N8N ULTIMATE COURSE 8+ Hours (Sell $10k+ AI Workflows)

0:00Hey, welcome to the most comprehensive

0:01NAD building master class. This is over

0:03eight hours of pure live automation

0:05builds and my goal is to take you from a

0:07pure beginner and turn you into somebody

0:08who can build professionalgrade AI

0:10workflows that a business is willing to

0:11pay thousands of dollars for. This is

0:13your first time here. I'm Nick. I scaled

0:15my own AI automation agency to 72K a

0:17month. And I now run the biggest paid AI

0:19automation community with around 3,000

0:20AI automation freelancers and AI agency

0:22owners. Most of them land their first

0:24paying client within 2 to 3 weeks of

0:26joining thanks to our daily

0:27accountability program and our proven

0:28frameworks. So, I don't just want this

0:30to be another cookie cutter NAND

0:31tutorial. There are millions of those.

0:33My goal instead is for this to be a

0:34comprehensive guide to real building. In

0:37particular, I'm going to be creating

0:38eight high value NADN workflows that you

0:40guys can start selling immediately. What

0:42that means, I'm not just going to show

0:43you a pretty finished product. In many

0:44cases, I'm actually going to build this

0:45entire thing alongside you live as if I

0:47have no idea what I'm putting together,

0:49and I'm going to do it while narrating

0:50my thought process out loud cuz I think

0:51it's much more instructive for people to

0:53see what a real build process looks like

0:55versus just the shiny finished product

0:57at the end. So, we're going to cover

0:58everything from the absolute foundations

1:00here to advanced systems that can

1:02command $3,000 to $10,000 per

1:04implementation. I've also added

1:05timestamps for every section in the

1:06description, meaning you guys can just

1:08jump around to what you need most and

1:09then come back for the rest. And also,

1:10just make sure to bookmark this video so

1:12you can reference the workflows again

1:13and again as you build and scale your

1:15automation company. Okay, whether you're

1:17looking to automate your own business or

1:18start an automation agency or maybe just

1:20add some highv value skills to your

1:21freelancing toolkit, welcome to the

1:23complete nadn building masterass. Before

1:25we start, this master class is all about

1:26live builds. The purpose is to show you,

1:28as I mentioned, what a real live build

1:29process looks like. And then I want you

1:31to internalize the parts of automation

1:32that you can't really learn by reading,

1:34only by watching somebody who knows what

1:36they're doing do it. So to make sure

1:38everybody's on the same page, this first

1:39section is going to include an

1:40introductory course on the foundations

1:42of NAD. I'm going to include content on

1:43simple logic, workflows, some basic

1:46nodes, and maybe some more. I have this

1:47elsewhere on my channel, so if you guys

1:49have already seen all of this stuff, you

1:50can jump right to the next section. The

1:52key here is to build a solid foundation

1:53that will support everything else we do

1:54later. Because when we start building

1:55those $5,000 plus automation systems,

1:58you do need to understand some core

1:59fundamentals inside and out before you

2:01can get to the good stuff. Let's now

2:03dive into the NAN foundations that are

2:05going to make everything else possible.

2:06So, if you guys have seen some of the

2:07other videos on my channel, you'll know

2:08that I put a very big emphasis on

2:10building things practically. I don't

2:11really care much for the academic side

2:13of things. I prefer we just dive right

2:14in and then teach some of these concepts

2:16by actually putting nodes together,

2:17making workflows that you could sell for

2:19business purposes or implement into your

2:20own companies. And that's what we're

2:22going to be doing today. We can't really

2:23get away from doing some of the academic

2:26stuff with JSON just because, you know,

2:27if you don't know JavaScript object

2:29notation, you are going to have to learn

2:30some things like types and objects and

2:32what variables are and stuff like that.

2:34But for the most part, we're going to be

2:35learning all of these concepts down here

2:37by building two workflows. The first

2:39workflow is up here. And essentially

2:41what this does is it feeds in a bunch of

2:43lead data to artificial

2:46intelligence. This is the Google sheet

2:48containing four leads with a bunch of

2:50information here. So we're going to feed

2:51in this information to AI. Then what

2:53we're going to do is we're going to have

2:54AI generate a subject line for an email,

2:56an icebreaker for an email, an elevator

2:58pitch for an email, a call to action,

3:00and then a post script little PS sign.

3:02The idea here is this is a real workflow

3:04that people pay me for. And so these are

3:06the ones that I want to I want to start

3:07with. Essentially, this will allow you

3:09to customize email outreach that it

3:10seems as if you've done a lot of

3:12research into the person, uh, which is

3:13very, very valuable in a business

3:15setting. The second workflow is a little

3:17bit more peculiar, I guess. There's this

3:19service out there called source of

3:20sources. It used to be called helper

3:22reporter at Haro. Basically, the way

3:23that it works is this lovely gentleman

3:25here, Peter, will send you uh a bunch of

3:28information basically where journalists

3:30are looking for professionals in a

3:32certain industry to weigh in on some

3:35developments. And then if you're a

3:37professional in the industry and you

3:38give the journalist some good info, they

3:40can actually tag you and then like use

3:41you in an article, it's a quick and easy

3:43way to basically get listed in a

3:44magazine or some very authoritative data

3:46source. And what this system does is it

3:49basically gets an email like this. Then

3:51it pumps in the titles into AI, does

3:53some cool processing, and then we

3:54actually write a draft of that email as

3:56if we were an expert in that field,

3:58which all you need to do is just like

3:59quickly review, give a once over, edit a

4:01little bit, and then send off to

4:02journalists. There's another system that

4:03I've sold a number of times. And so I

4:05want the videos that I create on NAND to

4:07be practical in nature. I want them to

4:09be on things that you're probably going

4:10to be using for business purposes. So

4:11these are the two systems that uh we're

4:13going to be building. They're cool

4:14systems and all, but you know, for the

4:15purpose of this video, I kind of want to

4:16build them from scratch. So why don't we

4:17just exit out of that puppy? That's the

4:19prompt that we're going to be using for

4:20AI. And for the rest of this, we'll just

4:23get rid of that. Okay. The very first

4:25concept I want to cover are fixed fields

4:27and expression fields. I'm going to be

4:28basically convincing you to just use

4:30expressions all the time. And I'm also

4:31going to show you how to map different

4:32field inputs because the last time that

4:33we jumped around Nadn, we built a couple

4:36of workflows, but I was sort of glancing

4:37over at some of the nuance behind fields

4:39and stuff like that. So, for the

4:40purposes of demonstration, it's going to

4:41be pretty easy. We're just going to

4:43build the first system out by clicking a

4:44button that is going to get a bunch of

4:47data from our Google sheet and then

4:49we're just going to pass all of that

4:50data in kind of rowby row into AI.

4:53Pretty simple, pretty straightforward,

4:54but ultimately something that is very

4:55useful and you'll find yourself doing

4:56quite a bit if you do cold outreach. So,

4:58first things first, I'm just going to

4:59press tab. That's going to open up this

5:01trigger thing on the lefth hand side.

5:02And then, you know, if you type trigger

5:04here, there'll be if you type trig, I

5:06should say, you'll see a ton of options

5:08here. Um, you could either do that or

5:09you could just scroll down to the bottom

5:10where it says add another trigger and

5:11then press trigger manually. Either is

5:13fine, but for the purpose of this demo,

5:15I'm just going to do that. And basically

5:17what I want to do is the second that I

5:19run this trigger, when I click on test

5:20workflow, um, I want to, uh, get all of

5:23the rows and all the data in my Google

5:24sheet. So, I'm going to go up here to

5:25nodes. I'm just going to type in sheet.

5:27We're going to get Google Sheets. Now,

5:29what I'm going to do is if you scroll

5:31down here, you'll see that there's this

5:32one node called get rows and sheets.

5:34That's what I'm going to click on. Now,

5:36I've already done a connection before.

5:38Um, what you're going to have to do if

5:39you want to connect to your Google

5:40Sheets account is go to create new

5:42connection. Now, because I'm on the

5:43cloud hosted offering, and keep in mind

5:45if you're not on the cloud hosted

5:46offering, um, these sorts of connections

5:47are a little bit more difficult. You

5:49have to go to Google Cloud Console and

5:50get set up there. But because I'm on the

5:52cloud hosted offering, all I need to do

5:53is click sign in with Google over here.

5:55And I'll actually go connect to my email

5:58account and then it'll create a

5:59credential for me, which is pretty

6:01handy. So you can see it's saying it's

6:03already has some access just because

6:04I've already done this connection. But

6:05for the purpose of this demo, I just

6:06wanted to show you guys what that looks

6:08like. And then I'm just going to save

6:10this connection here with a name so that

6:11it's just nice and organized. Okay. And

6:13what we want to do is we just want to

6:14grab this Google sheet up here, right?

6:16So, in NADM, there are a variety of ways

6:19to do this, but I'm just going to I'm

6:21going to be looking for a document.

6:23Sorry, a sheet within a document. My

6:24bad. The document we're going to want to

6:26select is from the list. And we're

6:27actually going to go down. We're going

6:28to choose this one called leads, comma

6:30space, January 27th, 2025. So, as you

6:34can see, we've already manually found

6:35that. Okay. And then the specific

6:38subshet that we want is this sheet one

6:41because I guess that's the only one that

6:42we have here. So, I'm just going to

6:43click on this. It's going to actually do

6:45an API call to Google Sheets. It's going

6:47to find that there's only sheet one

6:48here. Then I can give it a click and

6:49then voila. Let me just run test step

6:52and let's see what happens when I click

6:54test. Okay, great. So, I've offiscated

6:56this data. This data is not um actually

6:58one for one whoever this this person is.

7:00I've gone through and I've like renamed

7:02them and stuff like that just for

7:03privacy purposes. But as we see on the

7:05right hand side here, we we have a bunch

7:06of output. And we can see the output in

7:08a variety of ways. Tabular, JSON,

7:10schema. most people in and they like the

7:12schema look because it just kind of

7:14compresses the information nicely. It's

7:15a little bit easier for them to see. I'm

7:17a little zoomed in here too. Um but uh

7:19yeah, you know, normally like you see

7:21most of the node variables which is

7:23pretty handy. But this JSON one over

7:26here, this is really intimidating for a

7:28lot of people. And so we're we're going

7:29to cover this in detail. I'm going to

7:30show you exactly how you read JSON, what

7:32all those things mean. But just I just

7:33want you guys to know that for the

7:34remainder of this course, I'm going to

7:35be using primarily the JSON and the

7:37schema view. But I'm actually going to

7:38tend towards JSON. And the reason why

7:40I'm going to be tending towards Jason is

7:41because like JSON says all the same

7:44stuff that the schema view does anyway.

7:46But unfortunately part of the way you

7:47learn JSON is just by kind of staring at

7:49it a lot and squinting at it and kind of

7:50inherently and intuitively understanding

7:52the formatting. If we're going to be

7:53looking at outputs all day anyway, we

7:55might as well kind of get, you know,

7:56kill two birds with one stone. All of

7:57the same data in JSON is represented in

8:00schema anyway. It's just instead of like

8:02the quotes around key names and stuff,

8:03you just have sort of this light gray

8:05box alongside like a type sign here. So,

8:07I guess the point I'm making is we might

8:09as well double up and just learn how

8:10JSON looks while we're proceeding with

8:12the course. Um, and that's why I'm going

8:13to be using this. Even if it looks a

8:14little bit more intimidating, don't

8:15worry too much about it. Okay. So, I

8:16said that I'd talk about fields, right?

8:18So, fields in N&M just as I was covering

8:20on the previous video are stuff like

8:22this, right? We have this sort of center

8:24node config option here for our Google

8:26Sheets node. Um, you know, one of the

8:28fields we selected was this YouTube

8:31credential to connect with. Another one

8:33was this resource sheet within document

8:35operation document sheet. But I want you

8:37to know that these are actually all

8:39representable in code as well. So as you

8:43see over here, we have two different

8:44types of fields. One's called fixed and

8:46the other is called expression. So by

8:48default, just to keep your life easy and

8:50to not like freak you the hell out,

8:52especially if you're like a newbie and

8:53an um they're going to keep all the

8:55fields to the fixed type. But if you

8:57click on

8:58expression, you'll see that things are

9:00now a little bit different.

9:02Notice how when I went to fixed, we had

9:05like a nice little, you know, it said um

9:07leads January 27, 2025. And then when I

9:09jump over to expression, now we have

9:10some big long ID. So what is this? Why

9:13is it structured that way? And what

9:15exactly does any of this mean? Well, if

9:18we pay close attention, this ID field

9:20here, 1 lowerase o t r r r r r r r r r r

9:23r r r r r r r r r r r a t 4 C capital C

9:25and then the rest of this big long ID

9:27string over here. If we go to our Google

9:29sheet, what you'll see is that that ID

9:31string actually matches the URL of our

9:33Google sheet

9:35exactly. This takes me to a wider point.

