顧哥知識系統

Step-by-Step Tutorial: Build A.I. Agents with n8n (NO CODE!!)

閱讀時間約 30 分鐘

Step-by-Step Tutorial: Build A.I. Agents with n8n (NO CODE!!)

Transcript

[00:00] in this video we’re going to build a
[00:01] really cool AI agent that will have the
[00:04] ability to pull data from the Internet
[00:06] or call in a separate workflow that will
[00:09] have an A API attached to it so we’re
[00:12] going to build all of this step by step
[00:14] and I’m going to show you how powerful
[00:16] nadn is when it comes to uh building
[00:19] customized AI agents and how you can
[00:21] connect several different tools and
[00:23] obviously you can attach a large
[00:25] language model like um open AI in this
[00:27] case or anthropic or others as well I’m
[00:30] going to explain what nadn is and why
[00:32] this is such a cool tool to use when it
[00:35] comes to customize workflows and
[00:37] specifically building customized AI
[00:39] agents because it gives you the ability
[00:41] to connect different tools and therefore
[00:43] make the AI agent extremely powerful in
[00:46] in our scenario in this example that I’m
[00:48] going to walk through this AI agent will
[00:50] independently make a decision based on
[00:52] the question that you’re asking it
[00:54] either to call the Wikipedia tool to
[00:56] grab data from the Internet or call this
[00:59] separate nadn workflow that we’re going
[01:01] to build that will be connected to a
[01:03] third-party API app so this will be a
[01:06] really comprehensive tutorial on how to
[01:09] utilize n’s different tools and
[01:12] different customizability options to be
[01:14] able to build a really powerful AI agent
[01:16] all right let’s Jump Right In All right
[01:18] so a quick introduction for nadn um if
[01:20] you’re not familiar with nadn nadn is a
[01:23] workflow automation tool but it’s
[01:25] extremely powerful in my opinion it’s
[01:26] way better than make.com voice FL or all
[01:29] these other automation tools that are
[01:31] out there in the market and specifically
[01:33] when it comes to building AI agents so
[01:35] the reason why I really like an and
[01:37] again I’ve been building these
[01:38] Automation workflows and agents for a
[01:41] while now and the reason why I landed on
[01:43] nadn and why I think that this is such a
[01:45] good tool is because it gives you really
[01:47] a lot of option when it comes to
[01:49] customizing and being able to provide
[01:51] different tools and attach different
[01:53] apps to be able to really customize your
[01:55] AI agent another great Advantage for
[01:57] nadn is that you can actually self-host
[01:59] this on your local computer again this
[02:01] is going to be more of a introductory
[02:03] tool I don’t want to go too much deep
[02:04] into that but on the next videos I’ll be
[02:06] able to build a more complex automation
[02:08] tool or more complex AI agents where
[02:11] then I’ll be able to host or self-host
[02:13] nadn on my own local computer where then
[02:16] you basically will have the ability to
[02:18] own all of your data um and therefore be
[02:21] at ease when it comes to customizing and
[02:22] uploading your own data all right so
[02:24] with that being said um so you can
[02:26] create an um free n in account um and
[02:30] again you can do a cloud account so
[02:31] that’s going to be the free option you
[02:33] don’t have to worry about hosting self-
[02:34] hosting it for now we’re just going to
[02:36] go ahead and use their cloud account but
[02:38] you can as soon as you log into your
[02:39] account you’ll come into um your
[02:41] workflow and this is basically the
[02:43] starting point so I’m just going to get
[02:45] rid of this um so if you’re familiar
[02:46] with a tool like make.com it’s kind of
[02:48] similar to that but it’s again like I
[02:50] said there’s more custom customizability
[02:52] option here all right so first the first
[02:54] node we’re going to add is uh a trigger
[02:57] and again a trigger could be something
[02:59] whether uh it’s a manual trigger in this
[03:01] case obviously I added a manual Tri Tri
[03:03] trigger but it could be a chat it could
[03:06] be an email so you can have different uh
[03:08] trigger options available and again in
[03:09] the future videos I’ll create other AI
[03:12] agents that will have a trigger of an
[03:14] email or a chat or a slack message uh
[03:17] but for this one we’re just going to do
