Listen this is crazy didn't know that was a thing with the zod thing to send a schema and forcing it to responde in such a shape this is insane, thank you such a great video
Companies should be falling over themselves to get you to explain how to use their libraries. Langchain would have a monopoly on the AI helper SDK market if they hired you to teach how to use their stuff...I still don't know what the hell a 'runnable' is
I really like your content, I hope I can learn everything about typescript, to typize better and enjoy it, the truth entertains me and I enjoy using ts in my projects, so I have many expectations of the channel videos, first of all I appreciate all the effort that there is in each video, a greeting
Totally cool 👍 More like this. I would even say: more about Vercel's AI SDK ! Is Vercel AI SDK providing the same features as Langchain - Langgraph or are they complementary i.e. "built the basic stuff in Vercel AI SDK and for more advanced stuff integrate with Langchain"?
i havent try with Deepseek with api, but you can install it with ollama. So you should be able to use the API. I just dont know for the section and structured output.
What does structured data do if it can't find what are you ask it in the data? Would happens for instance if you pass it a blank invoice in your example, will it hallucinate?
You can make optional some field. So it wont fill when he dont find the value. Else, yes, it can hallucinate a lot. llama vision 3.2 hallucinate a lot more than chat gpt 4o
what platform are you using to host your lessons on ai hero. It looks good. If you made this on your own, can you publish a template? I would love to pay for it.
Hey Matt, do you know if it possible to work with a custom REST API instead of using any already integrated LLM? I am able to save the info, but do not see the stream response withouth refreshing the page. Thank you
If this was not the best youtube video tutorial in 2025. Then what else is ? Thank you a ton
Listen this is crazy didn't know that was a thing with the zod thing to send a schema and forcing it to responde in such a shape this is insane, thank you such a great video
behind the scene its transforming the zod schema to an json schema.
The image that you used for the imageUrl example is Lake Bled, Slovenia 🇸🇮 10mins away from my home town!
🤦♂should have watched the next 5s of the video before posting the comment, of course LLM describe it for you lol
Absolutely wild!
I also wanted to comment but did waited 5s haha, amazing
Absolute banger of a video. Looking forward to more of these long-form content styles.
Thank you for making such a good teaching video, it is very clear and can be understood very quickly.
nice wanted something like this as making agents with js was a pain. I am gonna try more stuff out, thanks a lot Matt
you never disappoint, another great video ! Thanks 🙏
Companies should be falling over themselves to get you to explain how to use their libraries. Langchain would have a monopoly on the AI helper SDK market if they hired you to teach how to use their stuff...I still don't know what the hell a 'runnable' is
Another one that will stay in youtube history! Epic! ❤
Awesome, it is indeed very useful for peeps who haven't tried it yet, or even for peeps who tried and didn't get the most out of it
👏
I really like your content, I hope I can learn everything about typescript, to typize better and enjoy it, the truth entertains me and I enjoy using ts in my projects, so I have many expectations of the channel videos, first of all I appreciate all the effort that there is in each video, a greeting
More of this from you, please! Great Job
Great video, relatively short and highly informative 💪
Amazing tutorial. Thank you very much
Totally cool 👍
More like this.
I would even say: more about Vercel's AI SDK !
Is Vercel AI SDK providing the same features as Langchain - Langgraph or are they complementary i.e. "built the basic stuff in Vercel AI SDK and for more advanced stuff integrate with Langchain"?
your best video yet 👑
Thanks for this fantastic tutorial!
Great stuff Matt!
Thank you
So helpful - many thanks! Much appreciated!
Absolute Banger! Thank You
Fantastic breakdown! Thanks Matt! 🎉😊
Amazing!
Can one use the open source Deepseek?
i havent try with Deepseek with api, but you can install it with ollama. So you should be able to use the API. I just dont know for the section and structured output.
great video!
great video ! thanks
Great video ! Thank you. Is it possible to make another video about langGraph ? 🙂
What does structured data do if it can't find what are you ask it in the data? Would happens for instance if you pass it a blank invoice in your example, will it hallucinate?
Great question! I'll think about that for the course.
You can make optional some field. So it wont fill when he dont find the value.
Else, yes, it can hallucinate a lot. llama vision 3.2 hallucinate a lot more than chat gpt 4o
what platform are you using to host your lessons on ai hero. It looks good. If you made this on your own, can you publish a template? I would love to pay for it.
how do we query vector database and parse it instead of the pdf
Longest if you don't count the lives ;)
Great stuff
I think i saw the book on the back: "Designing Data-Intensive Applications" is it worth reading?
It's still in the plastic wrap!
it's a loaded book. I'd consider it more as a "reference manual" than a "back to back read".
It is a great book. Highly recommended
Thanks!
Hey Matt, do you know if it possible to work with a custom REST API instead of using any already integrated LLM? I am able to save the info, but do not see the stream response withouth refreshing the page. Thank you
Genuine question, is that similar to pydanticAI?
pydanticAI for Python folks and Semantic Kernel for .NET folks
How efficient is this SDK to say something like Spring AI for the backend?
So good
Do you accept apprentice ? master
very cool
Awesome
matt stopped wearing his iconic high collar sweater
Gotta get the shirts out for the courses eh
Also I literally have 4 of those sweaters, I love them
Slovenia mentioned
We were both doing it today, apparently
🤯
great video !