9:37Most of the time, sorry about that. Most

9:39of the time, anytime you're accessing a

9:40resource on some API or even just on the

9:43internet, they will store the ID of the

9:45resource, which is sort of like a hidden

9:47representation of it, in the URL. So,

9:51you know, one of the examples that I

9:52provided the other day was I went over

9:53to ClickUp, right? And inside of

9:56ClickUp, I like searched around for some

9:58record. I'm just going to click on this

9:59here. This is this was my old content

10:00calendar. Um, how to send 1,000 cold

10:03Instagram DMs per day. That was one of

10:04the things I wanted to do. This right up

10:06here is the ID of the record.

10:1086B27A7Zm, right? If I wanted to do

10:13something with this through their API,

10:14this is the ID of the record that I

10:16would be calling. So, I want you to know

10:18that like ClickUp, uh, Monday, uh, like

10:20even Gmail, basically every service out

10:22there, they will store the ID of the

10:24thing you want to modify or update or

10:26whatever just in the URL. So if ever a

10:29field asks for an ID, you can almost

10:31always just go to the URL of the thing

10:33on the actual user um you know on the

10:35actual app like the user interface, then

10:37you can find that URL thing. You just

10:39hardcode it in here. Okay, so that's

10:41just a just a brief look at some of the

10:43differences between fixed and

10:45expression. Basically fix a lot of the

10:46time just to to keep you guys um to make

10:49sure that like we're on the same page

10:50here. Fix is just the simpler version.

10:53Then expression is sort of what's

10:54actually going on under the hood. So,

10:56you know, um, NAN just defaults to fix

10:58like it just did here because it doesn't

11:00really want to scare you away. But in

11:02order to really unlock the value of NAN,

11:05we have to go to the expression, um,

11:07field. And I'm going to show you here

11:08why, you know, I basically just use

11:09expression for everything at this point.

11:12Okay. So, looking at the output here,

11:14what we have is we have we clicked and

11:16then we got a click event that was

11:17counted as one item. And so, N8 actually

11:19shows you the number of items that are

11:20passed on. Then we pumped that click

11:23event into Google Sheets and then that

11:25click event outputed four items. So now

11:26we actually we're working off of array

11:28data or tabular data which I'll cover um

11:30in a moment. But essentially with these

11:32four items now what we want to do is we

11:34want to pass each of these items into

11:36artificial intelligence and we want to

11:37have AI tell us something about it

11:39before um writing some cold email copy

11:42for us to insert into an email or maybe

11:43we could just send a Gmail directly or

11:45something. So what I'm going to do is

11:46I'm going to click on this button. I

11:48zoom out a little bit. What we want is

11:50we want um if you go down advanced AI,

11:53we want is we want this open AI node.

11:55And specifically what I want is I want

11:57the message a model. So if you've seen

11:59me connect this in the previous video,

12:01you'll notice that you know in order to

12:02create a new credential, you actually

12:03have to go and you have to find the API

12:05key from OpenAI in order to do this.

12:07This is pretty simple to do. You just

12:08type

12:10platform.openai.com/ I think it's like

12:11account/appi or something. You'll have

12:13to click on this button here to open the

12:15documentation to tell you more. But

12:17anyway, you can basically just create an

12:18API key for Nadn. Um, it's very simple

12:21and very straightforward to do so. And

12:22I've already done this and I've called

12:24it YouTube. So, I'm just going to use

12:25that credential just so I don't have to

12:27like leak another API key. Now, again,

12:29we have a ton. We have a ton of fields.

12:31We have resource text operation message

12:33and model from list choose. Keep in mind

12:36the fields are again fixed, right? It's

12:38trying to make it really easy for us to

12:39select GPT40. So I'm just going to go

12:41down here, type

12:42GPT and four. And then I'm just going to

12:46click 4 O. And now we actually enter in

12:48the text that we're interested in. Okay.

12:51And this is really where you're going to

12:52start learning the differences between

12:54the fixed and the expression. So fixed

12:56again, fixed is just I mean it's what

12:57the name implies. It's fixed. It's text.

12:59You can't make this dynamic. You can't

13:01add variables to it. It's the simplest

13:02way to get up and running with a node,

13:04which is why nad will default to fixed.

13:07Um, but over the course of the next few

13:08minutes, I'm going to convince you to

13:09basically always just use expression.

13:12Okay. So, I I saved my prompt somewhere

13:14else. The first thing I'm going to do is

13:15I'm going to add a system prompt. So,

13:16I'm going to go down here to system.

13:17I'll just say you are a helpful

13:19intelligent writing assistant. Usually,

13:21the way that you will do um AI calls is

13:23you will have a system prompt first.

13:26Then you'll have a user prompt after.

13:28And the user prompt is where you

13:29actually give it the instructions you

13:30want it to do. So, you know, in our

13:32case, it's going to be like, hey, I want

13:33you to write a bunch of fields that are

13:34templates that we're going to insert

13:35into a cold email later. And then after

13:36you have the choice to provide a bunch

13:37of examples. So you could provide an

13:38assistant prompt and then you could do

13:40another user prompt. Assistant prompt.

13:41User prompt. Assistant prompt. You can

13:43do that however many times you want just

13:44to show it how things work. For the

13:46purpose of this example, I'm just going

13:47to be providing a single user prompt.

13:48And what I'm going to be doing here is

13:50I'm just going to copy over my prompt

13:52below. Let me paste that in. And let's

13:55just read through this together. Your

13:57task is to personalize an email. You'll

13:59do this by taking as input a prospect

14:01LinkedIn profile. Then editing five

14:03templates for different sections of the

14:05email. Subject line, icebreaker,

14:07elevator pitch, call to action, and a PS

14:09or postcript field. If you're unfamiliar

14:11with postcript, you know, at the bottom

14:12of an email, it'll just say PS, I really

14:15miss you. Can't wait to see you. That's

14:17what a postcript field is. We basically

14:18want AI to automate that for us because

14:20there's a lot of value in making those

14:22postcript fields seem human written.

14:25Anyway, now we're offering it some

14:26templates. subject line. Hey, name I

14:29think I have something for you. Re and

14:31then cool thing about them that we

14:33discovered. Let's just go unique thing

14:35about them or their company. Icebreaker,

14:39I know you're doing thing and I've been

14:41following related thing for a while, so

14:43I figured it made sense to chat. Next

14:45elevator pitch the TLDDR. I think I can

14:47add 5K a month to their paraphrase

14:50business with a few automated systems.

14:52And then there's a call to action. I

14:54just did this for a very similar

14:55industry company and we had 28,350 in a

14:58few months. They do related things. So

14:59I'm very confident I can duplicate this

15:01at minimum. Would be 100% risk-f free. I

15:04guarantee at least 20 appointments

15:05booked or you wouldn't have to pay.

15:07Pretty neat, huh? So then I give it a

15:09bunch of guidelines. And feel free to

15:10pause the video if you want to take a

15:11look at it. The last thing that I do is

15:13then I say respond in JSON using this

15:17format. And if you've at all used AI

15:19before, you'll see this JSON thing come

15:20up again. JSON. JSON. JSON. We're going

15:22to cover that in just a few minutes, so

15:24buckle up. Okay, great. So, we've given

15:26it a ton of uh instructions in the first

15:28user prompt. So, what I do next is I

15:30just give it a user prompt with the

15:33actual body of the input that I want to

15:35give. And if you think back here um you

15:38know what this Google sheet is just just

15:39so you guys um are all on the same page

15:41as me is I've basically gone and I've

15:42scraped a bunch of data about random

15:44people on the internet that fulfill some

15:46criteria that I have. So, chief

15:47executive officer, director of demand

15:49generation, director of marketing,

15:50business development, manager, some

15:52dentistry person or something. Okay. And

15:54then I have a bunch of fields here. One

15:56of the fields I have is I have a summary

15:57field where people basically write their

15:58own summary of who they are and what

16:00they care about. We have a ton of other

16:01fields as well. We have company

16:02location. We have a description of their

16:05title. We have um I don't know their

16:07their industry. And the cool part about

16:09AI is you could just feed this into a

16:10large language model and have it have it

16:11automate something for you. Have it have

16:13it write something customized. And so

16:15that's what we're going to be doing

16:16here. The thing is though, right, how do

16:19I get dynamic data into this? So, I

16:22don't know, let's say um one of the

16:24things I want is I want to feed the AI

16:25the person's full name. Notice how this

16:28is fixed here, right? If I just typed

16:29Amy Wabby, then that means that every

16:32time I do an API call that includes

16:35their full name, I'm going to have it

16:36say Amy Wabby. This is fixed. It's the

16:38same thing every time. If you want to

16:40make this dynamic, what you have to do,

16:42there are variety of ways to do this,

16:43but I'm going to use the expression

16:44field, is you have to click

16:45expression and then you drag the field

16:48that you want and then you drop it. And

16:51you'll see that when I do that, we've

16:53now just inserted a little bit of code.

16:56This is in N8N's

16:59um code format, the equivalent of the

17:02variable that we just pulled from our

17:04pinned data or our our data from the uh

17:06input. And what you see down here is

17:08this is separated into two halves. The

17:10top half is the code representation. We

17:13said full name just in like regular

17:14characters. And I can manipulate this

17:16how I want. Then a colon, then a space.

17:18And then there's uh curly bracket curly

17:20bracket space dollar sign Js N. FU L N A

17:26M E space. And then right curly bracket

17:28right curly bracket. Up here you have

17:30the code representation. And then notice

17:31that underneath here we have result. It

17:33actually shows us what the data that

17:34we're pulling in is from the input,

17:37which was right over here. Now, I'm

17:39going to be feeding it a bunch of data

17:41in order to have this personalized. I'm

17:42going to be feeding in their full name.

17:43I'm going to be feeding in the summary.

17:44I'm going to be feeding in the title.

17:45But I just want you guys to to notice

17:47how these these variables change between

17:50the full name, between the summary,

17:52between the title. And you're going to

17:53notice that there's kind of a pattern

17:54there. Okay. So, full name was that

17:57we'll go title. I'm just going to drag

17:59this

18:00in. Paste it in. Notice how the first

18:03one was jso nf full name. Second one was

18:06jso n.title.

18:08Right. Let's see what the third one is.

18:11Let's go down here to

18:13company. Company. If I drag and drop

18:15this, it now says JSON. Company. You

18:18know that the C is capitalized. That

18:19looks a little bit new, but for the most

18:20part, it's still pretty

18:22self-explanatory. It seems to me if I

18:25were an alien staring at this and

18:26looking to try and figure out what the

18:28pattern here is. It seems to me that

18:29every single time I drag and drop one of

18:31these fields in there, it says dollar

18:32sign jso n dot and then the name of the

18:36variable. And the name of the variable

18:37tends to be whatever I'm looking at on

18:39the left hand side here. So what if

18:41hypothetically instead of me doing this

18:43drag and drop what if I were

18:46to actually just try and write this

18:48myself? Well, let's see what happens. If

18:50I zoom in a little bit, just so we could

18:52all see. If I go curly bracket curly

18:54bracket, you'll see that I'm now

18:57entering sort of the next level up in

18:59NN. I'm now manipulating like code or

19:02JavaScript, their version of JavaScript,

19:03the JMSE path I believe it's called,

19:05directly in the expression editor. And

19:07this is where N gets really powerful

19:09because you also have a ton of built-in

19:10methods and built-in ways you can

19:12manipulate this data with literally one

19:14click, one little button tap without

19:15having to drag and drop all these

19:16modules everywhere. You can just do so

19:17in the convenience of your own field

19:20editor. Okay, so the very first thing

19:21that pops up is it says suggested JSON,

19:23if I just type that and then I press

19:24enter, you'll see that now I have access

19:26to all of the fields that I had access

19:29to earlier. So instead of me dragging

19:31and dropping all this stuff, what if I

19:33just wanted to write the word summary

19:34here to grab this. If I just type

19:36summary, notice how this now turned

19:38green. And we've added all of that

19:39information down here to the results

19:41tab. All that information is here. How

19:44cool is that? So now, you know, if I

19:47want to continue on, I'll go industry.

19:49I'll go dollar sign JSON industry.

19:52Voila. Next, we're going to go company

19:54location. I'm going to go um JSON.co

19:57company location. Voila. Notice how it's

20:00trying to autofill this for me, right?

20:02I'll go title description, dollar sign

20:04JSON dot to ital description. Voila. And

20:09I basically have the ability to do this

20:11um infinitely depending on how nested

20:13the data is in the JSON structure of the

20:15input. I I'll run through how to do all

20:17of that um in a moment, but I just want

20:18you guys to sort of pattern match look

20:20from the outside in. How am I actually

20:21referencing all these variables from

20:23from previous

20:24calls? Okay, great. So, to me, you know,

20:26as somebody that does this sort of

20:28personalization all the time, if we

20:29click on this little button here, we can

20:30actually open up we can see all of the

20:32um code and all of the text. To me,

20:34somebody that does this all the time,

20:35this looks like sufficient amount of

20:36information for us to personalize an

20:38email off of. So, I'm actually just

20:39going to call it there, and we're

20:40actually going to just run this puppy.