[03:18] a u manual when clicking test the
[03:22] workflow so this basically as manual as
[03:23] it gets all right so let’s go ahead and
[03:25] click on this add node button right here
[03:28] um so what we’re going to do is is
[03:30] actually go to Advanced AI again we’ll
[03:32] come back in other videos and take a
[03:33] look at these other options that you can
[03:35] add obviously there’s a lot of um
[03:37] different options when it comes to
[03:39] adding nodes but for this one we’re
[03:41] going to click on Advance Ai and when
[03:43] you click on Advance AI this gives you
[03:45] all of the different options you have
[03:46] when it comes to the AI tools that you
[03:48] can attach but we’re going to click on
[03:50] AI agent and as soon as you click on AI
[03:53] agent it’s going to pull up um the
[03:55] setting and the parameter we don’t have
[03:56] to worry about the settings for now but
[03:58] parameters this is where will be able to
[04:01] um attach actually different tools to
[04:03] this AI agent so for now let’s just
[04:05] leave this as it is we need to attach a
[04:08] uh language model to this so I’m going
[04:10] to get rid of this this node because
[04:12] automatically pulls that up so the first
[04:14] thing we need to do is add a chat model
[04:17] so if you just click on this plus button
[04:19] or you just drag it so this is all of
[04:21] the language models that are available
[04:23] for NN and they have this integration so
[04:25] they have obviously the the most popular
[04:26] one they have the entropic you know your
[04:29] grock and um Llama Or Lama the openi
[04:33] chat model obviously um I I have an
[04:36] account with open AI so I’m going to
[04:37] choose this but it depends on your
[04:39] choice you can definitely um uh choose
[04:42] whatever you want all right so once you
[04:43] do that then it’s going to ask you for
[04:46] your credential so you can connect to
[04:47] your API account again you have to have
[04:50] an open AI API account your keys that
[04:53] you need to grab from there so that way
[04:55] you can add to your credentials so I’ve
[04:56] already have my account here but if you
[04:58] don’t you just click on create new
[05:00] credentials and again you just have to
[05:02] go to your API account and be able to
[05:04] grab your API keys or you can create a
[05:06] new API key the organization ID is
[05:08] optional but you have to have your API
[05:10] key all right so once you do that you’re
[05:12] going to come back and this is going to
[05:13] get um your account is going to get
[05:15] connected here so the next obviously you
[05:17] got to choose your model again I’m going
[05:19] to use gp4 and that’s pretty much it so
[05:22] now you have connected your uh open AI
[05:25] chat model to this and as you can see
[05:27] this this attach is complete so another
[05:30] thing I’m going to add here is a memory
[05:32] if you just click on this so you know
[05:34] not to get into too much details into
[05:37] this but this just basically gives NAD
[05:40] the the ability to store your um data
[05:44] into your windows memory buffer so that
[05:48] way it can refer to your previous chats
[05:50] when you build this AI agent so we’re
[05:51] just going to click on the Windows
[05:53] buffer again you don’t have to do much
[05:54] you don’t have to add any credentials or
[05:56] anything like that you just got to
[05:57] attach it all right so the memory is
[05:58] done so this is what makes nadn really
[06:01] powerful the ability to add these
[06:03] different tools so they have a lot of
[06:05] tools available here uh so the first
[06:07] thing we going to add is a calculator
[06:09] and again a calculator just gives the
[06:11] this AI chat model the ability to uh run
[06:15] numbers so if you’re asking it a a
[06:16] question um that requires mathematical
[06:19] operations then it will have uh the
[06:22] ability to do that with this calculator
[06:24] I mean again we don’t have to worry
[06:25] about it too much all right so for the
[06:26] tools I’m just going to keep it separate
[06:28] and you can attach multiple tools here
[06:30] all right so another tool we’re going to
[06:31] add is the Wikipedia tool again just if
[06:34] you just click on this PL plus button
[06:36] The Tool uh will open up again uh and as
[06:38] you can see at the bottom right here it
[06:40] says Wikipedia so what this tool does it
[06:43] gives your agent the option uh or the
