The n8n AI Agent video is extremely helpful for progressing to the next steps. Having installed n8n through self-hosting, I'm now determined to start exploring n8n step by step.
@@dawitgebeta9263 Yes I am planning on making a video on how to deploy this in the near future! It's kind of a separate topic so I didn't want to include it here and make the video too long, but it's definitely important so I'll be covering it! And doing it justice by having a dedicated video just for deployment.
@paidoclock You sure can! This can be deployed really to any cloud machine no matter the provider. When I make a video on deploying this (or other applications), I'll be sure to call out how it can work for any cloud machine. I will be using Linux, so my instructions will apply as long as you are using Linux (and that's available in Oracle).
Hi good job! I use o1 for coding I noticed it survives simple prompts and it can code a bit more elegant (read serviceable) code. It can refactor code into classes etc with simple refactor prompts. Only problem I have is GPS problem. Can I still read maps? Do I even remember how to code? :)
How about a video that shows how to create a simple custom Chat html embed and UI for copy and pasting as an on a clients web page? 🙏 The default n8n chat ux design looks terrible and has a start chat activate button and powered by logo that’s a problem
I agree and thank you for the suggestion! You can actually embed this Streamlit app as an pretty easily so it would work for that too! But I get wanting something even more simple so I could create a video using a tool like Bolt.new to create an HTML embed with much better styling!
Hey man I love your videos. I’m having a hard time connecting n8n this way though. Do I have to run it on a local environment or can I get it working through the bolt.new development model?
Nice, can you build full stack SaaS AI project using n8n, Nextjs, Supabase and Lemon Squeezy as subscription payment gateway? It's gonna be huge. Some projects like PhotoRoom Clone, ChatPRD Clone, with credit system for multi pricing plans will be awesome
Thanks for the suggestions! I am certainly going to be creating content in the future taking these kind of apps further to make them full stack SaaS projects!
i want a version of this with a verification ID for users who want to use the agent on whatsapp for example - maybe connect it to twilio whatsapp number and then when a user talks to it it saves that phone number ID session perhaps asks for their name and email - and then this specific person has their own memeories for them specifically all based on their whatsapp phone number... though idk how to pull that off with n8n technically
Great question! You want to use your anon key in the Streamlit frontend and then your secret key in the n8n backend. If you mean the HTTP request node in the n8n backend, then use the secret key!
Great one. I've been taking a look at your videos and I've seen that you installed Postgres local with n8n at the same time. I've installed Postgress later one as I do it one by one... Now I'm trying to figure out how to create a table, connect it to n8n etc... which is the complicated part. This is the only part that I find in this video out of it. Anyways, I think your video is great.
Thank you very much! Let me know more specifically which part of getting a table set up and connected to n8n is causing you trouble and I've love to give some pointers!
@@ColeMedin I did create a docker postgres, then had to connected to the pgadmin postgres, then I did connect the dockers locally with addresses like 172.0.0.X and that made it. Now as I don't know which table to create I've seen in the n8n documentation that n8n creates the tables if they don't exist. So I'm waiting the workflow to work. Although I'm in the process of webhook, Google Drive, etc... Let's see how I can progress and I'll tell you. Thnks. It looks like in your Github you have two types of code though. And I've downloaded the more completed one. It's weird. Even if I create the file in the folder and I click on Test the Workflow. The Google Trigger never fires. UPDATE. I think it was the type of the file. PDF isnt' allowed? Working on txt...
Yeah PDF isn't supported by default (you have to add an extract text node specifically for PDF which n8n provides) but it still should trigger the workflow! You do have to make sure the workflow is toggled to active in the top right, maybe that isn't checked in your case?
Thank you very much! Could you clarify what you mean by having the n8n AI chatbot ask for parameters? Are you asking if the webhook can accept parameters not given by the user and use those to make API requests? If so, you can certainly do that! You can add extra parameters to the webhook, and then pass those into the "System Message" for the AI Agent (you can add the System Message by clicking on "Add Option" in the AI Agent node) so it knows what these values are and can use them to invoke tools that make external API requests. I hope that makes sense and please feel free to clarify if I didn't answer your question!
You can indeed create custom nodes for n8n! You have to make them with JavaScript/TypeScript though since n8n is an NPM package. docs.n8n.io/integrations/creating-nodes/overview/
I know that Cole's videos are very general, but that's all most people need to start their project. I did something similar for my project, except I'm not using Streamlit, and it's very easy to do. Now, I'm running it locally to help with my day-to-day tasks at work. Not to mention, I only started learning N8N two weeks ago. With all these AI tools available to us, there's no reason you can't figure things out.