20:41But basically what's going to happen is

20:42we're just going to be feeding in all of

20:44this stuff to artificial intelligence

20:46and we're going to be saying, "Hey man,

20:47based off of all of Amy's info, I want

20:49you to tell me something about her and

20:51then I want you to write an email um

20:53based off of the template that I

20:54provided you earlier." The last thing

20:56I'm going to do is I'm going to go down

20:57here and press output content as Jason.

20:58Give this a click. And then I'm not

21:01going to modify any of the options here

21:02either. Um but I'm just going to click

21:04test

21:05step. Okay. Okay, so what just happened

21:07or what is occurring as we speak is I'm

21:09feeding in four items to open AI. I'm

21:11basically Blitz going item one, item

21:14two, item three, item four, and it's

21:16happening all at once before they show

21:18us the output of each of these. So

21:20that's why it takes a little bit longer

21:21than usual. Um, but this is more or less

21:23what's happening under the hood. Okay,

21:25great. And we just received an output.

21:27Um, so what I'm going to do is I'm just

21:28going to zoom out a little bit just so

21:29we could see this in completeness. And

21:32I'll use schema for now just to make it

21:33easier for you guys. You don't have to

21:35like scroll all the way to the right to

21:36see it, but let's take a look. Um, the

21:38output was content, subject line,

21:41icebreaker, elevator pitch, call to

21:43action, PS. So, it actually went it

21:44outputed five fields for us and we can

21:47use those fields in future nodes very

21:50easily. The first thing in the subject

21:52line was, "Hey Amy, think I have

21:54something for you regarding boosting

21:55online strategies." The icebreaker was,

21:57"I know you're leading creative web

21:58solutions and net directives and been

21:59following innovative marketing

22:00strategies for a while, so I figured it

22:01made sense to chat." The TLDDR, I think

22:03you can add 5K a month to your client

22:05focused internet marketing efforts with

22:06a few automated systems. The call to

22:08action, I just did this for a very

22:10similar IT consulting company. We hit

22:12this amount. They do e-commerce and

22:13marketing, too. How cool is that? So,

22:15I'm very confident I can duplicate this

22:16at minimum. Would be 100% risk-f free.

22:18I'd guarantee at least 20 appointments

22:19booked. You wouldn't pay. P.S., even if

22:22we just chat, I'd love to hear about

22:23what you're doing with video marketing.

22:25That sounds pretty cool to me, right? If

22:26I were to receive an email like this,

22:28you know, aside from the subject line,

22:29which is a little bit vague, boosting

22:31online strategies, but you can't fault

22:33the AI for not being perfect 100% of the

22:34time. Um, you know, even if I were to

22:37get something like this, it would seem

22:38as if, uh, you know, like the person

22:40that's reaching out to me did their did

22:42their research at minimum and is

22:43reaching out to me sort of in a

22:44personalized customized way as opposed

22:46to just like blasting me a big sequence.

22:48So, uh, what else do I want to do with

22:50this? Well, if you guys, you know,

22:51remember back to the beginning of the

22:52video, the there weren't just two or

22:53three nodes here. there was um there was

22:55a note that updated the Google sheet.

22:56So, I'm going to show you how to update

22:57the Google sheet. And what we're

22:58actually going to do is we're going to

22:59take this one step further and I'm

23:01actually going to draft some emails to

23:02send to Amy and the rest of the people

23:04here. So, actually the first thing I'm

23:06going to do is I'm just going to go over

23:07here and pin this output. The reason why

23:09is because if you think about it, me

23:11calling um OpenAI there, that was a

23:13little bit like computationally

23:13expensive. It took me a little bit of

23:15time. I don't actually want to have to

23:16rerun that over and over and over again.

23:18pinning the data just allows me to

23:20capture that, cache it, and now I can

23:22just like test all subsequent nodes

23:23using that, which is very

23:25straightforward. Okay. So, what I want

23:27to do, I want to update this Google

23:28sheet. So, I'm going to click here,

23:29search nodes. I'll type sheets. And what

23:31I want to do is I want to um update row

23:34and

23:35sheet. I'm going to select my

23:37credentials again, YouTube. The resource

23:39will be sheet within document. The

23:40operation will be update row. Let's

23:42instead of using the fix, let's actually

23:44just use the expression so I could show

23:45you how this works. Just going to grab

23:47the ID of this field or of this sheet.

23:50Paste that in there. Voila. Then the

23:52sheet that I'm going to be picking, it's

23:53just going to be sheet one. So notice

23:55that the top here we use the expression

23:57field view and then down here we just

23:59use the fix. So it actually use the

24:00expression field view to do an API call

24:02to their backend to discover that sheet

24:04one was the only sheet and then we

24:06selected it there. We can also just feed

24:07in an expression. And as you see when

24:09you go from fix to expression for the

24:11specific sheet type, it just says G

24:13equals Z. G equals Z just refers to the

24:15first sheet. um basically here. So we're

24:17we're always going to be selecting the

24:19first sheet. But anyway, for this I'm

24:21actually going to go from list and then

24:22I'll just go sheet one. Keep this

24:24simpler. Okay. Now it's going to be

24:25fetching a bunch of columns for us. And

24:27basically in order to do this update,

24:29what we have to do is um we have to grab

24:32the data from the previous nodes and

24:33then we have to update every single

24:35column here with that data. If I were

24:37just to, you know, scroll all the way

24:39down and update the columns that I care

24:41about like subject, icebreaker,

24:43elevator, pitch, call to action,

24:44postcript, it would just leave the rest

24:46of these blank, which is kind of

24:47annoying if I'm being honest. But I

24:49don't want the rest of these to be blank

24:50and these to be filled. I want all of

24:51this data because I'm just going to

24:52import this into some cold email tool

24:53later, right? So, what we have to do is

24:56now, this is kind of the initial idea

24:58behind this system. I'm going to go

25:00through and I'm going to update every

25:01single one of these um using the

25:03expression tab and then I'm going to be

25:04pulling data in. But I'm not going to be

25:06pulling data in from here. I'm going to

25:07be pulling data in from one node behind

25:08it. So you guys could see what it looks

25:10like in code basically. Okay. So first

25:12thing we have to do is we just need a

25:13column to match on. In order for the

25:15automation to know which row should be

25:18updated, we have to find data that

25:20includes that email. So I'm going to go

25:22email. Then the first thing I'm going to

25:24do is I'm going to scroll down here to

25:26the previous node, not the OpenAI node,

25:28but the Google Sheets node, the one that

25:29like first listed us the data. I'm going

25:31to drag this feed that in there. And

25:34you'll notice that the format now looks

25:36different than it did before. Previously

25:38we had a dollar sign JSON dot item.

25:40Right now we have a dollar sign and then

25:43a bracket single quote sign the name of

25:45the node another single quote sign and

25:48then another bracket and then we go do

25:50item.json.e. So when you when you access

25:52node data from more than one node back

25:54you have to use this new format here

25:56which is kind of annoying but you'll see

25:58how easy it is when we just like kind of

25:59copy paste and spam our way through. So,

26:03uh, how about this civility? I'm going

26:04to go to expression. I'm going to paste

26:05this in. Then, instead of email, I'm

26:07just going to go

26:09civility. How about this first name?

26:12Let's paste that in there. We'll go

26:13first name. How about this first name

26:16suggestion? Paste that in there. Go

26:18expression. Go first name

26:21suggestion. How about the last name? I'm

26:24going to paste this in there. Go last

26:26name. How about the full name? I'm going

26:29to paste that in there. Go expression.

26:31We'll go full name. How about the title?

26:33I'm going to paste this in there. I'll

26:35go title, profile URL. So, you can see,

26:38you know, we're picking up the pace a

26:39little bit. It's getting a little bit

26:40faster and faster.

26:43Company, company, illegal

26:48name, company

26:51phone. And I'm just going to go ahead

26:52and um cut to me having actually filled

26:55all this stuff out just for brevity.

26:56Okay. Okay. Elevator pitch, call to

26:59action. Looking good. And then last but

27:03not least, we'll do um I think it was

27:05just PS or was it postcript? Yeah,

27:07sorry. It was just PS for that. This is

27:09the one situation which uh the column

27:11name is a little bit different from the

27:12variable name down here. Okay. Now that

27:14we're done with that, let's just quickly

27:16cover um just some little differences

27:18here in the formatting just so we can

27:20get a runup on the JSON which I'm about

27:21to teach you. You scroll all the way up

27:23here. You'll see that some of these or

27:25actually most of these followed this

27:26format. It was oops sorry about that.

27:30Let's go over here. It was um curly

27:33brace curly brace dollar sign the name

27:36of the node in single quotes dot

27:39item.json dot whatever the value was

27:41dots summary. In this case this one was

27:44regular company URL. This one down here

27:46wasvm ID. You'll notice that a few of

27:48these are actually a little bit

27:49different. A few of these in particular

27:51the variable names with spaces. They

27:52weren't just dot, you know, the the

27:54name. It was square bracket single quote

27:58and then the name of the variable that

28:00we're referencing. The reason why we had

28:02to do this instead of just doing the dot

28:05and then the variable name is because

28:08JSON in JavaScript object notation, you

28:11can't query I mean the technical term is

28:12you can't query a key that has spaces

28:15basically just because spaces aren't

28:16really represented. So, you know, if we

28:18scroll down here to some of these

28:20variable names, first space name space

28:22suggestion, there's a bunch of spaces in

28:23there, right? So, the way that you get

28:25around this in uh Nad's formatting is

28:28instead of calling, let me get that

28:30specific example, company legal name or

28:32was it uh first name suggestion. So, the

28:36way that you get around this is you

28:37can't just go first name suggestion cuz

28:40that kind of breaks the formatting here,

28:41right? The way you can get around this

28:43is two things. one, you can make it so

28:44that the input data aka the columns in

28:46your Google sheet are all just one word.

28:48So if instead it was first name

28:50suggestion, this would be fine. You

28:52could also do something like first name

28:54suggestion. Some people do that format.

28:55I don't really like that. I don't know

28:57why. What I do is um called camelc case.

29:00It's kind of like a programming

29:02convention. And here we can get into

29:03like a a lifelong mutually assured

29:06destructive battle where some people

29:07prefer camel case, other people prefer

29:09underlining. Um, but I'm team camel

29:11case. So, go team camel case. Um,

29:13instead, you know, if we want to

29:15represent the spaces, what we have to do

29:16is we have to go brackets here. Um,

29:18single colon first name suggestion,

29:21another col uh another quote sign, and

29:23then another square bracket. So, don't

29:25don't sweat the small little formatting

29:27stuff too much. I just wanted to give

29:28you guys like the best way that I found

29:30to learn this is literally just to like

29:31spam a bunch of examples. Um, that's

29:33what I did when I was picking this stuff

29:34up. I didn't read a bunch of books on

29:37JavaScript object notation or

29:38expressions or whatever. I just spammed

29:40a bunch of examples. And the human brain

29:42is such that if you squint at it long

29:43enough, you'll sort of figure it out

29:45intuitively. Uh, okay, great. So, now

29:46that we've mapped all these, let's

29:47actually go and let's, um, let's update

29:49this data, right? That's the whole point

29:50of this. So, uh, now that we've mapped

29:52all the data, if I click test step, then

29:53I go over here to my Google sheet,

29:55scroll all the way to the right, you

29:57see, and it just took us a second, but

29:58you see that we just, boom, we just

29:59updated all four of these

30:00simultaneously. Uh, we got the subject

30:02line for Amy, subject line for Joe,

30:04subject line for Mercedes, subject line

30:05for Susan. Same thing with all the

30:07elevator pitches, call to actions,

30:09postcripts. You'll see that the the the

30:11copy of the email is pretty similar, but

30:13um it it changes. So, this one's, I know

30:15you're leading creative web solutions. I

30:17know you're le into leveraging customer

30:18voices. I know you're focused on

30:19building relationships. I know you're

30:20advancing digital dentistry. These are

30:22all basically like reframing or

30:25paraphrasing

30:26um the things that they said in their

30:28profile which ultimately uh you know is

30:30meant to make them go like oh okay this

30:31person did their research they read a

30:32little bit about me. So that's pretty

30:34cool. Um why don't we take this one step

30:36further now. Why don't we pin this and

30:37then I'm actually going to create a

30:39Gmail draft in my inbox just so we can

30:41see what's going on here. So I'm type

30:43draft create a draft. I'll connect with

30:46my Gmail credential and this is before I

30:48did the naming convention so it was

30:49probably number three. We're just going

30:51to create a draft. The subject line is

30:54going to be, let's just go to fix now

30:55that we know how to do this. It's going

30:57to be JSON dot uh subject right

31:02here. And then the actual message. It's

31:05going to be pretty interesting, but I'll

31:06show you I'll show you how we put it

31:07together. First thing we're going to do

31:08is we're going to go JSON dot. And then

31:10what I want to do is I want an

31:12icebreaker.