[06:46] ability to to have a conversation with
[06:48] you and have access to the knowledge
[06:51] resource of Wikipedia okay so now that
[06:53] our AI agent has the um knowledge base
[06:57] from Wikipedia it means that we can cat
[06:59] with it right so right now as you can
[07:01] see in the bottom here it just says test
[07:03] flow um because right now this is a
[07:06] manual trigger but let’s go ahead and
[07:08] get rid of this and we want to be able
[07:09] to add that chat trigger here so I’m
[07:11] going to go ahead and click on this um
[07:15] add first step and again you can click
[07:16] on that top right corner there as well
[07:19] and as you can see on the uh on the list
[07:20] here you have that trigger manually
[07:22] option on an app event on schedule on a
[07:24] web H hope call uh but as in the bottom
[07:27] you can see here it says on chat message
[07:28] so we’re going to click on on that all
[07:30] right so
[07:35] oops so let’s grab this over here and
[07:38] bring it here all right and then let’s
[07:40] go ahead and connect this there you go
[07:42] so you just got to do is you just got to
[07:44] drag that and connect it to and you can
[07:46] always delete it from here this
[07:47] connection um all right so now that
[07:49] that’s connected um so now again you
[07:53] don’t have to do anything here this is
[07:54] just a trigger and as you can see at the
[07:57] bottom here little chat box appears
[07:59] which means now you have the ability to
[08:01] chat with this AI tool um so let’s go
[08:03] ahead and actually do that first of all
[08:04] another thing is I always have the habit
[08:07] of saving so I’m just going to click on
[08:08] Save
[08:09] there that way you don’t use lose your
[08:12] progress all right so now let’s go ahead
[08:13] and test this out so I’m going to come
[08:15] here and chat with this so I’m going to
[08:17] say what is the capital of Spain for
[08:21] example
[08:22] right so it should be able to have
[08:25] access to that data via that Wikipedia
[08:27] tool that we that we that we gave it and
[08:30] there you go the capital of Spain is
[08:31] Madrid and on the bottom you can see the
[08:33] log for this message it means that it uh
[08:35] the AI agent use the windows buffery
[08:38] member again this is just to save so
[08:40] that way you can refer to it um um later
[08:44] on uh and it used the open AI chat model
[08:47] and it used the Wikipedia tool as you
[08:49] can see right here to search for that
[08:51] and if we get out of this you can see
[08:52] now that the AI agent used the following
[08:55] tools by looking at this um with the
[08:57] little green markings here all right all
[08:59] right cool now we’re going to add
[09:01] another workflow to this and what you’re
[09:03] going to do is click on this plus button
[09:06] again right here it says called NN
[09:08] workflow tools what this means is that
[09:09] this gives you the ability to call other
[09:13] workflows that you can create in NN and
[09:16] attach it to that AI agent and the way
[09:18] you do that is right here on the
[09:19] workflow ID you can attach the ID of a
[09:22] separate workflow that you’ve created so
[09:24] therefore um this AI agent will have
[09:26] access to that so let’s go ahead and
[09:28] actually do that so let’s go ahead and
[09:30] create a new workflow that grabs weather
[09:33] data using a thirdparty API okay so
[09:37] let’s back out of this Fel and as you
[09:39] can see right now there’s this little
[09:40] error message just because there is no
[09:42] workflow connected to this so in order
[09:44] to create a new workflow again you can
[09:46] just go back to your homepage and just
[09:48] add another uh workflow here so I’m just
[09:50] going to go ahead and do that all right
[09:52] so now I’m here with my new workflow so
[09:55] what we’re going to do is leave this um
[09:57] the test workflow as manual trigger and
[09:59] click on the plus button here and here
[10:02] we’re going to grab this data
[10:03] transformation so when you click this
[10:05] gives you the option to manipulate uh
[10:07] filter or convert different data that’s
[10:09] coming in so we’re going to click on
[10:10] this and we’re going to select this um
[10:13] edit field again this is kind of the
[10:16] popular ones but you can see at the
[10:17] bottom they have all these different uh
[10:19] data transformation tools that you can