@@Slyndc I agree with what @SDChillRide said, but also I am planning on making more content in the future that goes way in depth taking something to production like you are saying! I don't want to do that for most videos though because most of the time people appreciate something concise to get them started. But there is a time for longer content too for sure!
Has anyone actually got Claude to use relevant libraries for other work? I find too many uses of deprecated libraries being used. I want a system which incorporates the latest docs into the LLM and the only tool I know that is cursor but unfortunately it's doc crawling system is pretty shit. Surely there must be a tool outs there that automatically crawls and uses relevant docs
Typically I can get Claude to use up to date libraries just by prompting it to do so if it hallucinated and used deprecated ones. But that of course depends a lot on the libraries you are using. Which ones have you seen Claude used that are out of date? I'm not aware of the kind of tool you are looking for, but that would be insanely cool and useful! One thing I'll say is some documentation pages, like the one for LangChain, have a chatbot that you can use to ask how to code certain things using the library. Of course its knowledge is more just limited to that library, but it's sill partly what you want!
@@ColeMedinI am playing around with web scrapers to extract documentation and feed it into the 2million token limit of Gemini and hope to god it can help. Let's see what happens
Hi Cole! You wanna build something that many, many people will be interested in? How about a simple server setup (standard PC with an RTX4090 or RTX3090 ) in a local environment offering inference to LAN members, with a super useful RAG engine? Small business, having emails, PDFs, CSV and Excel files, Meeting protocols etc. They don´t have the funds to get a 3rd party service come in and build a RAG pipeline for them. But they might have an IT Manager who can build a simple setup for the team to be used for querying, wiki etc.! Maybe even optimize the RAG engine with multi stepping so that the most relevant data is retrieved from the vector database? A possible stack could be privateGPT with qdrant and ollama - langchain for optimizing the retrieval workflow.. Open source models are fast enough to do the reasoning on the data even today, so all can be private and local. You are a Pro so I guess you could put something like this together, right?
Yeah I could put something like this together, I appreciate the suggestion a lot! What are your thoughts on doing it this way versus having businesses pay for a cloud machine to host everything?
Indeed it is! Supabase has an awesome free tier and you can even self host it since it is open source. You can host n8n and Streamlit yourself as well so the only thing you really have to pay for is the machine in the cloud to host it!
Great question! A Streamlit app wouldn't be suited for a WhatsApp or Instagram deployment. You would want to use the WhatsApp or Instagram API to interact with the n8n agent through the webhook, similar to how I do it with the Streamlit UI.
Fair question! I just love the Claude website UI, but the convenience of Claude dev and having that right in the IDE is definitely a plus. So maybe I should use that in the future!
The n8n AI Agent video is extremely helpful for progressing to the next steps. Having installed n8n through self-hosting, I'm now determined to start exploring n8n step by step.
I'm glad you found it helpful - thank you!
This video is a gold mine. Ill be looking more n8n videos. Thanks
Awesome, I appreciate it a lot!!
Hi Cole, thank you for following up on this. Great to see n8n connected to a webhook. Keep up the good videos.
My pleasure, thank you man!
Can you show us how you can deploy these to production? I find my challenge to be when I try deploying to the web/production instead of just localhost
@@dawitgebeta9263 Yes I am planning on making a video on how to deploy this in the near future! It's kind of a separate topic so I didn't want to include it here and make the video too long, but it's definitely important so I'll be covering it! And doing it justice by having a dedicated video just for deployment.
On this thread, can you deploy it on oracle cloud free tier? Thank you
@paidoclock You sure can! This can be deployed really to any cloud machine no matter the provider. When I make a video on deploying this (or other applications), I'll be sure to call out how it can work for any cloud machine. I will be using Linux, so my instructions will apply as long as you are using Linux (and that's available in Oracle).
vercel is quickest and easiest but maybe not best - but i like it.
@littledaddi3 I do like Vercel as well!
Hi good job! I use o1 for coding I noticed it survives simple prompts and it can code a bit more elegant (read serviceable) code. It can refactor code into classes etc with simple refactor prompts. Only problem I have is GPS problem. Can I still read maps? Do I even remember how to code? :)
Thank you! And yes - o1 is fantastic for coding because of its behind the scenes reasoning!