31:14Uh, sorry, I don't want an icebreaker. I

31:17want to go hi. And then I want to grab

31:19the person's name. So I'll go Jason

31:21first name right

31:23here. So hi

31:26Amy, then we have the the JSON

31:28icebreaker. Next up, I want to go JSON

31:30dot uh sorry, dollar sign JSON dot and

31:33then what were we doing here? Was it the

31:35elevator pitch? Yeah, it was the

31:36elevator pitch. Paste that in there.

31:39Then we want to go dollar sign JSON dot

31:41what's the next one? Call to action.

31:44Beautiful. And we also want to go dollar

31:47sign JSON dot and I'm sure you guys can

31:48guess what the last one is, but very

31:50quickly make sure you do make sure you

31:52know u postcript. Now if we open up this

31:55thing in the bigger example window,

31:57you'll see the email says, "Hi Amy, I

31:59know you've been leading creative web

32:00solutions and net directives. Been

32:01following innovative marketing

32:02strategies for a while, so I figured it

32:03made sense to chat. The TLDDR, I think I

32:04can add 5K a month to your client

32:06focused internet marketing efforts with

32:07a few automated systems. I just did this

32:08for and even if we just chat, I'd love

32:10to hear about what you're doing with

32:11video marketing." I guess I actually

32:13need to add a PS sign here. And it looks

32:14like none of these have periods.

32:16Actually, I think this last one has a

32:17period, but not all of these do. So,

32:19just because I was a little bit um fast

32:21in kind of designing this and I didn't

32:23put periods over, I'm actually just

32:24going to add the periods directly into

32:25the expression editor. I'm going go PS

32:28here. Okay. And now, if we open it up,

32:30this is what it looks like. We got

32:32periods everywhere. Cool. Wonderful. And

32:34then we have a little PS sign here. Even

32:36if we just chat, I'd love to hear what

32:37we were doing with video marketing.

32:38Very, very cool. Awesome. Awesome. So,

32:40now that we have that pinned data, um,

32:42why don't we just draft these emails?

32:44So, I'm going to, uh, create a draft.

32:45I'm not actually going to send this cuz

32:46I don't just want to spam a bunch of

32:48these people. I've also changed the

32:49email addresses and stuff, so I'd get a

32:50bunch of bounces. We're just going to

32:52test step. It's executing. We just

32:55executed four nodes. So, now if I go

32:56over here, go to

32:58drafts, you'll see here that we have,

33:01"Hey, Susan, think I have something for

33:02you. Red digital dentistry." Oh, gez,

33:04I'm realizing I didn't put the um email

33:05address in actually. Yeah, I did not put

33:08the email address in. I just did the

33:10draft. U in options, you have to go to

33:12email. And then what we want to do is we

33:13just want to drag that back here. JSON

33:15email to get that address. Let's create

33:17four more drafts just for shits and

33:19gigs. Why don't I just delete the ones

33:21that I just generated uh right over

33:25here. Going to just discard

33:27these. And then you see that the four

33:30that I just did now have just popped up.

33:31And they also have the email addresses

33:34there. So yeah, that's that with that

33:36example. Um, I think at this point you

33:38guys probably have an intuitive

33:39understanding of how these fields work.

33:41And hopefully I've made a case for why

33:43you should just always use expression.

33:44Like there's no real need to do fixed

33:46because if you think about it, like you

33:47could just write the same fixed thing.

33:49Like if I wanted to type, hey, think I

33:50have something for you regarding

33:51marketing strategies. And I just wanted

33:52to send the same thing every time. I

33:54could do so with the expression view of

33:56the field, right? Like same thing. It's

33:58just here I also have the ability to u

34:01modify uh code a little bit and like do

34:03something if I wanted to. So, I

34:06personally am always going to be using

34:07this moving forward. And the reason why,

34:10sorry, it can be kind of a lot to see if

34:12you muck around with. Just always put a

34:13dollar sign first if you're referencing

34:14data from the previous node. Uh, and the

34:16reason why is it's just going to be the

34:18easiest for me. Um, you know, you get

34:20all the variables up here anyway. It's

34:21no big deal. So, we're doing subject.

34:23Stick that in there. We got the subject

34:24line. We are good to

34:26go. So, uh, I think we probably learned

34:28a fair amount about the fixed fields and

34:30expression fields at this point. Um, the

34:32last thing I wanted to cover, I say

34:33mapping different field inputs here.

34:34Last thing I wanted to cover was uh if

34:37we go back to Gmail

34:39here and if we scroll down to actually

34:43this is a bad example. Why don't we go

34:45back to open

34:46AI? You see here how um I selected the

34:50RO field and it was fixed and then I

34:52selected from a dropown user assistant

34:54system whatever. Well, I can actually

34:55just go expression and what you see here

34:57is this is just text that we are feeding

35:00this API the uh the node. We're actually

35:02just writing user. So instead of writing

35:04users, you could also write a system.

35:05This fix stuff. This is just made to

35:07make our life a little bit easier. But

35:08anytime you want to get the actual data

35:10representation, just move over to

35:11expression. So what are some examples of

35:13that? Well, like you see these simplify

35:15output and output content is JSON

35:16fields. We go to expression. You'll see

35:18that what we're actually passing is

35:20we're passing this true value. We're

35:22literally just writing true. Um output

35:25content is JS. If we move it to

35:26expression, you can see we're actually

35:27just passing true here for output

35:29content is JSON. The point that I'm

35:31making is um all of this is basically

35:33just um you know if we just strip away

35:35all of the basic simple stuff. You can

35:37start getting into what's actually going

35:39on behind the scenes. How we're actually

35:41communicating with these nodes. The way

35:42that we're actually communicating with

35:43these nodes in practice is we're sending

35:45the term true. Now the reason why this

35:47one has quotes around it I believe is

35:50because um true is a special case. It's

35:52called a boolean which I'm I'm about to

35:54cover right now. Um so I believe you

35:55need these quotes just for NAN to like

35:57not bug out at you. um just one of those

35:59unfortunate peculiarities of the

36:01platform. But that's that for fixed

36:03fields and expression fields. Okay, next

36:05up, let's talk more about JavaScript

36:07object notation. And if you guys already

36:09know JSON, if you've already

36:10experimented and and understand the

36:12various formatting types available here,

36:14feel free to skip on until we move on to

36:16how data in N8 is represented. Um should

36:18be about 15ish minutes or so. Uh but for

36:20everybody in the audience that doesn't

36:21understand JavaScript object notation, I

36:23just want to go real deep into it and I

36:25want to make sure you understand

36:26everything about JSON because ultimately

36:28as opposed to a lot of other no code

36:30platforms, NAD it doesn't ignore code or

36:33try and shove code away. It actually

36:35embraces code. So understanding a little

36:36bit of some of the um code types like

36:39JSON for instance makes you way more

36:41powerful. This is really what like

36:42everybody that makes money with this

36:44platform uses. We just use JSON to send

36:46data back and forth. Um, and you know,

36:48if you don't know JSON, everything's

36:49going to be a little bit trickier for

36:50you. Okay. So, what's the best way to um

36:54intuitively understand JSON? Well, what

36:56I'm going to do next is I'm going to

36:57walk you through the various data types

36:58in JSON. Uh, I'm going to give you just

37:00like a a brief little um structure and

37:02format, and then we can actually just

37:03walk through JSON kind of step by step

37:06depending on different variables and

37:07stuff like that. So, I'm just going to

37:08go and type in JSON formatter. This is

37:10just the simplest and

37:13freesting platform I found online. As

37:15you can see, we got tons of ads in the

37:16middle here. you don't pay anything to

37:17use it. But basically what's happening

37:19under the scenes is behind the scenes is

37:21it's running um with every time you

37:23press a keystroke it's just double

37:25checking to see if it's a valid JSON.

37:26This is actually pretty useful for us

37:28because we can we can verify if

37:29something is indeed JSON or not just by

37:31looking at it. This isn't the only way

37:32to do it. Obviously we could do a ton of

37:34different things. We could I could open

37:35up like a code platform like VS Code or

37:37something and do the same thing. But I

37:38just wanted to keep this thing

37:39accessible for all of us. All right. All

37:41right. So, zooming way in here, um, just

37:44so I don't get ads in my face 24/7. So,

37:47what is JavaScript object notation?

37:49Basically, JSON um is just a way to

37:54represent data in a structured and

37:58standardized

37:59format that minimizes the number of

38:02characters and it also minimizes the

38:04ambiguity so that when we send data to

38:07and from some API or something, we could

38:09do so as efficiently as possible.

38:12So the way that JSON works, the way that

38:15you send and receive data is based off

38:17of two um two concepts. The first

38:20concept is a key. So I'm going to write

38:22a key here. This key is going to involve

38:24my name. The first name is going to be

38:27sorry um the the key name is going to be

38:28called first name. So the key is just

38:31you can think of it as like the name of

38:32a variable. So the variable is called

38:35first name, but what the variable equals

38:38is a whole different matter. And that's

38:40where the value comes

38:42in. So first name I'm going to set to

38:46Nick. So this here is proper or well

38:49formatted JSON. It's not yelling at us

38:51or anything like that. It's it's it's

38:53actually good, which is nice. So this is

38:54an example of one of the simplest

38:56JavaScript objects that you could build.

38:59It is a onekey uh one value object. The

39:02key being first name and the value being

39:04neck. You've undoubtedly seen examples

39:07of this before if you tried working with

39:08any noode or coding platform um you know

39:12nadn included. The thing is in JSON

39:16there are a few simple but consistent

39:18formatting quirks that you just have to

39:20pay attention to. So the first is there

39:23are variety of different data types. Now

39:25as you can see here what I've done is I

39:27have this open curly bracket and then

39:28close curly bracket. I have the key name

39:31over here, a a colon, that's just these

39:33two dots, and then I have the

39:35value, but what I've done is I've

39:37wrapped everything in these double quote

39:39signs here. So, first name, double

39:41quote, Nick, double quote. The reason

39:43why is because the data type that I'm

39:45going with here is called a string. It's

39:47actually a particular data type. A

39:48string is just like written text. Okay?

39:52So, this is if if instead of first name,

39:54I say ID. This here, this is a string

39:57data type, but it's not like you don't

40:00only have strings available to you.

40:01Although a lot of platforms prefer

40:02strings, you also have a variety of

40:03other data types. Here's another data

40:05type we have access to. This is a number

40:08data type. It's numeric. So, in order to

40:11um you know send and receive numbers,

40:13you don't actually have to wrap them in

40:14quotes. This red here just corresponds

40:17to it being linted or formatted or

40:19whatever by this JSON formatter as a

40:21number data type. So this is still valid

40:23JSON even though we don't have the

40:24quotes around this. Okay, this is not

40:28valid JSON, right? This does not mean

40:31anything because we're using string

40:33characters uh and we're inserting it in

40:35something that the software that we're

40:37going to be using NN is going to be

40:39expecting to be a number. So basic rules

40:42of thumb are uh you can't just write a

40:44string without wrapping it in

40:46quotes. And the unfortunate reality is

40:49numbers can be both numbers and strings.

40:52So three ways to do X. Maybe we call

40:55this titles, right? The three here, this

40:56is a number, right? But we are still

40:58wrapping this within the quotes of a

41:00string. However, this is different from

41:04this. And so in practice, the reason I

41:06harp on this is because in practice,

41:09this usually doesn't matter that much.

41:10No code platforms will take care of like

41:12the type conversion for you between

41:13number to string. Um, if you're using a

41:15number, if you're sending it as like

41:17just the three without the quotes around

41:18it, um, for the most part, that's that's

41:20okay. you'll be able to use this as a

41:21string later on. But in like pure

41:23JavaScript and a couple of programming

41:25languages, you can't actually just like

41:26if you wanted to, I don't know, run a

41:28function after this that added a number.

41:31Um, you know, maybe it was like here,

41:34let me show you number of things, then

41:38title template.

41:40Let's say I had some function where I

41:42wanted to add the number of things to

41:44the beginning of the title template so

41:45that it said three ways to create NAN

41:48flows, five ways to create N inflows, 15

41:50ways to create NAN flows. Some

41:52programming languages wouldn't let you

41:54do that in which case you'd have to

41:55convert this to um you know a string and

41:57then you could just go number of things

41:58plus title template equals 15 ways to

42:00create NAN

42:01flows. Okay, great. So we've covered

42:04numbers. We've also covered uh and

42:06numbers I believe they have some fancy

42:07name int or or technically they could be

42:09floats. It could be a number of things,

42:10but um in our case, we're just going to

42:12go numbers. Next up, I want to cover a

42:14couple of additional data types. So, one

42:17of the data types is called bool. So, a

42:20bool stands for boolean. Boolean just

42:22means zero or one, true or false. So,

42:25true. See how this just turned into

42:26orange instead of gray. Um true is an

42:29accepted bool. And you don't actually

42:30need to wrap uh quotes around this true

42:32in order for it to technically be valid

42:34JSON. Now, for the most part, I will

42:36always just wrap quotes around all this

42:37stuff anyway because, as I mentioned,

42:38they do type conversions and it doesn't

42:39really matter to me too much. But I just

42:41wanted you guys to sort of explain why

42:43sometimes you see stuff that is not

42:45wrapped in quotes. And that's one of

42:46them. You also have a few higher level

42:48data types. And one of the higher level

42:50data types um that I'm going to show you

42:51guys, I'm going to show you two. The

42:53first is I'm going to show you an array.

