Here we go! Someone that actually knows what he is talking about ! Thanks a lot for the content, I have quite the same stack and it’s efficient. I thank you for the inspiring and helpful information, I will rework it a lot thanks to you 😊.
FINALLY , after watching months of youtube video's listening to all the hype and fluff , Ai Agency's hype etc etc. It great to see someone who is most interested in building real life workable solutions without all the fluff and Hype. Youv'e won me, keep up the great content
Good set up. N8N is the right tool if agencies are serious about these things. If you are targeting SMB and perhaps the lower end of SMB’s, you may want to invest in creating for yourself and outreach agent schema that discerns between those companies where the $3000 automation tool is a no-brainer versus going near companies with very thin margins who aren’t likely to be customers because of it
A Custom AI Agent Framework video is super insightful! The innovation in creating and managing AI agents for various tasks is remarkable. #AI #Innovation #CustomAgents
With *Phlanx's caption generator* , you can effortlessly craft captions that resonate with your audience. It’s a valuable tool for any business looking to strengthen its social media presence.
Thanks a lot for this great video content! I have a question : when you talked about the apps you developped in the part for your customers using AI Agents, did you used an agentic framework like CrewAI, LangChain Studio or n8n to develop these apps ? or hard coding approach like using Python, Visual Code and LangGraph framework ? Thanks!
Thats so awesome. I would love to develop something for the distillery. I'm working on something for about 6 month to learn, test, discover. What a journey ! Would love to discuss with you about it more in details.
Where do you store the business knowledge (ie. Business background, contact knowledge, product knowledge) into your AI n8n node? Do you upload it into the vector store?
I think you are one of the best out there for this kind of stuff. I'm learning so quick I'll have a bunch of agents doing my things before the end of the year. I just have a question about tools, why do you need openai chatgpt plus on top of platform.openai assistants? Can't you just do it all with the assistants connetected to other workflows outside in n8n or other tools? I didn't find a way to connect a custom gpt as a an action instead of a trigger..
Yes you could! However, I like to use the Langchain nodes in n8n for the tool calling rather than the assistants model. I'm not sure it matters which way you do it though.
@@customaistudio Thanks for the answer! Ill check it out. Zapier don't connect to chatgpt, only to assistants, so I thought it was not possible. I'll probably create an account with n8n soon, do you have an affiliate link or something?
Very informative presentation! however I am missing one big point which was not mentioned, PRIVACY How can I make sure that my data don't become compromised and stored on a random server? Do I need to host the LLMs on my own server?
n8n let's you host your workflows on your own server. OpenAI's API doesn't use your data to train their model (supposedly). So we should be fine from a data security perspective.
@@customaistudio Not sure about that, but it's all about learning! This has some decent explanations if anyone is interested in understanding more about these semantic vectors. th-cam.com/video/OATCgQtNX2o/w-d-xo.html
If you're building simple things I guess you can get away with not knowing how to code, but anything non-trivial you're going to hit a wall. Guaranteed. Pick up Python or Javascript, it's not hard you can learn the basics in a weekend. Enough to get your mind thinking more like an engineer and able to start hacking away at things. There's just so much that you're not even able to think about unless you know how to code- like adding a validation layer to an output in case things break (and they will). Something as simple as a malformed JSON packet can break an application. There's no getting around learning to code IMO but it doesn't have to be this daunting thing.
We add in that stuff when we're really pushing to production. The goal with our approach is to show you can build an agent in minutes without knowing how to code. I agree though, knowing how to code would accelerate everything.
I build relatively complex economic games without knowing coding. I have a programmer for the advanced stuff... until you get into things like networking and databases you don't need to know how to code. This might change in a few years though if AI keeps getting better
Here we go! Someone that actually knows what he is talking about ! Thanks a lot for the content, I have quite the same stack and it’s efficient. I thank you for the inspiring and helpful information, I will rework it a lot thanks to you 😊.
FINALLY , after watching months of youtube video's listening to all the hype and fluff , Ai Agency's hype etc etc. It great to see someone who is most interested in building real life workable solutions without all the fluff and Hype. Youv'e won me, keep up the great content
Awesome man!
Good set up. N8N is the right tool if agencies are serious about these things.
If you are targeting SMB and perhaps the lower end of SMB’s, you may want to invest in creating for yourself and outreach agent schema that discerns between those companies where the $3000 automation tool is a no-brainer versus going near companies with very thin margins who aren’t likely to be customers because of it
Yea I completely agree
This price is a steal. Hiring one more person to support your business in California is $42,000 a year at minimum wage. YMMV.
A Custom AI Agent Framework video is super insightful! The innovation in creating and managing AI agents for various tasks is remarkable. #AI #Innovation #CustomAgents
Another video well-done. Informative and to the point.