[10:21] attach and again the following videos
[10:22] I’ll kind of go through some of these
[10:24] because they’re very powerful but for
[10:25] now we just click on this edit field all
[10:27] right so we’re going to go ahead and
[10:28] change the name here so I’m going to
[10:30] change this to query if I can type
[10:36] oops all right so click on rename we’re
[10:39] going to leave the mode as manual
[10:42] mapping um so Json is if you’re you know
[10:45] familiar with code you can actually add
[10:47] the mode via Json as well but for now
[10:49] we’re just going to leave this as manual
[10:51] mapping so for the fields to set so this
[10:53] is where you can um grab uh a field from
[10:57] a previous input and in this case I
[10:59] obiously ours is a manual trigger so if
[11:01] you just execute the previous node it
[11:03] just says when clicking test workflow
[11:06] because we don’t have anything added to
[11:07] this but obviously if you have a
[11:08] previous node that’s attached to it and
[11:10] is doing something you’ll be able to
[11:12] have access to that as well all right so
[11:14] now what we need to do is just manually
[11:17] add a field okay so the name you can
[11:19] name this again same thing I’m going to
[11:20] name it Cory you can leave it as string
[11:23] because this is just going to be um a
[11:25] value that we’re going to provide to our
[11:27] API endpoint in the next note that we’re
[11:29] going to add all right so for the value
[11:31] now we’re going to add um a name of the
[11:33] city again this is just so we can test
[11:35] our workflow so for now I’m just going
[11:37] to say San
[11:38] Francisco that’s my city so I’m just
[11:40] going to put that here all right and
[11:42] that’s pretty much it so now let’s go
[11:44] ahead and test this step and by testing
[11:46] this step you you want to make sure that
[11:49] uh the output is what you’re expecting
[11:51] and in this case we want the output to
[11:52] be just be San Francisco and as you can
[11:54] see right here the quy it’s the name of
[11:56] the field and it’s outputting San
[11:58] Francisco based on what be put in the
[11:59] bottom here all right so let’s back out
[12:01] of that and as you can see it’s
[12:02] successful uh if there was any kind of
[12:05] error this green box won’t appear it
[12:07] it’s just going to show the error there
[12:09] okay so now we need to add a node that
[12:12] will have access to weather data and
[12:15] we’re going to do this by grabbing the
[12:18] weather API from a third party so the
[12:20] way we do that is just again same thing
[12:22] add a node I’m going to search weather
[12:24] here and I’m going to use this open
[12:26] weather map uh tool that they have and
[12:29] again this is uh from a third party so
[12:31] it just you gives you the option as far
[12:33] as actions you can return current data
[12:34] or return weather data for the next 5
[12:36] days so let’s just click on the return
[12:39] current weather data because we want to
[12:41] be able to have you know the current
[12:43] data from our Cy which is San Francisco
[12:46] all right so once you click on that this
[12:47] will open up your credentials again to
[12:49] connect with so it’s the same kind of
[12:50] concept you have to go to your open
[12:53] weather map um account on their website
[12:56] and if you click on docs here this also
[12:58] opens up um kind of a guide on how to
[13:01] grab your credentials so go ahead and do
[13:02] that once you do that again same thing
[13:04] you’re just going to come here and uh
[13:06] add your access token which is your API
[13:08] key then at the bottom here you have
[13:10] your operation and you’re just going to
[13:12] say current weather 5day for uh forecast
[13:15] I’m just going to leave it with current
[13:16] weather uh the format you can choose as
[13:18] metric or imperial you know obviously
[13:20] Imperial is the Fahrenheit metric
[13:22] Celsius I’m just going to leave it as it
[13:24] is for now um location setting so this
[13:27] is going to be the city name that’s
[13:28] going to going to be coming in so the
[13:30] city this is where um we’ll be able to
[13:34] grab the city and in this particular
[13:36] this field could be or this parameter
[13:38] could be coming in from our previous
[13:40] node and in our case in our previous
[13:42] node we added San Francisco as our query
[13:46] so that’s exactly what we’re going to do