42:54An array looks like this. It's with

42:55square brackets. So, you have a left

42:57square bracket and then a right square

42:58bracket. And the other is another

43:01JavaScript object. And the really cool

43:03part about JSON is you can infinitely ne

43:06nest different data types within the

43:08values of uh of a key. So I could

43:13theoretically wrap an array which is

43:16just a number of things and then inside

43:18of that

43:19array I could wrap a number of other

43:23JavaScript objects that go infinitely

43:25deep. And I'll show you guys how to do

43:27this in practice, but maybe we'll just

43:29go items. Then here we'll go first name

43:32Nick. Here we'll go first name and we'll

43:35go Sally. Then here uh let's just do two

43:37so I don't run off the page. So now what

43:40we've done is we've created JSON where

43:42we have a key called items. Then inside

43:46of said key we have an array and inside

43:49of set array we have two more objects.

43:52Both objects have the key called first

43:55name. And then they have different

43:56values. The first has a value of Nick.

43:58The second has a value of Sally. Now, to

44:01make people's lives easy, usually the

44:02way that this works is you will do some

44:04level of uh formatting like this string

44:08formatting just to make your life really

44:10really easy. You'll be able to

44:12see like where the item starts and then

44:15there's usually like a fair amount of

44:16indentation and I don't know the exact

44:17amount of indentation, but usually it

44:19looks something like this. Now, at a

44:20glance, you can kind of see the

44:22structure of this. It's sort of nested.

44:23You have um this top level array, and

44:25inside of the array, you have two

44:27objects. first name Nick, first name

44:29Sally. Okay, great. So, let's create a

44:32hypothetical JSON object. And just

44:34because I know the most about myself,

44:35I'm going to be creating this for

44:36myself. Let's hypothetically just create

44:38a JSON object. And let's just do like a

44:40user object. Okay. So, what I'm going to

44:43do is I'm going to go user. That's going

44:45to be the key name. Then I'm going to go

44:47uh colon. And then I'm going to create

44:50another object.

44:54Inside of my user object, I'm gonna have

44:55first

44:56name. First name is Nick. I'm gonna have

44:59a last name. My last name is

45:02Sarafh. Um, I'll have city. My city for

45:06the time being is Calgary. I'll have

45:09um foods he enjoys or foods. Let's just

45:15call it food preferences, right? Because

45:16if you say foods he enjoys, now you're

45:19um insinuating that has to be a he. And

45:20then what if you add a user in the

45:22future that is a woman? Do you want to

45:23change the key name? No, obviously not.

45:25So, food preferences. I'm just going to

45:26add an array. And inside of my array are

45:28going to be a bunch of strings. So, one

45:29of my food preferences is um I don't

45:31know,

45:32Thai. I'm going to have another one that

45:35is uh I don't know, like Japanese. Okay.

45:38So, I like Thai food and I like Japanese

45:39food. Apparently, I like absolutely

45:41nothing else. I have a very strict diet.

45:43Okay, cool. And then why don't we add

45:46um let's just add one more. And then I'm

45:49just going to add one called friends

45:50hypothetically. Um, but because as we

45:52all know, I have absolutely no friends.

45:54Um, and then inside of friends, we're

45:55going to add a another object. We're

45:58going to add an array. And inside of

46:00that, we're going to have first name.

46:02We're going to have Peter, my uh dearest

46:04and longest friend, of course. Thank

46:06you, Peter. Then underneath that, we're

46:08going to have um last

46:12name Griffin because I am now

46:16uh love and family guy. And that's one

46:19of my friends. And then I'm just going

46:20to copy this over. I'm going to add a a

46:23comma because you need a comma in

46:24between all items in an array and also

46:25all items in an object. So we have our

46:28object here. It's a pretty intense user

46:31um object. We have a first name, last

46:33name, the city the person lives in, the

46:34food preferences. We have a list of

46:36friends, right? The reason why I go into

46:39this much uh detail here to create this

46:41object is just because I want to impress

46:42upon you that you can make an object

46:43arbitrarily detailed. You can make it as

46:46detailed as a human being could ever

46:47possibly want. And a lot of the time

46:49you'll see that these APIs that you're

46:50accessing, these these calls that you're

46:52making to these different nodes that

46:53return you data about a particular

46:55software platform, they will be really

46:57really deep and they will have nested

46:58data like six or seven things uh levels

47:01deep essentially. So the easiest ones to

47:04use in my experience are just the ones

47:06that are all um you know surface level.

47:09You just have like a user object and

47:10then a first name and then a last name,

47:11a city and then maybe some food

47:12preferences. But in practice, you know,

47:14we need to be a little bit more capable.

47:16And understanding how data is structured

47:18in JavaScript object notation is uh

47:20probably like half the battle to be

47:21completely honest. So, I believe I've

47:24covered everything I wanted to cover.

47:25The last thing I'll mention is um you

47:27can't have a comma, I believe, be the

47:29last. Yeah, it doesn't allow you to have

47:31a comma for the last item in an object

47:35or an array. So, notice here how we have

47:36a curly bracket, then we have the key

47:38name user, then we have another object

47:40sign, and then we have the key name, and

47:42then the value. the key name and the val

47:44the key name and then the value. Between

47:46each of these, we have a comma, but just

47:48know that at the last item in an array

47:51or the last item in an object, you can't

47:53have a comma. So, you just need to

47:55remove that. If I press format or

47:56beautify, which is just my button that

47:57tells me everything's okay. Um, then you

47:59see that it is indeed formatted and

48:00beautified. Everything's everything's

48:02all right. Okay, cool. So, that's the

48:04meandering um you know, that's sort of

48:06how to understand this stuff from a

48:07bird's eye view. Moving forward, what

48:09I'm going to be doing is just to make

48:10our lives a little bit easier. I'm going

48:11to be showing you guys in every module

48:13or every node I should say on what the

48:15JSON is of the input and the output. The

48:17reason why is it's just going to make it

48:18a lot easier for us to ultimately, you

48:21know, reference stuff and then do cool

48:22things with the data. So, I go back to

48:24our Google Sheets example from earlier.

48:26I'm just going to move the input, sorry,

48:28I'm going to have the output be JSON.

48:30And you'll see that in this case, what

48:32we have is we have an open JSON curly

48:34bracket. Then we have a bunch of top

48:37level keys. So row number over here,

48:40well that's a key. And you can see that

48:42it's represented as a two. Very similar

48:44to what we talked about earlier, right?

48:45You'll notice that the formatting here

48:47is a little bit different. Notice that

48:48this one has an underscore and then

48:49these other ones are all camelc case.

48:51And some of these are capitalized

48:52because these are within quotes. None of

48:54this stuff really matters. But just

48:55showing you guys that there's a there's

48:57a variety of like conventions that

48:58people do. Some people like capitalizing

49:01things, some people don't.

49:03So my recommendation to you just to keep

49:04everything really clean and and expected

49:06is just to pick one format and stick

49:07with it. If you like the underscores,

49:08use the underscores for everything. If

49:10you like the camel case like me, use the

49:11camel case for everything. But anyway,

49:13civility is Mrs. First name Amy. First

49:15name suggestion, it's empty. You know,

49:17when something is empty in JSON, you

49:19just use these quotes. Notice that every

49:21item here has a comma in between it.

49:24Last name, comma, full name, comma,

49:26right? Just going all the way down. And

49:28then if I zoom out, I scroll, you'll see

49:31that what do we have here? This was just

49:32one item. So, if I just click on this

49:34little blue left curly bracket, it'll

49:36actually nest the item for me. And

49:38you'll see that inside of this item are

49:4038 other items. But, it'll just nest it

49:41for me. And then I can see the next one.

49:43Now, I can see the next one. Now, I can

49:44see the next one. Right? We only have

49:46five. So, that'll be it. But now, I want

49:48to show you guys a little bit about how

49:50an format stuff because it's a little

49:52bit different than Jason I just showed

49:53you. But once you understand this,

49:54basically everything from here on will

49:55be child's play. Notice how the very

49:59first character that you see on the

50:01output page on the JSON is a square

50:04bracket. Well, if you think back to the

50:05example that I gave you earlier, if we

50:07click out of this um Pizza Hut ad, as

50:09much as I love Pizza Hut, this array

50:12here starts with these square brackets.

50:15And so we see over here we have a square

50:16bracket as well, that means that this is

50:17an array basically. So the way that data

50:21is represented in NADN is all data on

50:23the platform is represented as an array

50:26of objects. Okay. So if I scroll down

50:29here to how data in NAN is represented.

50:31All of it is an array of objects. All of

50:33it will basically be those two square

50:36brackets on the outside of a number of

50:38items. And you could have as many items

50:40as you want. As we saw, we had four back

50:42over there. And inside of those four

50:44items, we had an additional 38 items

50:46nestled inside. But all data is an array

50:48of objects. Meaning that that's how you

50:50send and receive multiple objects.

50:53Remember at the beginning we had one

50:54item. We just outputed a single variable

50:56there. Well, now after this Google

50:57Sheets call, we're doing four items.

50:59Those four items are represented in our

51:01JSON as an array inside of which is

51:04another object that has the key name row

51:06underscore number, civility, first name,

51:08first name, suggestion, and so on and so

51:10on and so forth. So all data in NAD is

51:12an array of objects. The number one

51:14gotcha that I always see when beginners

51:16start out with this is they end up not

51:19fully understanding how many objects the

51:21previous node is sending or the current

51:22node is receiving. And so then when they

51:24try and reference a particular item

51:26inside of the data structure, they find

51:28that it just doesn't exist or they need

51:30to like index it, they need to slice in

51:31to find a specific item inside of the

51:33array of items or some other problem.

51:35But this is the number one problem that

51:37I see in practice just running through

51:39reading like N8 and help threads and

51:40stuff like that that people struggle

51:41with on this platform. So if you

51:44understand this the fact that all inputs

51:46and all outputs are an array of items

51:48represented as follows and now if you

51:50understand how to format JSON what JSON

51:52looks like and now further if you just

51:54make it so that you look at the JSON and

51:56not necessarily the table or the schema

51:58format every time you're sending and

51:59receiving data you will solve the

52:00biggest gotcha that people have in their

52:02net it's just going to be a lot easier

52:03for you moving forward to you know

52:05create flows you're you're basically

52:06going to take so much of the load off

52:08that most other people spend like hours

52:10and hours and hours trying to debugging

52:12um right from the get- go, which is

52:13ultimately what I wanted to do here.

52:15Okay, great. So, now we know that data

52:17in NAN is represented as an array of

52:18objects. So, how do you actually

52:20reference that data? You guys remember

52:21earlier in that example here where we

52:23were using this Google sheet module,

52:25sorry, this Google sheet node, I should

52:27say. And if we scroll down here, we were

52:30updating a field using data not from the

52:33previous node, but from a few nodes

52:35back. Well, basically if I go back to

52:37JSON here, what you can see is we can go

52:40one node back to open ad, two nodes back

52:42to Google sheets, three nodes back to

52:43when clicking test workflow. Okay, let's

52:46just go to this Google sheet. In order

52:48to reference data, actually in this

52:49case, we will use the schema cuz you can

52:51actually see sort of them logically top

52:53to bottom. In order to reference data

52:55more than one node back, what we have to

52:56do if we just want to redo this email

52:58one, we go curly bracket curly bracket

53:00dollar sign. Then what we want to do is

53:02we go this left round bracket. Then we

53:06have access to all of our earlier nodes.

53:07We can select the current node open a

53:09but we can also select Google sheets and

53:11we can also select when clicking test

53:12workflow we want is the Google sheets.

53:14And then if you just dotindex if you put

53:15a period then you'll get an

53:18item. And the reason why I bring up this

53:20item is because basically all of this

53:23stuff is buried under this dot item

53:28syntax. And it's kind of annoying but in

53:30order to reference I don't know the row

53:32number what we have to do is we have to

53:33reference the Google sheet. Then we have

53:35to reference the item. Then we have to

53:36reference JSON. Then we have to

53:38reference the row number. Why? Because

53:40NAN actually hides some of this

53:42information from us. We don't actually

53:43see the nested data structure. All we

53:46see is the nice representation of it.

53:47Okay. So this isn't actually represented

53:50this way despite the fact that it looks

53:52kind of like it is the way that this

53:54actually looks kind of underneath the

53:56scenes behind what NAD is showing us is

53:59it looks like this. And if you

54:00understand this, you'll understand how

54:01to reference basically any item in NAN.

