The magic begins ! You have peaked my interest in self hosting N8N (i may need to upgrade my MacBook) and build a few automations…obsidian being my destination to accumulate all the content I discover on YT, podcasts, news articles and various and sundry RSS feeds. Thank you so much ! You are presenting and teaching at the exact level that I’m at right now.
I mainly use NodeRed as my low code and AI Workflow platform. Yes i know it is more rudimentary than n8n but back in the time when I used n8n for my various email related workflows I wasn’t happy with the performance as it often stalled doing things. My nodered instance sits in HomeAssistant (so it can also easily access my smarthome) and my ollama installation on a discrete machine in my 19” rack. I’m doing all kind of stuff like summarizing and filter my emails, telling me (via homeassistant text-to-speech) that my various devices have finished working, that my car finished charging and how long it needed/electricity it consumed - all with different and funny texts. Most of my ollama knowledge I got from this channel 😉
Ive got two instances of N8N running. One handles an email account, responds to emails and there are some things I can ask it to do in email. It also summarizes news articles, etc. The other one is for the AI. It handles specific requests. For example, I wrote a python script that gathers system info on my ollama server. Things like cup and memory usage, gpu usage for all the gpus, etc. So when I ask the AI how it is doing, or how it is feeling, it pulls system information and feeds it to the model to write the response. If I ask it to run a full internal diagnostic it uses that info plus some other info it gathers to tell me about the status of the system it is running on and the subprocesses that handle other parts of the AI.
Video very much appreciated. Taking a side detour to look at standing up a Windmill instance in the proxmox cluster to immediately start reducing some workloads, +1 for the extra tip!
Thank very much Matt, I've been using n8n for the past 2 months and I have created many cool workflows! The one I'm most proud of sends random messages to my parents on WhatsApp, sharing updates about daily events!
node-red is way more versatile than n8n imo, I wouldn't dismiss it so quickly. I created a chat & image generation interface for ollama/comfy ui running in node-red that I can access from anywhere, which in turn can use/fire off anything I want. Based on the trigger words i use in the interface, it is routed to the AI i want. There are 1000's of nodes/plugins available for any type of integration. I do agree that oauth2 integration out of the box is very nice, as it is a horrible mechanism to deal with, but also nodes available in node-red, but a bit more finicky. Creating your own api in node-red takes like 5 seconds, which means that you can use it in other flows or tools.
This is what made me drop everything and focus on AI. When I saw Matt Wolf embed a call to an LLM from within a workflow, my world changed. For me, this showed a way of blending deterministic tasks with inference. It showed a glimpse at how AI could be utilized to make traditional approaches work better without falling victim to what AI was not good at - deterministic tasks. I hope n8n can continue to find a way to blend the corporate and community model. NOW: Is it useful to use Huginn to trigger or act as agents for your workflow, e.g. trigger your webhook?
I’m very interested in learning more about n8n. A couple moths back I had it running locally on my pc but when I reloaded windows I haven’t reinstalled it mostly because I was still in the learning stages and well some new image generation models (Flux) came out and I was off on that adventure lol. They honeymoon time is over and now I’ve back to working with llm’s and actually just installed Preplexica using your video, thank you. I appreciate all your efforts. Jason aka SouthbayJay
Very interesting video, thank you! I'm currently playing around with Flowise, but this looks even more powerful and intuitive. Would love to see a video about creating custom nodes in n8n like you suggested!
Excellent content! I am very interested in any follow-up suggestions you may have, as I believe many viewers would benefit from seeing additional approaches to these processes-even if we are already implementing similar methods.
That was super interesting. I would love to hear more. I am currently using IFTTT. I was looking to make the move to Make but N8n looks like a better fit.
I was commenting on comments and deleted one accidentally. Yes I will do a video about why I installed via npm vs docker on my local machine. I should have learned my lesson that it’s never good to work with RU-vid comments from the hot tub.
This is close to what I am looking for. I will admit to being a long term coder, but reasonably new to AI. What I am looking for is a way to have one LLM running on one device call a LLM running on a different device, then get the results passed back. My thought is, have fine tuned models for solving specific types of problems, with one model to rule them all. Creating a Langchain tool that SSHes into the other computer sounds like a solution, but not a fun one. Any ideas? Again, this is very early in the project. I have a bit of sample test running, but nothing close to solving the problems.
Great video! Langchain 😅 Lot’s of juggling involved, feels like calling APIs in your dreams, very creative and non deterministic. Any help taming this beast would be appreciated.
Why not NodeRed? It's opensource and much more powerful than N8N and you can use custom plugins from npm. I use NodeRed with Ollama and HomeAssistant. In fact NodeRed talks to all the MQTT and REST stuff. I selfhost it all though.
I didn’t mention it because they didn’t. This is a tool I have used for years that I really like. You don’t like me sharing tools I like to use? That’s kinda the point of the channel.
@@technovangelist no, don’t take it so negative, please. I didn’t mean to imply that. We are constantly bombarded by sponsored video-reviews, it just felt like another one. My humble advice, you could explicitly mention they did not sponsor it. Great you shared the tool. Knowing it is not sponsored, I will watch it again from a new perspective. I love your content, specially the course! Thanks! 😊
@@technovangelist Wow. At the hype of no-code tools when n8n was just the decent alternative to zapier and nodered, I’ve heard an influencer mentioning n8n as Neithan, so the pronunciation stuck with me. I saw it written n8n in multiple articles since, but I was always reading it as Neithan. Today I’ve checked, and of course you’re right.