[13:47] and then as you can see on the left hand
[13:48] side on input you have a query that’s
[13:51] coming in and it shows San Frisco right
[13:54] so there’s the schema there’s a table
[13:56] View and there’s a Json view right so if
[13:58] you’re you know amiliar with code you
[13:59] can uh work with Json and that also is
[14:02] very helpful um because it gives you
[14:05] exactly the code version of it but you
[14:06] can just stay in schema because this if
[14:09] you’re not familiar with code or you’re
[14:10] not good at coding you just grab
[14:12] basically uh this schema from here so
[14:15] what I’m going to do is just basically
[14:16] grab this and bring it over here to the
[14:18] city and as you can see this show this
[14:21] uh converts it into um a Json field
[14:24] which again it’s a JavaScript field so
[14:26] all right so that that’s pretty much
[14:28] we’re done with this point we’re just
[14:29] going to leave the language is English
[14:31] so let’s go ahead and test this step out
[14:32] and again I’m going to click on test
[14:35] step and there you go so because my um
[14:39] API is connected one thing I want to
[14:41] point out that when you create your open
[14:43] weather map API account it does take
[14:45] about 5 to 6 minutes for your API to key
[14:48] to get operation so just once you create
[14:50] that account just be on the lookout
[14:51] there because if you click on this it
[14:53] just going to give you an error so once
[14:54] you create your account your API key
[14:56] just wait a couple of minutes before you
[14:58] test this Stu out all right so now let’s
[15:00] take a look at the output here as you
[15:02] can see it gives us the longitude
[15:04] latitude and again all this data is
[15:06] coming into to us via our open weather
[15:08] map account or our open Weather API
[15:10] there uh it gives you the temperature
[15:12] what it feels like pressure humidity all
[15:15] of this right in a in a table format
[15:17] because you output you also have the
[15:19] option to have the output in table
[15:20] format in Json or in a schema view but
[15:24] the table view this also gives you the
[15:26] data in a really well structured way and
[15:29] as you can see the ID and the name is
[15:31] the San Francisco because this is coming
[15:32] in from our query here all right so that
[15:36] looks good so now let’s go ahead and add
[15:38] another field to this where we want to
[15:40] convert this data that’s coming in from
[15:42] this API in this format and convert it
[15:45] into simple text so that way we can
[15:47] attach it to our um to our AI agent here
[15:50] so that way it can absorb that in plain
[15:53] text so in order to do that we’re going
[15:55] to go ahead and add another uh note to
[15:58] this and that is going to be again our
[16:00] open um open AI chat model so again
[16:03] we’re going to use our large language
[16:04] model to be able to um convert this
[16:07] table format output that we’re getting
[16:09] from our open weather map uh API into
[16:12] plain text so I’m going to go ahead and
[16:14] search for open AI click on open Ai and
[16:18] you have all these different options you
[16:19] have assistant actions you have text
[16:21] actions image actions audio actions so
[16:23] it gives you a lot more comprehensive
[16:25] ways to use this um open AI tool but I’m
[16:29] going to go ahead and click on message a
[16:31] model all right so same thing your
[16:33] parameters the credentials is going to
[16:35] be your account um the resource we’re
[16:37] going to leave it as text operation but
[16:40] but you do have option uh to um have
[16:43] image audio file and Etc all right the
[16:46] operation we’re going to leave as
[16:47] message a model um you can have custom
[16:50] API call or you can classify the text
[16:52] violation all right so now for the model
[16:54] same thing we’re going to choose from
[16:55] the list I’m going to use
[16:57] gp4 all right so for the message here
[17:00] let me give myself some more room here
[17:02] so this is where we’re going to put our
[17:03] prompt right so let me go ahead and put
[17:05] my prompt here so you don’t have to see
[17:06] me typing okay so I just basically said
[17:09] uh please convert the input from open
[17:11] weather map into plain English with a
[17:13] friendly tone to make it prettier use
[17:15] relevant emojis again you can work with