54:04So this is equivalent to what we're

54:06seeing here. What we actually have is

54:07there's an additional two layers between

54:09us accessing the data through the dot

54:11thing. It's actually

54:12item.json and then we have the row

54:14number stability, first name, first

54:15name, suggestion, stuff like that. Okay,

54:17so that means that when we want to

54:18access it kind of just just looking over

54:19here, the very first thing we have to do

54:21is we have to reference the Google

54:23sheet. Then think about it logically.

54:25Then we have to reference the item. So

54:27go

54:28item. Then we have to reference the

54:30JSON.js.

54:32And then we have to reference the row

54:33underscore number if we want to like

54:35that. And then we get that data. Okay.

54:38So big issue, big misunderstanding I

54:40would say with Naden. Um unfortunately

54:42their documentation in my experience is

54:44not clear enough to really elucidate

54:46what's going on here unless you have a

54:47programming background. Um but just want

54:49you guys to know that this is how it's

54:50done when you reference nodes that are

54:52more than one module back. If you wanted

54:53to just reference something in the

54:54previous node, it's a lot easier. You go

54:56dollar sign and then you go jso n. then

55:00you can just go immediately into the

55:01variables. Message content or something

55:04like that dot subject line for instance.

55:07All right, so let's start looking at

55:08some of these foundational nodes over

55:09here because I think at this point we've

55:11done four different workflows and I

55:13think you guys probably have a

55:14reasonable understanding of how to put a

55:15workflow together now as well. Some of

55:17the more nuanced portions of NAN like

55:19the way that they do their JSON, some of

55:21the the nuances behind burying names

55:24inside of various keys and how to do

55:26arrays of objects and backtrack and all

55:28that fun stuff. Let's actually cover

55:30like nodes because nodes are ultimately

55:31what you guys are going to be using on a

55:33daily basis. After we're done that, then

55:34I'll I'll rebuild this flow up here and

55:36I'll show you guys how all that stuff

55:37works under the hood. This will allow us

55:40to use some of the more um I guess code

55:44oriented features of N&N. Then I think

55:46after this point like you guys have a

55:48reasonable enough understanding to build

55:49most things. It's just a matter of like

55:51what can you build and how exactly do

55:52you put the various Lego blocks I've

55:54shown you together. That's what the

55:55subsequent videos in this course are

55:57going to are going to ultimately just

55:58show you. just going to be non-stop

55:59building 24/7, baby. Anyway, so let's

56:03cover these foundational nodes. The

56:04first thing is we're going to cover

56:06nodes for doing things. So, HTTP

56:07requests, web hooks, and open AI nodes.

56:09And we'll cover nodes that modify flows.

56:11So, if filter, merge, split into a

56:12batches. There are a couple of other

56:13ones, but these are the ones that we're

56:14going to be focusing on, at least for

56:16today. So, you know, in terms of nodes

56:18that do things, um, if you guys aren't

56:19familiar with, um, HTTP request, this

56:22stands for hypertext, I don't know,

56:24hypertext transfer protocol. And what

56:27we're doing when we do an HTTP request

56:29is we're doing the same thing that your

56:31browser does when it goes to a website.

56:33So just like when I go to google.com,

56:36I'm sending a request to Google,

56:38receiving a bunch of information what

56:39this web page looks like, and then I'm

56:41using my browser Chrome to render it

56:43into the beautiful, wonderful image you

56:45have in front of me. The HTTP request

56:47does the same thing. It just it just

56:48returns the thing to you in in the code.

56:50You don't have that rendering portion.

56:52So I'm going to do a get request here.

56:54And the get request is just the simplest

56:56and most basic request. And this is

56:58ultimately what you're probably going to

56:59be using for for most cases. And the URL

57:01I'm going to be doing is just my own.

57:03Leftclick.ai. So leftclick.ai just for

57:05the purposes of this discussion. Looks

57:07like this. We build hands-off growth

57:09systems for B2B founders. There's a

57:10bunch of information about what our

57:11clients get, our leads. The lovely Joe

57:13Davies left me a testimonial. I should

57:15probably touch this website up at this

57:16point. It's been the same for a while.

57:17But anyway, I'm just doing an HTTP

57:19request using a method get to

57:22leftclick.ai AI with no authentication,

57:24no nothing. We're just going to see what

57:25happens. So, let me click that test

57:26step. And the response that we've

57:28received is again an array of objects.

57:31Our one object has a key name data. Then

57:34inside of that key name, you'll see that

57:37there is a ton of code, HTML it's

57:40called. Okay. And just because this is

57:42not the easiest to see here, I'm just

57:43going to go to schema view so you can

57:45see a little bit more of it. But

57:46basically, this here is the code of my

57:48website. This is what my website looks

57:50like to the browsers. um you know and

57:52like what ultimately my browser uses to

57:54render it. You can see it says leftclick

57:56space vertical carrot space AI and

57:59amperand. This is just a symbol that um

58:01allows it not to break any characters

58:03process optimization. Left click is an

58:05AIdriven performance optimization agency

58:07is cutting edge tech to scale your

58:08company. This is all the code right it's

58:10pretty badass. If I were to go to

58:11leftclick and then I were to go to

58:14inspect. If I were to open this up this

58:17is like the Chrome dev tool which allows

58:18you to see the code of the website. It's

58:20the exact same thing that's over here in

58:23the bottom lefthand corner, right?

58:24Literally no differences whatsoever. You

58:26built your agency, we'll scale it. The

58:28only difference is some symbols like the

58:29amperand or maybe the at sign, they just

58:31have like little replacements here or

58:33there so they don't break any um string

58:34formatting. So that's the HTTP request.

58:36Why why does an HTTP request matter and

58:38why would we want to do an HTTP request?

58:40Well, the reason why is cuz you can do

58:42pretty cool stuff with this. Like if I

58:45go um HTML and then I say extract HTML

58:48content. What we can do is we could use

58:51the code of the page to pull out all of

58:53the text. Okay. So why don't I just go

58:55text? I'll go P. Uh let's just do H1,

58:57H2, H3, H4, H5, H6. Let's do P and

59:01return this value as text. And then I'm

59:03going to return uh let's just test this

59:04and see what happens.

59:06What we're doing is we're basically

59:08feeding in the HTML and then we're using

59:10various CSS selectors they're called to

59:13extract a bunch of text for us. And it

59:14looks like I didn't do this right

59:15because we're not actually retrieving

59:17anything. Let's just try P for

59:19now. We'll do this. You're seeing

59:22already that we've extracted a bunch of

59:24P text. We've extracted basically like a

59:27specific type of

59:28tag inside this website that starts with

59:32a P. Okay. Now, we can do the same thing

59:34with a variety of other tags. Let's say

59:36I want to do H1. This is going to return

59:37all of the um top level selectors

59:40basically. So, let's do H1

59:43test. We build we build hands-off growth

59:45systems for B2B founders. That's pretty

59:47cool. If I return an array, let's return

59:49all of them. We build hands-off growth

59:51systems for B2B founders. A better way

59:52to build ops. What our clients get. If I

59:54do uh I think if I want to select

59:56multiple, I just would I just do P like

59:58this. H1 and P. Yeah, there we go. Okay.

1:00:02So now I'm getting all the text of the

1:00:04website. We build hands-off growth

1:00:05systems for B2B founders. Find the

1:00:06perfect offer. Automate your lead

1:00:07acquisition. Solve your project

1:00:08management. A better way to build ops.

1:00:10This is all the text of the website. And

1:00:12all I needed to do in order to feed this

1:00:14in was I basically went through and then

1:00:16I fed in a bunch of elements which I

1:00:17know just correspond to text like this.

1:00:20And then I pressed test step. It went

1:00:22through and it extracted all of them

1:00:24into this big fat array which is really

1:00:26cool. Now like if you think about it, I

1:00:28could actually do some pretty cool stuff

1:00:29with this. I could have AI tell me

1:00:32something about the site very easily.

1:00:34You're a helpful intelligent uh website

1:00:36scraping

1:00:37assistant. We'll go got to add my

1:00:40credential first. We'll go down here and

1:00:43I'm just going to have

1:00:44AI tell me a little bit about this

1:00:46website. Return in JSON just some data.

1:00:49I'll do that as my system prompt. And

1:00:51then here, your task is to take as input

1:00:54a bunch of scraped website text and

1:00:58return as output a JSON that follows

1:01:01this

1:01:03format. I'm going to say summary. I'm

1:01:07going to say three unique

1:01:10points. We'll do an array. We'll do um

1:01:14probable customer demographic. Okay, I'm

1:01:17going to have it return an object. we'll

1:01:20see contact

1:01:22information if any and then we'll do I

1:01:25don't know some sort of like array of

1:01:27objects. Okay. And then that's that's

1:01:30what we're going to have it return.

1:01:30We're going to output the content as

1:01:32JSON and then as input you add a

1:01:34message. What we want to do is we just

1:01:36want to join this is um being output as

1:01:38an array right now. Right? Arrays as we

1:01:40see are many options or many different

1:01:42um things on various lines. What we want

1:01:45to do is we want to take all of this

1:01:46output data. We just want to turn it

1:01:48into one big long string. The way you do

1:01:50that here is I would use the expression

1:01:52tab. And then I'm just going to add some

1:01:54lines so the AI knows that this is my

1:01:56input. Then I'm just going to go

1:01:58JSON.ext and then dot we'll just go

1:02:01join. Join is just a way to convert an

1:02:03array of items into just one big long

1:02:06string. And the thing you put inside of

1:02:07the join is you just put what you want

1:02:10to separate it by. So in my case, I'll

1:02:12just separate it all with a new line. If

1:02:14I enter the detailed editor here, you

1:02:16can see that now I've just turned all of

1:02:17this into one giant long string. I'm

1:02:20going to feed this into AI and I'm going

1:02:21to have it just tell me something about

1:02:22the

1:02:23website. So, this took us just a few

1:02:25seconds and we already have a scraper

1:02:27that's basically capable of coming up

1:02:29with and then outputting a summary of

1:02:31what I do, three unique points that

1:02:34separate me from the competition,

1:02:36probable customer demographic

1:02:37information maybe I can use to do

1:02:39something, then contact information if

1:02:41any. And it looks like it separated that

1:02:43into a subobject that says method, book

1:02:44a call, details, get started today,

1:02:46platform, website. As I'm sure you guys

1:02:48could imagine, I spent 15 seconds

1:02:49putting this puppy together. If you guys

1:02:51wanted to maybe scrape emails or do

1:02:52something like that, you could put

1:02:54something together that does this pretty

1:02:55easily. So that's the HTTP request node.

1:02:58Pretty simple, pretty straightforward.

1:03:00The next thing I want to show you is I

1:03:01want to show you basically the inverse

1:03:03of the HTTP request node. Instead of us

1:03:05sending data, I want to show you a quick

1:03:07and easy way that we can receive a

1:03:09little bit of data if necessary. And

1:03:11it's nowhere near as hard as you think.

1:03:12There are a variety of different ways

1:03:13that you could do this, but I'm going to

1:03:14show you guys a really simple and easy

1:03:16one that I personally use all the time.

1:03:18It's called the web hook. So, I'm going

1:03:20to go over here. Then, I'm going to type

1:03:22in web hook. Starts the workflow when a

1:03:25web hook is called. Basically, what a

1:03:27web hook is is it's just a server URL

1:03:29that you spin up, almost like my website

1:03:31leftclick. And every time that server

1:03:33gets a request to it, it'll show up here

1:03:36with all of the data. Why is this so

1:03:38valuable? Well, it allows us to do a

1:03:40million things. Connect workflow to

1:03:42workflow, add up and create our own API

1:03:44integrations, do a variety of things

1:03:46that otherwise we wouldn't really be

1:03:47able to do without them. This is

1:03:49basically the glue that holds the

1:03:50internet together. And this is a quick

1:03:51and easy way for you to make your own

1:03:53piece of that. So, this is what the web

1:03:54hook fields look like. We have a test

1:03:55URL, production URL. Don't worry about

1:03:57the distinction there for now. We'll

1:03:58just use test URL. HTTP method we want

1:04:01to allow to access our service. Think we

1:04:03might actually allow multiple. One of

1:04:04these I think allows us to do multiple,

1:04:06but we're just going to go with get for

1:04:08now. The path, we're going to leave the

1:04:10path as fixed. The path is just this URL

1:04:12string. Authentication, we're going to

1:04:14turn this off. We don't want any

1:04:15authentication. They recommend that you

1:04:16have authentication, but for simplicity

1:04:18sake, I'm just not going to have any

1:04:19because otherwise it might be a little

1:04:20bit too much at once. Then we're going

1:04:22to do respond immediately. What I'm

1:04:24going to do is I'm going to set up a

1:04:25test event for this URL. Okay. Actually,

1:04:27I don't know if I can do this with the

1:04:28same workflow. Actually, no, I can't

1:04:30because the workflow is already running.

