I agree. Leon is truly good. I feel some of these videos are rampantly copied without being given credit. I found one eerily similar to another a month later here with no credits.
Thanks for sharing. Small tip . If you want to limit AI problems related to hallucinations especially before connecting to or sending requests to external service, e.g. gmail , use wait node and give it some time, example 10 sec, to prevent repetitive requests.
Thanks Nate! I’ve been loving your content recently, definitely going to join the community now. You mentioned what we would like to see- so much of business happens through Slack. Maybe a pinecone store with Deepseek to talk to slack messages?
Absolutely love your videos-they're so informative and well-made! Keep up the amazing work! 🙌 Would you consider making a video on how to update data in Pinecone?
Great work. The only thing is so far without tools the cost savings is not there since we still have teo use a model that can use tools. Unless I missed that sorry 🙁 But I really like how you show to use it directly and the different chat vs deepseek-reasoner really helpful.
Thanks for the video Nate, I am trying to follow along here as a non tech guy. It seams you skipped a step at 3:19. How did you get to that screen where you input the api data and the url? It says test r1 open ai on the top left. I searched for the deepseek tool and could not find it.
I had the same issue, as I'm also new to n8n as well, just choose open AI as your chat model then edit it, to change the model to deepseek-reasoner, make sure to switch input type to expression because fix input wont allow you to input text.
Amazing video man! Thanks for sharing this with us! Do you have a page where you list all the content available in the paid community? I would like to know what is there before joining
Has anyone noticed that when using Deepseek V3 as the chat model to call tools, it keeps repeatedly calling the tool with the same query until reaching the max iterations, whereas 4o Mini does not have this issue?
Encounter the same problem. Using Deepseek chat instead of chatgpt mini keeps repeating the calling of the tools. So problem addressed at beginning of the video is not completely solved for the second step. It works for the planner using Deepseek reasoner. Strange behavior... Nevertheless great tutorial!
It can still call functions. Function calling is just structured output. prompt it to output your “Function calling” in json format and just verify and parse the JSON. Old school. lol
Hi. I build AI agent based on your tutorial . I want to access chat from my APP or Website . How can I do it ? I don’t want to log in to n8n each time I want to ask question. Does anyone knows how to ? Any tutorials ? ❤ thanks
i don't get it where do you get the deekseek r1 in n8n? th-cam.com/video/tjaD65OCoE8/w-d-xo.html and how do you gt to that screen to et up a new credential with baseurl?
Yes that is true, still super cheap models. If this framework is more efficient though, I will take better performance for an extra cost 10 out of 10 times!
Could you create a video demonstrating how GPT handles situations where key information is missing? For example, imagine I want to send an email to a client. I draft the entire email but forget to include the client's email address. In this case, GPT should respond with something like: "Thank you for the email draft, but I need the client's email address to proceed. Could you please provide it?" Additionally, it would be great to show how GPT can remember the email address for future interactions, so the next time I want to send an email to the same client, GPT already has the information and doesn’t need to ask again.
Huge fan of your content 💪 I just released a community node on top of Vercel AI SDK that has an implementation of a DeepSeek Node 😄 it's probably pretty much the same as you did with the baseurl, but looking to also build more around it: th-cam.com/video/rgpKD7BZHc8/w-d-xo.htmlsi=n5n1Hbe-vv0zk0ri&t=608
Hi. I build AI agent based on your tutorial . I want to access chat from my APP or Website . How can I do it ? I don’t want to log in to n8n each time I want to ask question. Does anyone knows how to ? Any tutorials ? ❤ thanks
duuuude, i just spent an all nighter troubleshooting this, figured this out this morning and went to sleep, when i woke up you posted this 😔
Haha unfortunate timing!
Credit to Leon Van Zyl on showing us on have to use R1 without open router! I’ll give him the credit for it!
I agree. Leon is truly good. I feel some of these videos are rampantly copied without being given credit. I found one eerily similar to another a month later here with no credits.
@sleepless-nite Please give me the yt video link..
Thanks for the credit 😊.
OpenRouter is definitely not needed when you can hit the DeepSeek endpoint directly.
Nate, keep up the incredible work, bud.
@@leonvanzylJust watch one of your videos yesterday. Looking forward to more of your content.
I'll give the docs the Credit... 🙄
Thanks for sharing. Small tip . If you want to limit AI problems related to hallucinations especially before connecting to or sending requests to external service, e.g. gmail , use wait node and give it some time, example 10 sec, to prevent repetitive requests.
Interesting. I’ll have to play around with that concept.
great video, big work around man. I will get involved in the community. This stuff is fun to learn.
Thanks Nate! I’ve been loving your content recently, definitely going to join the community now. You mentioned what we would like to see- so much of business happens through Slack. Maybe a pinecone store with Deepseek to talk to slack messages?
Absolutely love your videos-they're so informative and well-made! Keep up the amazing work! 🙌 Would you consider making a video on how to update data in Pinecone?
Great work. The only thing is so far without tools the cost savings is not there since we still have teo use a model that can use tools. Unless I missed that sorry 🙁
But I really like how you show to use it directly and the different chat vs deepseek-reasoner really helpful.
Awesome tutorial. You are amazing
is there any tool capable LLM that can run in a 3090 able to do the task the tool agent just did ?