[17:18] the prompt and try to change it based on
[17:20] the result you get to make sure that
[17:22] it’s exactly the way you want it but for
[17:24] now this is good for this example right
[17:26] and the data so I’m going to say here is
[17:31] the data and I’m just going to again
[17:34] grab the data that’s going to be coming
[17:36] in from here so you have several options
[17:38] you can uh grab individual notes so for
[17:40] example I could just grab this and if I
[17:42] bring it here on the bottom as you can
[17:44] see it just says uh json. weather so
[17:46] this going to give you the weather there
[17:49] but if you want to again grab everything
[17:50] and not worry too much about it you’re
[17:52] just going to you come click on that
[17:53] Json field here and literally just grab
[17:57] everything right so I’m just going to go
[17:58] ahead and like select everything and
[18:01] come here and paste all right so that’s
[18:04] good let’s go ahead and test this out so
[18:06] on the bottom of the Roll you just leave
[18:07] it as user for now so let’s go ahead and
[18:09] test it out let’s see if this works and
[18:12] again all this data here it’s coming in
[18:14] from our previous note which is our um
[18:16] open weather map API that we called uh
[18:19] to grab the data for San Francisco and
[18:21] Bam there you go as you can see on the
[18:23] right hand side here really cool way to
[18:26] put all of that data and convert it into
[18:28] a m Mage and again this is coming in
[18:30] because of our prompt here right um so
[18:32] same thing play around here if you see
[18:35] that hey you know what you don’t like
[18:36] some of this or if you don’t like the
[18:38] Emojis just change the um change the
[18:41] prompt to to your liking but for now
[18:43] this is good to go again this is a table
[18:45] format but the Json format gives you a
[18:47] more uh structured way to see this this
[18:49] data all right cool so we’re pretty much
[18:51] done from here so now what we’re going
[18:53] to do is add another quy to this so that
[18:55] way we can convert this data because
[18:57] again like I said this is still in a
[18:59] format that has the role in it the
[19:01] content in it this refusal all the stuff
[19:03] in there but what I want to do is just
[19:05] make this really nice and simple so that
[19:08] way we can have access to it with our AI
[19:10] agent so what I’m going to do is again
[19:11] same thing add another query here so I’m
[19:14] just going to go ahead and go to data
[19:16] transformation edit fields and this time
[19:19] I’m going to change this to
[19:24] response change the name to response
[19:26] cool and then the same thing you just
[19:28] just leave the modee as manual mapping
[19:31] um for add field you click and you can
[19:33] name this so I’m going to same thing
[19:34] name this
[19:36] response keep it as string because we’re
[19:39] going to grab the text here right and
[19:40] for the value now all I have to do is
[19:43] just grab this uh this content here
[19:46] right so all I have to do is just if I
[19:48] go to
[19:49] schema I can
[19:51] whoops I can just grab the
[19:56] content perfect okay so that’s cool so
[19:59] we’re done there and now let’s go ahead
[20:01] and test this out again so I’m going to
[20:02] click on test app and there you go you
[20:04] got uh the nice response here this an
[20:06] adjacent on from let’s go ahead and
[20:07] table and this gives you um you know the
[20:10] better view that hey like this now the
[20:13] me this message is only coming in from
[20:15] this the execution of this workflow
[20:17] right and this messages is under the
[20:20] response field that we have set so now
[20:22] we going to do is grab this respond
[20:24] field and add it to our AI agent so that
[20:28] way it can has access to this message
[20:30] only when we’re chatting with it so
[20:33] let’s go ahead at this point we’re done
[20:35] with this as you can see this is
[20:37] successful uh so now we’re going to
[20:39] switch to our previous workflow our a uh
[20:43] AI agent workflow and we’re going to now
[20:46] connect this workflow tool or sorry
[20:48] connect this workflow to this AI agent
[20:51] via this n8n workflow tool so this is
[20:54] what as I mentioned before makes NN
[20:56] really uh different from other tools
[20:58] tools because you can add separate
[21:00] workflows to it all right so you can
[21:01] again rename it but I’m just going to
