Thoughts on n8n tool
Yesterday I visited on Hamburg n8n meetup. I met a lot of nice people with different ideas and experience. My goal was to understand two things: how people use n8n to solve real problems and how can I use it in my projects. I got answers to all my questions on this meetup.
What is n8n?
n8n - AI automatisation platform which allow users to connect different apps, APIs and AI models in pipeline. It has relatively simple visual interface where user can drag-n-drop services and build complex chains. The platform provides a list of tools for integration, templates and boilerplates.
Some people call n8n as Lego for AI automation.
Alternatives: Zapier, Make.
Founder: Jan Oberhauser, 2019, Berlin, Germany.
Website: https://n8n.io/
How people use n8n?
On the internet you can find a lot of cases how people automate their daily routine using 8n8 and AI tools. Yesterday there was a presentation how to use n8n and local models to create a second brain.
A "second brain" is an external, digital database where you save and organize your ideas, notes, and learnings. Popularised by Tiago Forte, it acts as a trusted memory storage system so your biological mind is free to focus on thinking and creating instead of memorising data.
For me problem of building the "second brain" is not new: for the last several years I tried different approaches to solve this problem. The latest working tool for me was Obsidian vaults with different levels of automation.
For using local models I also built a plug-in to connect LM Studio to Obsidian and use models for summarisation, grammar, structure tasks, searching and so on. So I already solve this problem for myself and I wanted to see how other people approach similar task.
Building a second brain using n8n is a complex task which includes zoo of tools. I saw a "spaghetti" of pipelines to structure the request and generate the respond. It was complex overloaded system which took days for setup. And because of complexity I don't know how to support and scale it in the future.
The person who presented the n8n-backed "second brain" was absolutely proud of the final result: got the answer in seconds: Telecom costs or brand of the watch with a black screen. It is working system which takes a lot of time and effort to build it. The solution works locally on his mac and Hetzner server using local models, so it costs relatively zero for him.
But the only thing is - it is not efficient and scalable. If you are an old-school techie, real DIYer, like to build a telegram bot on the Raspberry Pi with orchestration tools - it is your way. The only problem that average internet user who really needs some simple AI automation cannot do it by her own.
How could I use n8n in my projects?
On the meetup I have a chance to discuss the ways of using n8n tool in my projects. My questions was "Can I build a content factory using n8n?". Three bright minds and I sat on the table and brainstormed the ways to inject content factory into shortifi.me project.
There were a lot of good ideas how to do it. But all of them I can do faster, easily and cheaper without n8n. Adding n8n into the project additionally connects a lot of supportive tools and increase complexity of the project. Which I don't want to make.
The future of n8n
n8n was a great innovation in 2019. It attracted a lot of innovators at that period. After that came early adopters who loves to automate and do things. But today the time of the mainstream market for n8n and the company should cross the chasm.
I see the one biggest issue is complexity. The average user won't spend a lot of time to setup the pipeline or a tool. It should be simple as send an email.
If the n8n will find the solution for reducing complexity - it will be on each laptop. If not - in the next several months we will see another tool addresses the complexity problem carefully.