Thanks for the video Nate, I am trying to follow along here as a non tech guy. It seams you skipped a step at 3:19. How did you get to that screen where you input the api data and the url? It says test r1 open ai on the top left. I searched for the deepseek tool and could not find it.
I show it at around 2:30. This is just configuring a new credential in n8n
I had the same issue, as I'm also new to n8n as well, just choose open AI as your chat model then edit it, to change the model to deepseek-reasoner, make sure to switch input type to expression because fix input wont allow you to input text.
you have to update your n8n
Such a clever system. What is the pricing like for these software?
n8n can be free if you host it locally, otherwise you can start at $27/month
Amazing video man! Thanks for sharing this with us!
Do you have a page where you list all the content available in the paid community?
I would like to know what is there before joining
I would just check out the “About” section and please don’t hesitate to reach out if anything is unclear!
Has anyone noticed that when using Deepseek V3 as the chat model to call tools, it keeps repeatedly calling the tool with the same query until reaching the max iterations, whereas 4o Mini does not have this issue?
This is one of the issues I address in this video
Encounter the same problem. Using Deepseek chat instead of chatgpt mini keeps repeating the calling of the tools. So problem addressed at beginning of the video is not completely solved for the second step. It works for the planner using Deepseek reasoner. Strange behavior... Nevertheless great tutorial!
cool, thanks I just built it
Did it work , following this video?
It can still call functions. Function calling is just structured output. prompt it to output your “Function calling” in json format and just verify and parse the JSON. Old school. lol
That would work in code. But can you make it work in AI Agent in n8n as a model?
Thanks for this. Have you tried the reasoner model with react ai agent instead of just tool agent?
Do this work if you are running n8n locally via docker? I see no base url option no matter what I do.
You may want to try installing DeepSeek locally using Ollama
how do you use pinecone vector databases with the Deepseek model?
As of now, that would be considered tool calling, which is something this model doesn’t support.
how to fine tune the ai agent to improve performance
hey Nate I got the same error " The resource you are requesting could not be found No endpoints found that support tool use." how did you resolve it ?
Bad request - please check your parameters
deepseek-reasoner does not support Function Calling, it's giving me this error. Can someone help?
Please watch the beginning of this video, I address this issue.
the best!
If Gpt 4o mini agent comes after deepseelk agent, is n't it a risk that gpt 4o mini ruins the request by deepseek? May be I don't get it entirely.
can't find the model.. any tips?
You have to change the base URL of the OpenAI chat model. I show this in the video
do i need to learn json?
It would be helpful to understand JSON, yes!
why this video dont have 1080p or 1440p?
Traveling and recorded on my laptop rather than my usual setup :/
Hi. I build AI agent based on your tutorial . I want to access chat from my APP or Website . How can I do it ? I don’t want to log in to n8n each time I want to ask question. Does anyone knows how to ? Any tutorials ? ❤ thanks
i don't get it where do you get the deekseek r1 in n8n? th-cam.com/video/tjaD65OCoE8/w-d-xo.html and how do you gt to that screen to et up a new credential with baseurl?
You click on “Create a new credential” within the OpenAI chat model. I show this at 2:26.
For me it doesn't show the base URL just API key and ID
@@coachXYZ update your n8n. It only show on the new versions
@@ronnaldhawk thank you
Wouldn't this cost you more money than just using the one model since you're paying for deepseek + open ai's tokens?
Yes that is true, still super cheap models. If this framework is more efficient though, I will take better performance for an extra cost 10 out of 10 times!
👍👍
Incrível!
Could you create a video demonstrating how GPT handles situations where key information is missing? For example, imagine I want to send an email to a client. I draft the entire email but forget to include the client's email address. In this case, GPT should respond with something like: "Thank you for the email draft, but I need the client's email address to proceed. Could you please provide it?"
Additionally, it would be great to show how GPT can remember the email address for future interactions, so the next time I want to send an email to the same client, GPT already has the information and doesn’t need to ask again.
Nate Herk, please allow your videos to be dubbed, please
Nate Herk, permita que seus vídeos sejam dublados por gentileza.
is it free?
Please am having issues on joining your Skool community
You can join from this page for free: www.skool.com/ai-automation-society/about
why do people even bother with openrouter then, with deepseek n8n setup if it's crap?
OpenRouter is a pretty popular tool, so wanted to show it off
Dude, why don't you enable multi-language?
Openrouter has gone awful rate limiting like a crazy man. Even when u pay it still limits all the time I gave up with it.
Good thing we aren’t using OpenRouter!
openrouter costs are devious tho
Good thing we aren’t using OpenRouter!
Huge fan of your content 💪 I just released a community node on top of Vercel AI SDK that has an implementation of a DeepSeek Node 😄 it's probably pretty much the same as you did with the baseurl, but looking to also build more around it: th-cam.com/video/rgpKD7BZHc8/w-d-xo.htmlsi=n5n1Hbe-vv0zk0ri&t=608
Hmm
DeepSeek so slow!
Its reasoning capabilities are super powerful.
too basic
Hi. I build AI agent based on your tutorial . I want to access chat from my APP or Website . How can I do it ? I don’t want to log in to n8n each time I want to ask question. Does anyone knows how to ? Any tutorials ? ❤ thanks