[21:03] leave it as it is for now uh parameters
[21:05] so let’s go ahead and name this
[21:07] parameter so I’m just going to say
[21:08] weather tool right because again this is
[21:11] uh um our weather workflow that’s coming
[21:15] in from that separate tool there so I’m
[21:16] just going to name it weather tool you
[21:17] can name it whatever you want all right
[21:19] so for the description oh give me a oh
[21:21] yeah you’re not allowed to have space so
[21:23] for the description this is where where
[21:24] you want to instruct this workflow you
[21:27] know to what to do so again gives you
[21:29] that uh placeholder call this tool to
[21:31] get a random color the input should be a
[21:33] string with comma separated name so this
[21:34] is just a um place order but it just
[21:37] gives you the format of how to write a
[21:39] description so what we’re going to say
[21:40] is use this tool to get the current
[21:45] weather data for a
[21:48] specific City and we’re going to say
[21:50] input should be the name of the city all
[21:55] right
[21:57] perfect that is good all right so now
[22:01] here’s kind of the magic here’s how you
[22:03] attach um your other workflows so you
[22:06] come here and select Source database um
[22:10] so if you Source this defined below this
[22:12] just gives you the option to just attach
[22:15] or copy and paste the Json uh so if
[22:18] you’re familiar with code again same
[22:19] thing you can just bring the Json from
[22:21] your other workflows and just paste it
[22:22] here or the easiest way to do is just
[22:25] click on database and you can grab the
[22:26] workflow ID the workflow flow ID is
[22:29] always if you just look at the URL on
[22:31] top here so let’s go back here if I
[22:33] click right here as you can see the
[22:34] workflow ID is this ID at the end of it
[22:37] so you know it says AI workshop. app.
[22:39] nn. Cloud
[22:56] slworking2 which means that your AI
[22:59] agent has um access to this workflow
[23:02] over here because we added this tool all
[23:04] right perfect so now we’re done there
[23:06] you go now that error disappeared
[23:07] because this workflow tool is now
[23:10] attached to this one because of the fact
[23:13] that we added that ID there right okay
[23:16] so now let’s go ahead and test this
[23:17] thing out right so I’m going to go ahead
[23:18] and click on save so now if you’re
[23:21] asking a weather related question this
[23:25] AI agent will anonymously decide to grab
[23:28] that data from this workflow tool but if
[23:31] you’re asking it anything nonrelated to
[23:34] uh weather it’s going to grab that data
[23:36] from Wikipedia and we’re going to see
[23:38] that in uh real life so let’s go ahead
[23:40] and start a chat here so earlier we
[23:43] asked it what is the capital of Spain
[23:45] and it answer the capital of Spain is
[23:47] Madrid and it used the Wikipedia tool
[23:49] here so now let’s go ahead and ask it
[23:52] what’s the weather in San Francisco all
[23:55] right perfect there you go now as you
[23:56] can see it’s using it’s giving us the
[23:58] weather in San Francisco that is 23.4 6
[24:01] celsus varying between low blah blah
[24:02] blah and on the right hand side as you
[24:04] can see it’s using our n8n workflow and
[24:07] if we get back out of here and you can
[24:09] see right now it’s using this workflow
[24:11] instead of the Wikipedia and again if
[24:13] you were to ask it another question
[24:15] about what’s the capital of another
[24:16] country whatever you will see that it it
[24:18] will use the Wikipedia instead of the
[24:20] nadn tool and this is what it’s gives it
[24:23] that ability for the AI agent to
[24:24] independently decide which tool to use
[24:26] based on what question you’re asking
[24:29] all right so hope you found this helpful
[24:30] just give you a good idea of what NN is
[24:32] again in the next few videos I’m going
[24:34] to do like comprehensive tutorials on
[24:36] other AI agents and building multiple AI
[24:39] agents or using these adult tools that
[24:41] are available in Inn to make really
[24:43] powerful W fro um and then also do kind
[24:46] of like an introduction video of what
[24:47] makes nadn better compared to these
[24:49] other automation tools so stay tuned for
[24:51] that all right hope you found this
[24:52] helpful thanks for watching I’ll see you
[24:54] on the next one
[24:56] [Music]

關係圖譜