How about a video that shows how to create a simple custom Chat html embed and UI for copy and pasting as an on a clients web page? 🙏 The default n8n chat ux design looks terrible and has a start chat activate button and powered by logo that’s a problem
I agree and thank you for the suggestion! You can actually embed this Streamlit app as an pretty easily so it would work for that too! But I get wanting something even more simple so I could create a video using a tool like Bolt.new to create an HTML embed with much better styling!
I am really looking forward to seeing how OpenWebUI and n8n can be integrated seamlessly. When will your explanation be available?
I'm glad!! I'll be putting a video out with OpenWebUI + n8n within the next month!
Spectacular! Too beautiful, thank you Cole!!!
Thank you very much Anna!!
Hey man I love your videos. I’m having a hard time connecting n8n this way though. Do I have to run it on a local environment or can I get it working through the bolt.new development model?
Thanks Brian! Either should work! What is the error you are running into right now?
Nice, can you build full stack SaaS AI project using n8n, Nextjs, Supabase and Lemon Squeezy as subscription payment gateway? It's gonna be huge. Some projects like PhotoRoom Clone, ChatPRD Clone, with credit system for multi pricing plans will be awesome
Thanks for the suggestions! I am certainly going to be creating content in the future taking these kind of apps further to make them full stack SaaS projects!
@@ColeMedin Yeah, really excited to watch these content from you!
Can you make the stream-on output available?
Could you clarify what you mean by stream-on output?
Hello Cole, thank for the video. I would like to see a video of Open Web UI with n8n.
My pleasure and thanks for the suggestion! I will be doing exactly this in the near future!
i want a version of this with a verification ID for users who want to use the agent on whatsapp for example - maybe connect it to twilio whatsapp number and then when a user talks to it it saves that phone number ID session perhaps asks for their name and email - and then this specific person has their own memeories for them specifically all based on their whatsapp phone number... though idk how to pull that off with n8n technically
i basically wanna do a RAG AGENT on whatsapp with automatic identification, with conversational memory if thats possible.
That is certainly doable - I love your use case! I might do something like this in the future!
In the HTTP Request node, do I enter my SupaBase anon key or secret key?
Great question! You want to use your anon key in the Streamlit frontend and then your secret key in the n8n backend. If you mean the HTTP request node in the n8n backend, then use the secret key!
@@ColeMedin ok thanks!
You bet!
Great one. I've been taking a look at your videos and I've seen that you installed Postgres local with n8n at the same time. I've installed Postgress later one as I do it one by one... Now I'm trying to figure out how to create a table, connect it to n8n etc... which is the complicated part. This is the only part that I find in this video out of it. Anyways, I think your video is great.
Thank you very much!
Let me know more specifically which part of getting a table set up and connected to n8n is causing you trouble and I've love to give some pointers!
@@ColeMedin I did create a docker postgres, then had to connected to the pgadmin postgres, then I did connect the dockers locally with addresses like 172.0.0.X and that made it. Now as I don't know which table to create I've seen in the n8n documentation that n8n creates the tables if they don't exist. So I'm waiting the workflow to work. Although I'm in the process of webhook, Google Drive, etc... Let's see how I can progress and I'll tell you. Thnks. It looks like in your Github you have two types of code though. And I've downloaded the more completed one. It's weird. Even if I create the file in the folder and I click on Test the Workflow. The Google Trigger never fires. UPDATE. I think it was the type of the file. PDF isnt' allowed? Working on txt...
Yeah PDF isn't supported by default (you have to add an extract text node specifically for PDF which n8n provides) but it still should trigger the workflow! You do have to make sure the workflow is toggled to active in the top right, maybe that isn't checked in your case?
Great content! Is it possible to make n8n ia chatbot ask for parameters not given by user to make and external api request?
Thank you very much! Could you clarify what you mean by having the n8n AI chatbot ask for parameters?
Are you asking if the webhook can accept parameters not given by the user and use those to make API requests? If so, you can certainly do that! You can add extra parameters to the webhook, and then pass those into the "System Message" for the AI Agent (you can add the System Message by clicking on "Add Option" in the AI Agent node) so it knows what these values are and can use them to invoke tools that make external API requests.
I hope that makes sense and please feel free to clarify if I didn't answer your question!
Very awesome. Very nice job and good content.
Thank you Anthony, that means a lot!!
is it possible to create own functions written in python to import like a module in n8n?
You can indeed create custom nodes for n8n! You have to make them with JavaScript/TypeScript though since n8n is an NPM package.
docs.n8n.io/integrations/creating-nodes/overview/
Great tutorial!
Thank you!!
Another banger!!
