useful (or just amusing), share it with your crew—let’s spread the word about N8N’s magic. And as always, drop me a comment or DM if you’ve got questions or tips of your own.

Here’s to more AI adventures and less stress!