Thanks for watching
With *Phlanx's caption generator* , you can effortlessly craft captions that resonate with your audience. It’s a valuable tool for any business looking to strengthen its social media presence.
Thanks a lot for this great video content!
I have a question : when you talked about the apps you developped in the part for your customers using AI Agents, did you used an agentic framework like CrewAI, LangChain Studio or n8n to develop these apps ? or hard coding approach like using Python, Visual Code and LangGraph framework ?
Thanks!
Thanks! We used a hard coding approach. This was before CrewAI came out (well he started developing before that).
Great video!
Thanks!
Thats so awesome. I would love to develop something for the distillery. I'm working on something for about 6 month to learn, test, discover. What a journey ! Would love to discuss with you about it more in details.
Let's talk! You can schedule a meet by going to our site: customaistudio.io
Where do you store the business knowledge (ie. Business background, contact knowledge, product knowledge) into your AI n8n node? Do you upload it into the vector store?
Yea I save that in the vector store.
Thanks for such value. How many people work in your agency?
We're still pretty small, ramping up now.
I think you are one of the best out there for this kind of stuff. I'm learning so quick I'll have a bunch of agents doing my things before the end of the year. I just have a question about tools, why do you need openai chatgpt plus on top of platform.openai assistants? Can't you just do it all with the assistants connetected to other workflows outside in n8n or other tools? I didn't find a way to connect a custom gpt as a an action instead of a trigger..
Yes you could! However, I like to use the Langchain nodes in n8n for the tool calling rather than the assistants model. I'm not sure it matters which way you do it though.
@@customaistudio Thanks for the answer! Ill check it out. Zapier don't connect to chatgpt, only to assistants, so I thought it was not possible. I'll probably create an account with n8n soon, do you have an affiliate link or something?
@@kbsoundcollection yep! It’s in the description.
Very informative presentation! however I am missing one big point which was not mentioned, PRIVACY
How can I make sure that my data don't become compromised and stored on a random server? Do I need to host the LLMs on my own server?
n8n allows you to host your workflows on your own server and you can use open source LLMs. Data privacy is a big thing.
i think we need to create an AI Agency Franchise World Wide 😁
How can you assure the customer that his private data isn't being used to train the LLM (i.e OpenAI).
OpenAI doesn't use API data to train the model.
it is mostly informative video!
I´d would like to set up a call but your calendar isn´t adapted to European time, let me know how we can contact! Thank you
Hey can you shoot me email. I'll be able to work around your schedule.
Thanks for your efforts
Great video. Question from a tinkerer. Why vector database? I thought chatGPT embedding sort of eliminated the need?
I'm using chatGPT to embed the data, then I'm storing it in a third party database (Pinecone).
OpenAI Assistants API will do it all for you, but I don’t think n8n supports that
Useless due too using Chatgpt , or any online services. Need an local install high security option.
n8n let's you host your workflows on your own server. OpenAI's API doesn't use your data to train their model (supposedly). So we should be fine from a data security perspective.
How do you mean a vector db is '3 dimensional'?
Poor phrasing on my part. The vectors are placed on a graph that has three axis. So I imagine it as a cube, which I said 3 dimensional.
@@customaistudio I guess I missed that example. In general, of course, semantic vectors are typically high (1000+) dimensional.
Nah my understanding of vector databased has 10xed since posting this video. You probably know more than I do about them.
@@customaistudio Not sure about that, but it's all about learning! This has some decent explanations if anyone is interested in understanding more about these semantic vectors. th-cam.com/video/OATCgQtNX2o/w-d-xo.html
If you're building simple things I guess you can get away with not knowing how to code, but anything non-trivial you're going to hit a wall. Guaranteed. Pick up Python or Javascript, it's not hard you can learn the basics in a weekend. Enough to get your mind thinking more like an engineer and able to start hacking away at things. There's just so much that you're not even able to think about unless you know how to code- like adding a validation layer to an output in case things break (and they will). Something as simple as a malformed JSON packet can break an application. There's no getting around learning to code IMO but it doesn't have to be this daunting thing.
We add in that stuff when we're really pushing to production. The goal with our approach is to show you can build an agent in minutes without knowing how to code. I agree though, knowing how to code would accelerate everything.
I build relatively complex economic games without knowing coding. I have a programmer for the advanced stuff... until you get into things like networking and databases you don't need to know how to code. This might change in a few years though if AI keeps getting better
you got the use cases right, but when you covered vector databases, you totally butchered the tech details.
I barely went into it in this video. There's a lot more to be said. I'll dive deeper into the tech details in later videos.