Thank you Cam! 😎
We need to see this in production and a lifer video step by step, no lazy edits.
I know that Cole's videos are very general, but that's all most people need to start their project. I did something similar for my project, except I'm not using Streamlit, and it's very easy to do. Now, I'm running it locally to help with my day-to-day tasks at work. Not to mention, I only started learning N8N two weeks ago. With all these AI tools available to us, there's no reason you can't figure things out.
@@Slyndc I agree with what @SDChillRide said, but also I am planning on making more content in the future that goes way in depth taking something to production like you are saying! I don't want to do that for most videos though because most of the time people appreciate something concise to get them started. But there is a time for longer content too for sure!
where can I host this publically? somewhere else than localhost?
Sure can! I just made a video on this!
ua-cam.com/video/259KgP3GbdE/v-deo.html
Has anyone actually got Claude to use relevant libraries for other work? I find too many uses of deprecated libraries being used. I want a system which incorporates the latest docs into the LLM and the only tool I know that is cursor but unfortunately it's doc crawling system is pretty shit. Surely there must be a tool outs there that automatically crawls and uses relevant docs
Typically I can get Claude to use up to date libraries just by prompting it to do so if it hallucinated and used deprecated ones. But that of course depends a lot on the libraries you are using. Which ones have you seen Claude used that are out of date?
I'm not aware of the kind of tool you are looking for, but that would be insanely cool and useful! One thing I'll say is some documentation pages, like the one for LangChain, have a chatbot that you can use to ask how to code certain things using the library. Of course its knowledge is more just limited to that library, but it's sill partly what you want!
@@ColeMedinI am playing around with web scrapers to extract documentation and feed it into the 2million token limit of Gemini and hope to god it can help.
Let's see what happens
Wow that's pretty cool, I hope it works well for you!
Thank you. Nice work....
Thank you very much, my pleasure :)
Did you try anythingLLM?
Not yet!
Hi Cole! You wanna build something that many, many people will be interested in?
How about a simple server setup (standard PC with an RTX4090 or RTX3090 ) in a local environment offering inference to LAN members, with a super useful RAG engine?
Small business, having emails, PDFs, CSV and Excel files, Meeting protocols etc. They don´t have the funds to get a 3rd party service come in and build a RAG pipeline for them. But they might have an IT Manager who can build a simple setup for the team to be used for querying, wiki etc.!
Maybe even optimize the RAG engine with multi stepping so that the most relevant data is retrieved from the vector database?
A possible stack could be privateGPT with qdrant and ollama - langchain for optimizing the retrieval workflow..
Open source models are fast enough to do the reasoning on the data even today, so all can be private and local.
You are a Pro so I guess you could put something like this together, right?
Yeah I could put something like this together, I appreciate the suggestion a lot! What are your thoughts on doing it this way versus having businesses pay for a cloud machine to host everything?
Is that all free resources??
Indeed it is! Supabase has an awesome free tier and you can even self host it since it is open source. You can host n8n and Streamlit yourself as well so the only thing you really have to pay for is the machine in the cloud to host it!
thanks for sharing, this helps
Of course Alex, I'm glad to hear! 😀
Can I create this chat for other languages?
You sure can! LLMs are great at working with really any language. They do perform the best with English, but they still do well with others!
Great video, as usual 😅
Btw, isn't 'n8n' pronounced as 'natan'?
Thank you! I haven't heard it pronounced that way but maybe it is 😂
Just found you today, been watching a few AI channels, will be checking out more of your content, 👍 good job
I appreciate it a lot!! Thank you!
How can I deploy it on WhatsApp, Instagram?
Great question! A Streamlit app wouldn't be suited for a WhatsApp or Instagram deployment. You would want to use the WhatsApp or Instagram API to interact with the n8n agent through the webhook, similar to how I do it with the Streamlit UI.
dang cool
Thank you!
Why not Claude dev
Fair question! I just love the Claude website UI, but the convenience of Claude dev and having that right in the IDE is definitely a plus. So maybe I should use that in the future!
Respect to the AI Arabic translation
Haha thank you, turns out someone had just commented "The French" on one of my videos. Not sure what that meant...
Pretty awesome! Got insights from you !
Thank you! I'm glad you did!
I was looking for this; Thanks, you're awesome!
@@PrashadDey I'm glad to hear, my pleasure and thank you!
People are using Lang chain to build AI agents ;)
I'm using a lot of LangChain myself to build agents! That plus LangGraph.
Combo
We love a good combo 😂