1:04:31So, I'm actually going to make another

1:04:32workflow really quickly. And I'm going

1:04:34to use that to call this web hook to get

1:04:36some cool data. So why don't I create a

1:04:39workflow. We'll just call this three NAN

1:04:42concepts. And then we'll go web hooks

1:04:44HTTP

1:04:45requests. I'm going to make a new HTTP

1:04:49request here. It's going to get this

1:04:51long thing that I've done over here.

1:04:53Okay, I'm going to test it. Then over

1:04:55here, I'm going to grab that data. And

1:04:57as you can see, like what just happened

1:04:59if you were paying close attention is

1:05:00this just ran the second I sent the data

1:05:02from this node which is in another

1:05:04workflow over to this node which is on

1:05:06the current workflow. We received a ton

1:05:08of info and the info that we received

1:05:10was received a big object one item

1:05:12inside of our array of objects here. So

1:05:14a single object with a key name headers

1:05:17which had a bunch of other data

1:05:19underneath params query body web hook

1:05:22URL execution mode test. So this might

1:05:25look a little bit dry and a little bit

1:05:26boring to us right now. But what's the

1:05:28value here? The value is you can run

1:05:30something from one workflow, send it to

1:05:31another workflow really easily. So in

1:05:33our our first example, we sent no data.

1:05:35But what if I go back to my other

1:05:37workflow? Okay, that's number three

1:05:39here. And then I send some query

1:05:40parameters and I say first name and I

1:05:43say Nick. Then I go here a last name

1:05:47surive. Then here maybe I go UU ID.

1:05:50That's just like user ID. And then I

1:05:52type something like this. Now, if I go

1:05:54back here and I listen for a test event,

1:05:57I go back here and then I send the test

1:05:59event. Now, when I receive it, not only

1:06:01do I get the headers, sorry, the headers

1:06:04over here, not only do I get the params

1:06:06inside of our query are the variables

1:06:09that I just sent over inside of JSON.

1:06:11First name, last name, UU ID. So, man, I

1:06:14can do so many cool things with this.

1:06:15It's crazy. I could connect this to any

1:06:19web service out there. just give this

1:06:21URL as the URL where the events are sent

1:06:23and then voila, I basically have like an

1:06:25infinite machine. I'll show you guys a

1:06:28quick example right now using

1:06:30ClickUp, but you guys can extend this

1:06:32example and do whatever the hell you

1:06:33want with it. So, ClickUp is just this

1:06:36project management platform that I have

1:06:38that allows you to send out web hooks.

1:06:40And most services that are good will

1:06:42allow you to do stuff like this. But you

1:06:44see there's a call web hook feature

1:06:46here. I'm going to go back to my web

1:06:48hook and I'm going to copy this URL.

1:06:50Now, this is for get only. I don't

1:06:52actually remember if ClickUp sends a get

1:06:54or a post. So, we're going to see what

1:06:55happens here. And then what I'm going to

1:06:57do is I'll just have status change.

1:07:01We're going to say from specifically

1:07:04hook to specifically

1:07:08outline. And we're going to create this.

1:07:10And basically what this means is when I

1:07:11change a field called status inside of

1:07:13my my project manager, it sends a web

1:07:15hook over to this

1:07:17address. So hook to outline finds

1:07:19something that's hook. Maybe this one

1:07:20here. Now what I'm going to do is I'm

1:07:22going to go over here, listen for a test

1:07:23event. Then I'm going to change this to

1:07:25outline. And I don't remember if it's a

1:07:27get or a post request. So we might have

1:07:29to wait and try two or three. This

1:07:32unfortunately does not respond to the

1:07:33methods that are not the same. methods

1:07:36are, you know, kind of a deeper story,

1:07:37but basically there are variety of ways

1:07:39you can query a website or a web

1:07:41service. HTTP, most commonly, you'll do

1:07:43either a get or a post. Um, in the

1:07:47specific instance of ClickUp, it looks

1:07:48to me like they're probably using post.

1:07:49So, I'll go HTTP method. I'll go

1:07:51post. Now, I'll listen for this test

1:07:53event. I'll change this back to

1:07:56hook. Then, I'm going to change this

1:07:58back to outline. And now that it's

1:08:00outline, um, we're waiting for this this

1:08:02post request. Let's see if we we are

1:08:04receiving it from ClickUp. Okay, great.

1:08:06Looks like we received it. Now, what did

1:08:09we receive in reality? Well, excuse me.

1:08:11I'm like really close to sneezing, but

1:08:13I've decided not to, so I won't. Uh,

1:08:16before we updated the query field, now

1:08:18we're updating the body field. We see

1:08:19it's built into ClickUp. What we did,

1:08:21ClickUp has a number of default things

1:08:23that they send over when you do this web

1:08:24hook integration. One of them is the ID

1:08:26of the record. Now there's the trigger

1:08:28ID ID the trigger payload ID the name of

1:08:31the thing which was three chat GBT

1:08:33prompt engineering hacks you need to

1:08:34start using next some hooks some order

1:08:38in text some information about the

1:08:39person that created it the point that

1:08:41I'm making is we just created our own

1:08:42integration with ClickUp and it took me

1:08:44like 30 seconds realistically after I

1:08:45got off the HTTP method hump so in

1:08:48practice sometimes APIs like ClickUp I'm

1:08:51sure they have API documentation

1:08:52somewhere but sometimes they just don't

1:08:53tell you right at the point of creating

1:08:55the web hook whether it's going to be a

1:08:56get or a post request. So if your get

1:08:58request doesn't come in, just change the

1:08:59HTTP method to post and then rerun it

1:09:02with the post uh example and then one of

1:09:04those will work, which is pretty cool.

1:09:06Okay, so that's how you do it as a test

1:09:08URL. Ultimately though, we don't just

1:09:10care about having this workflow be sort

1:09:12of a test. We want it to be live. And so

1:09:13when your workflow moves to production,

1:09:16aka you publish it and you make it live

1:09:18to actually interact with the internet,

1:09:19you're going to want to go over to this

1:09:20production URL and then copy this URL

1:09:22and update all of your web hooks to send

1:09:24here. This will enable you to activate

1:09:26this. If it's in test mode, you actually

1:09:27won't be able to activate this. Um, just

1:09:29I guess for safety or security purposes.

1:09:32It just makes the transition to publish

1:09:33it unfortunately involve an additional

1:09:35step, but it also makes your workflows a

1:09:37little bit more secure. So, that's how

1:09:39you do web hooks. The last thing I'll

1:09:41mention are these OpenAI and AI nodes.

1:09:43Now, I'm going to have many, many videos

1:09:45after this one be all about AI agents

1:09:47since that's obviously the big thing

1:09:48that's blowing up right now. What I'm

1:09:50going to do here is I'm just going to

1:09:52cover them super super briefly. I'm not

1:09:54even going to like run anything, but I

1:09:55just wanted to show you guys that NAN is

1:09:57very AI native. And so whereas I've been

1:10:00doing some very basic um OpenAI calls

1:10:02with this OpenAI module, there's a

1:10:04variety of things you could do. You

1:10:05could create an AI agent which generates

1:10:07an action plan and executes it. Uses

1:10:09external tools. You can have OpenAI

1:10:11message an assistant or GPT. This is

1:10:13what we've been using. Some basic LLM

1:10:15chains and a bunch of like specific

1:10:17tools that are used to do things like

1:10:19categorize information, summarize

1:10:21information, and so on and so forth. AI

1:10:23agent. Just to give you guys a very

1:10:25brief example, probably is one of the

1:10:27most intimidating looking modules or

1:10:28nodes, but it's actually one of the

1:10:30simplest in practice. When you create an

1:10:32AI agent, it'll automatically open up

1:10:34this when chat message receive node on

1:10:36the side. And then you'll see that down

1:10:37at the bottom of my screen, there's an

1:10:38additional button that allows me to chat

1:10:40with my with my model. But in order to

1:10:42make this work, what we need to do is we

1:10:44need to hook this up. Notice how there's

1:10:45this little um warning here. What we

1:10:47need to do is we need to go down to chat

1:10:49model. We actually need to select the AI

1:10:51module or AI node that we want the AI

1:10:54service I should say that we want to

1:10:55use. For most intents and purposes, I

1:10:58use OpenAI. This is just the best to me.

1:11:00But you could use O Lama, Mistral,

1:11:02Google, Gemini, Anthropic. Feel free to

1:11:04play around with this for whatever your

1:11:05use case is or whatever your data

1:11:06privacy security requirements are. So,

1:11:09I'm going go down to my OpenAI chat

1:11:10model. The model I'm going to be using

1:11:11for this is going to be GBD40. I just

1:11:13find it has better answers. And then now

1:11:16you'll see that the warning sign is

1:11:17gone. Now we have an additional node I

1:11:19can drag and drop here. And now I can go

1:11:21to chat and I could say, "Hey, how are

1:11:22you doing?" We've essentially opened up

1:11:24our own chat window. Hey, I'm just a

1:11:26computer program, but I don't have

1:11:28feelings, but I'm here and ready to

1:11:29assist you. Thanks, Chat GPT. And on the

1:11:31right hand side, probably the most

1:11:32valuable part about this is you could

1:11:33see a log of what is happening and how

1:11:35many nodes were called in order to get

1:11:38you the result. This is important

1:11:39because the whole point of AI agents is

1:11:41their ability to call other tools to do

1:11:43things for you. So this says system,

1:11:45you're a helpful assistant. human, hey,

1:11:47how are you doing? This is just their

1:11:49prompt setup. So, the input to OpenAI

1:11:52was this right here. And because we just

1:11:54asked how it's doing and so on and so

1:11:56forth, um, you know, it didn't it didn't

1:11:57really do anything special. There were

1:11:58no additional tools. This is the same

1:12:00thing as you just sending a message to

1:12:01chat GBT essentially. Now, in order to

1:12:04make an AI agent like really work,

1:12:06you're going to want to add two things.

1:12:07The first thing you're going to want to

1:12:08add is you're going to want some sort of

1:12:09memory. If an AI agent doesn't have

1:12:11memory, then basically if I go back here

1:12:14to chat and then say, "What did I say in

1:12:17my last

1:12:18message, it will have no context or no

1:12:21idea. I don't have any access to past

1:12:23messages or personal data. Each session

1:12:25is independent." Basically, this is like

1:12:26a one it's like it's like a send a

1:12:29question, receive an answer sort of

1:12:30window. But we don't want that. We want

1:12:32this to actually have access to our chat

1:12:33history. We want it to see what we've

1:12:34been talking about over the course of

1:12:35the last like 20 or 30 minutes and be

1:12:37able to reference those. So, there are a

1:12:39variety of ways to do this. Basically,

1:12:40you need to implement some sort of

1:12:41database. I'll show you guys how to

1:12:42implement some more complex databases in

1:12:44the future, but the simplest one, the

1:12:45one that Naden provides right out the

1:12:46gate, the one that most people on

1:12:47YouTube are going to be talking about is

1:12:49this window buffer memory. Window buffer

1:12:51memory basically just allows you to

1:12:53store it here inside of the test window,

1:12:55which is the easiest to do. And the

1:12:57default is five messages. So, u

1:13:00basically every time you send it a

1:13:01message, it will send up to five pass

1:13:03messages. But just for the purpose of

1:13:05this discussion, I'm going to go 10. So

1:13:06now what I'm going to do is I'll go to

1:13:08chat and then I'll say hey how are you

1:13:11doing and I'm going to say what did I

1:13:14ask you in my previous message and

1:13:16you'll see that it's asking how it's

1:13:18doing right so we're actually now

1:13:20accessing the previous message using

1:13:23this buffer memory and on the right hand

1:13:25side you'll see here that the log has

1:13:27gotten a little bit more intense too so

1:13:28basically we called the AI agent up at

1:13:30the top the next thing that happened was

1:13:32we went through the buffer memory we fed

1:13:35this in and basically ally added this to

1:13:37some big long stack of message history.

1:13:40Then we fed that in plus the previous

1:13:42message. Then we said, "Hey, what did I

1:13:45ask you in my previous message?" This is

1:13:46the input that is currently being fed

1:13:48into the model on that second call. And

1:13:51then we went down into buffer memory. We

1:13:53saved all of this again. And then uh we

1:13:55kind of came up and then and sent the

1:13:56answer. How about

1:13:59now? Cool. So now we're three levels

1:14:02deep. And as you can see, this is just a

1:14:03quick and easy way to load the memory.

1:14:05make sure that we're always having some

1:14:06sort of topical contextual conversation,

1:14:08which is pretty cool. Now, the real

1:14:10juice in AI agents, the reason why

1:14:12they've gained so much popularity is

1:14:14just because of this tool section here

1:14:16where you can essentially call an action

1:14:18just like we were doing before

1:14:19procedurally, but you can call it using

1:14:21AI and you can automatically format and

1:14:23get information from different tools.

1:14:25So, there are variety of tools that are

1:14:26sort of set up for you. Air table, base

1:14:29row, calculator, Gmail, Google calendar,

1:14:31Google Docs, Google Drive, all this

1:14:32stuff.

1:14:34What I'm going to do in this example,

1:14:35just because I don't want to spend all

1:14:36day on it, um, before I do my more

1:14:38detailed AI agent tutorials, is I'm just

1:14:40going to select Google Calendar. I'm

1:14:42going to create a new credential, sign

1:14:43in with my Gmail

1:14:46account, and then I'll go over

1:14:48here. I'll close this window. Now that

1:14:52I've uh I've connected my Google

1:14:53calendar agent, what I'm going to do is

1:14:55I'm going to select my specific

1:14:56calendar, which is Nick

1:14:58leftclick.ai. And then now that it's

1:15:00connected to my AI agent, this is the

1:15:02create event. I don't want to create. I

1:15:04just want to get so get all of my events

1:15:06basically from my calendar. Okay. And

1:15:07then in addition to that, what we have

1:15:08to do is we have to use this dollar sign

1:15:10from AI feature here. And we just have

1:15:12to paste that into the expression field.

1:15:14Um what this does is this just tells AI,

1:15:16hey, I want you to provide uh you know

1:15:18your own details for the value of a

1:15:21field. In our case, I wanted to feed in

1:15:23some options that say, hey, you know, I

1:15:25want you to grab data that is after this

1:15:27date but before this date. So, if I'm

1:15:29asking

1:15:30AI like, "Hey, what's going on? Can you

1:15:34tell me what I'm doing tomorrow?" And

1:15:36I'll just say, "Jan 28, 2025." The whole

1:15:38idea here is AI now has access to my

1:15:40calendar. It also has the ability to

1:15:42call that API. Then it can actually go

1:15:44and retrieve specific events from my

1:15:46calendar and then return them here. So,

1:15:48as you can see, I have this call uh me

1:15:50and my my buddy Zach, and then it just

1:15:51has all of this information. Okay,

1:15:53great. How are you more generally? you

1:15:55know, I can also just chat with it like

1:15:57I'm chatting to chat GBT or something

1:15:59like that. So, I don't always have to

1:16:00use like the tool that I'm I'm calling.

1:16:02And the idea is you basically stack on

1:16:04three, four, 5, 10, 15, 20 of these

1:16:06tools. Although, I find in practice when

1:16:08you get um past maybe six or seven,

1:16:10instead of calling a tool, what you want

1:16:11to do is you want to call another agent

1:16:12which then decides to call a tool. It's

1:16:14basically like a big almost like a

1:16:15search tree or something. But anyway,

1:16:17that's more or less the AI side of

1:16:19things. The last thing I'll mention here

1:16:21is uh this OpenAI note doesn't just have

1:16:23the message a model text action like

1:16:25we've been doing before. There's a

1:16:26variety of other things you could do.

1:16:27You can create an assistant, delete an

1:16:29assistant, list assistance, message

1:16:30assistance, update assistance, analyze

1:16:32images, you can generate images,

1:16:34generate audio. Like you have a ton that

1:16:35you could do here, which is pretty

1:16:36sweet. I go down to generate audio and

1:16:39I'll say um I don't know, Nick is

1:16:42awesome and very pretty and I generate

1:16:44it using the Nova voice. Click this test

1:16:47step. Not only can I generate text and

1:16:50stuff like that, but I can also have

1:16:51this generate me an audio output I can

1:16:53then listen to. Variety of cool things

1:16:55you could do with this. Nick is awesome

1:16:57and very pretty. You're damn right I am.

1:17:00Um variety of cool things you can do

1:17:01with this, but definitely don't sleep on

1:17:03the AI nodes. Uh you know, don't just

1:17:05like stick to the one that I've shown

1:17:06you guys so far. Okay, great. So, those

1:17:08are the foundational ones. In practice

1:17:09with NAN, you're probably going to be

1:17:11using these quite often. What I'll do

1:17:12next is talk about some nodes that

1:17:13modify flow. And here a bunch more that

1:17:15like you're going to want to read their

1:17:17docs and add them to your toolkit

1:17:18because this is like an everyday sort of

1:17:19thing. The first is an if. The second is

1:17:22filter. The third is merge. And the

1:17:23fourth is split into batches. So let me

1:17:26show you guys a very quick and simple

1:17:28example of the if. If I go

1:17:30here and then I add my own trigger. Uh

1:17:33and I just want to trigger manually. Oh,

1:17:35sorry. It looks like I already have one

1:17:36somewhere in here. Right. So let's just

1:17:38repurpose this um manual trigger for an

1:17:41example workflow that I'm going to build

1:17:43down below.

1:17:44this example trigger. I'm going to click

1:17:46it. And then what we're going to do is

1:17:48we're going to use the edit

1:17:50fields. We're going to go down to JSON.

1:17:53This is just a handy dandy tool that

1:17:54allows you to set your own inputs and

1:17:56outputs. So I can now set my own um

1:17:58output. And I could say first name Nick,

1:18:02last name uh Sarif. And if I if I test

1:18:05this, if I test my whole workflow,

1:18:06you'll see this broke because um Sarif

1:18:08was not in quotes there. Got to make

1:18:10sure that all of your strings are in

1:18:11quotes. If I check out the JSON, you see

1:18:13the output of this module is now first

1:18:14name, Nick, last name, survive. Okay,

1:18:16I'm just going to pin this. Now, let's

1:18:18say I want to do something else, you

1:18:19know, if

1:18:20the if the input is Nick, I want to do

1:18:24something really cool. I want to provide

1:18:27prize. I don't know, um,

1:18:30$100. If the input is Nick, I want to go

1:18:33through my my sequence and then I want

1:18:34to generate another variable called

1:18:36prize, and I want to I want to have it

1:18:37be $100. Okay. But I only want to do

1:18:40that if the input is Nick. If the output

1:18:42is something else, then I want to have

1:18:46my prize be just $5. So Nick gets all

1:18:50the prizes here. He's very greedy. Okay.

1:18:52So how do you actually implement this

1:18:53sort of logic? Well, the simplest way is

1:18:54if I click this plus button and I just

1:18:56type if, you'll see I'll have this node

1:18:58pop up that says if wrote items to

1:19:00different branches, true or false. So

1:19:01I'm going to add that in there. And what

1:19:03I'm going to say is if first name, which

1:19:05by the way, we could still just drag if

1:19:06we wanted to. You can say if first name

1:19:08is equal to Nick, then proceed through

1:19:11the true node, which is up here. And if

1:19:15not, we're going to want to proceed

1:19:16through this false node. Isn't that

1:19:19cool? So now we basically have two

1:19:21things that are occurring. Okay, prize

1:19:24up here was 100. Prize down here was

1:19:27five. I'm going to click test workflow.

1:19:29This is now going to run. And I just

1:19:31want you guys to see what's happening. I

1:19:32clicked test workflow. We then edited

1:19:35our fields. We added Nick as the first

1:19:37name. We then went to the if and then as

1:19:39we saw here, first name was equal to

1:19:42Nick, meaning oops, if I double click

1:19:44this again, the output is now only going

1:19:47down the true branch with one

1:19:50item. And then the uh upper branch was

1:19:53illuminated. It's green. And then that's

1:19:54how we get to edit fields with the prize

1:19:56equal to $100. Now, if I change this,

1:19:58instead of Nick, if it's like Sally or

1:20:01something, and if we run this again,

1:20:04what you'll see is the

1:20:06data is flowing through here cuz I

1:20:08didn't unpin it. So, let's unpin it. If

1:20:10we if we test this now, uh what you'll

1:20:12see is the data didn't go through the

1:20:13top uh route anymore. Went through the

1:20:15bottom route. Okay, great. So, now with

1:20:16this example here, why don't we just

1:20:18pretend like we're emailing somebody.

1:20:20So, I'll go draft an email.

1:20:23create my thing here and we'll say

1:20:25congrats you won good expression. Then

1:20:29what I want is dollar sign

1:20:31Jason.prise right over here. So like

1:20:34congrats you won $5. How cool is that? C

1:20:38title

1:20:40loser. Uh okay, great. And then I'm

1:20:42going to go here and I'm going to create

1:20:43a

1:20:44draft. Voila. I have it. Now if I go

1:20:46back here to my SOS media queries page,

1:20:49see it says, "Congrats, you won $5."

1:20:51There's nobody to because we didn't set

1:20:52the email to. But pretty neat, huh? We

1:20:55gave the same thing with this $100

1:20:59field. Go over here. You'll see that I

1:21:01just copied like all of the same logic.

1:21:03It says, "Congrats, you won Jason.

1:21:05Prize." It's grayed out right now

1:21:06because there's no data coming in. But

1:21:08you'll see that it'll work if I change

1:21:09the um input back to Nick. Then if I

1:21:12test this, I'll see it'll follow the top

1:21:14route and then also send me an email.

1:21:16Okay, great. So now let's look to use

1:21:17the filter node. Um, what I have here is

1:21:19I have my little first name Nick here.

1:21:22Um, what I'd like to do instead is I'd

1:21:23like to just change this a little bit.

1:21:25So, instead we'll create an array of

1:21:28names. And inside of this, we'll have

1:21:30Sally, John, and then Nick. Okay. And so

1:21:34now, if we test this, we see three

1:21:36entries in an array called names. We

1:21:37have our top level array, which contains

1:21:39an array of objects, and we have our

1:21:41object. And inside of that, we have key

1:21:43whose value equals a list of other

1:21:45objects and array of other objects. I

1:21:46know the terminology can be kind of a

1:21:48lot and unfortunately there are many

1:21:50ways to refer to the same thing. So if

1:21:52something doesn't make sense just bear

1:21:53with me here and we'll in a moment.

1:21:54Okay, great. So let's say what we want

1:21:56to do if this names array includes Nick

1:22:00then um I want to continue with the

1:22:02flow. So I'll go array. Then what I want

1:22:05is I want to see if this array of of

1:22:07names feed that in here contains and I

1:22:11just want to say Nick. I'm going to test

1:22:13this out.

1:22:15What you'll see is that we've kept it

1:22:16because it does in fact contain Nick,

1:22:18which is pretty cool. If instead we

1:22:20wanted to see if it contains Peter, we

1:22:22test this. You'll notice that we are now

1:22:24following the discarded route. Okay,

1:22:26there's kept and then there's discarded.

1:22:28The thing is um it just goes down the

1:22:29same flow. Whereas if sort of split into

1:22:32two, there's a true and a false route.

1:22:33Uh this one actually just like continues

1:22:34and proceeds down the same flow. If

1:22:36something matches the filter, it will

1:22:38continue. Something doesn't match or the

1:22:40fil match the filter, then it won't. So

1:22:42what we could do if I just paste in this

1:22:44Gmail node. So we could basically build

1:22:47uh a very similar flow but what we could

1:22:50do is if it is kept then we could send

1:22:52the prize of $100 instead. We could

1:22:54hardcode that $100 in. So I'm going to

1:22:56do is I'm going to check to see if it's

1:22:57kept. Okay. So right now name contains

1:22:59Peter. It's probably not going to be

1:23:00kept, right? So test workflow. It's

1:23:01going to stop right here. Does not

1:23:03proceed any further. If instead I change

1:23:05the filter so that it contains Nick,

1:23:06then we test it. we see is we're going

1:23:08to move on and we're actually going to

1:23:10like proceed with the rest of our flow.

1:23:12So, this filter here just allows us to

1:23:13kind of stop if it doesn't match our

1:23:15condition or or continue. And you can

1:23:17add as many conditions as you want. You

1:23:18can go and or or um you can add whatever

1:23:20sort of logic you'd like. I used array

1:23:23logic here too, but there's also a lot.

1:23:24There's like string. You can check to

1:23:26see if something exists or matches array

1:23:27X, number, date and time, boolean,

1:23:30array, and then there's also a bunch of

1:23:31object ones as well. So, that's filter.

1:23:34Pretty straightforward, I would say. The

1:23:36last thing is two more. There's one

1:23:38called merge and then there's split into

1:23:40batches. What I'm going to do here is

1:23:42I'm going to have two routes or two

1:23:44outputs of a module and then I'm going

1:23:46to combine them back into one. And I'll

1:23:47show you guys what I mean with this.

1:23:49Remember earlier how we had some HTTP

1:23:51requests. What I'm going to do is I'll

1:23:53go first name

1:23:57Sally and I'm going to

1:24:00have let's just say second name.

1:24:03Actually, let's

1:24:05go person one. Go person two. Just for

1:24:09the purpose of this example, we're going

1:24:10to have two people. Person one, Sally.

1:24:12Person two is Nick. Okay. Next, what

1:24:14we're going to do is we're going to add

1:24:15an AI node. You know, down to open AI,

1:24:18and I'm just going to message a model.

1:24:20Then here, I'm going to say write a

1:24:23detailed fun story about what I'm going

1:24:26to do is I'll go I don't know, person

1:24:29one. So, JSON.person one. This example

1:24:32is sort of silly if I'm honest because

1:24:34we could just hardcode the names in

1:24:36there. But I just wanted to do this to

1:24:37show you guys how this logic of the

1:24:39merge node would work. I'm going to

1:24:41select

1:24:43GPT40. And then I'm, you know, because