# 01 2026-09-20 AI Workflow Automation n8n APIs

course: AI Clinic
module: AI-Clinic
date: 2026-09-20
type: transcript
video_url: https://personal-learn.armco.dev/files/_Recordings/AI-Clinic/01_2026-09-20_AI_Workflow_Automation_n8n_APIs.mp4

---
[00:09:02] Hello everyone. Good evening. Good evening. Good evening.
[00:09:11] Pradesh. Just FII we will begin our session by 7 0 5.
[00:09:18] Okay. Let's wait for everyone to join in. In the meantime,
[00:09:26] if anyone has any questions about today's session about
[00:09:30] previous sessions, etc, etc. Please feel free to ask me.
[00:09:42] I said 12.6 minute.
[00:09:45] requested for certificate for reimbursement. Still not that
[00:09:50] issue.
[00:09:51] That is something the manager will take care of. Is the meeting
[00:09:57] manager present here?
[00:10:01] Oh, hey, so bro, I think someone will be joining in a minute.
[00:10:04] You will be taking all the questions and concerns.
[00:10:07] Okay. By the way, what kind of a crowd are we expecting? How
[00:10:11] many people?
[00:10:13] Actually, this is where every cohort of every programs
[00:10:18] learners comes usually joins in so we can expect something
[00:10:22] around 100, 150. I just share once from Salman as well.
[00:10:29] Okay.
[00:10:30] I saw may ask it out.
[00:10:48] I think you have me up.
[00:10:49] Yeah, please go ahead.
[00:10:51] If I'm
[00:10:56] I'm clear if I'm creating a chatbot for actually, I'll
[00:11:03] ask a real question.
[00:11:06] In see I'm a civil engineer and I'm working with construction
[00:11:10] related some documents.
[00:11:13] Okay, the rules.
[00:11:17] Actually, I'm from Kerala and we are using Kerala building
[00:11:19] group or Indian standard code for building practices.
[00:11:24] If I have I need a chatbot.
[00:11:29] If I have a doubt on this building rule, I can ask and
[00:11:33] the clerk.
[00:11:35] If I want to create such a chatbot where I have to start
[00:11:40] actually.
[00:11:42] Okay, let me understand your problem statement better.
[00:11:45] You come from a particular niche in which to work.
[00:11:50] Yes.
[00:11:50] Yes.
[00:11:51] And you want to build a chatbot in and around the knowledge
[00:11:56] of the industry or and around your work.
[00:11:58] Is it?
[00:11:59] Yes.
[00:12:00] Exact periods, some periods.
[00:12:02] Okay.
[00:12:03] So that is something we will kind of cover in today's lecture
[00:12:07] only.
[00:12:07] Okay.
[00:12:08] There are various ways you can go about it depending on
[00:12:11] how much let's say knowledge you are talking about.
[00:12:15] Okay.
[00:12:15] If it's let's say what we will cover today will get you across
[00:12:20] for let's say 5, 10, 20 pages, maybe 50 pages.
[00:12:24] Okay.
[00:12:25] Actually, yes, well, number of pages.
[00:12:31] More than average.
[00:12:32] Like, are we talking about books?
[00:12:34] Okay.
[00:12:34] Are we talking?
[00:12:35] Actually, I will show the athletes roughly 300 pages.
[00:12:46] We are talking about 300 to let's say a thousand, 2000 pages.
[00:12:52] When in that range, we usually suggest a rag.
[00:12:55] Okay.
[00:12:56] We will not be building a rag today that but I do teach
[00:12:59] rags as well.
[00:13:00] Okay.
[00:13:01] To other cohorts.
[00:13:03] A custom built rag is the kind of solution you're looking for.
[00:13:08] And there are let's say these standard ready made
[00:13:12] hosted versions that you also get.
[00:13:16] There's something known as Gemini file store that you can have
[00:13:21] a look at and it will actually it's basically like a rag out
[00:13:26] of it's built out of the box.
[00:13:29] Okay.
[00:13:29] It works and it's also free to start with.
[00:13:33] So we will cover the essentials of how to let's say get the
[00:13:36] interface up and running and let's say if you have 300 pages
[00:13:40] of knowledge and you want to index that so that it can
[00:13:42] make a chat pod search through it.
[00:13:44] Use Gemini file store.
[00:13:47] Okay.
[00:13:48] Yeah.
[00:13:52] Also guys, I do know that a lot of you prefer let's say texting
[00:13:58] in the meeting chat just FYI it becomes a bit clunky for me to
[00:14:05] open the chat and scroll through it and then close it.
[00:14:09] And it also takes like a lot of screen space.
[00:14:12] Okay.
[00:14:12] So I would rather have you interject and ask me a question.
[00:14:17] Okay.
[00:14:18] When you have a doubt and let's say you do not understand
[00:14:21] something or you feel like, okay, if I don't understand this
[00:14:24] part, I'm going to get left behind.
[00:14:27] Ask me then and there.
[00:14:28] Let's get the doubt or the error clarified and then let's
[00:14:32] proceed.
[00:14:32] Okay.
[00:14:33] Otherwise, either you will get left out or you will keep on
[00:14:37] fixing the error while I explain and then you will miss out on
[00:14:41] the explanation and the, you know, practical or the demo.
[00:14:45] Okay.
[00:14:46] So yeah, please do that.
[00:14:49] You are free to interrupt and ask me and may continue one more
[00:15:03] sub question.
[00:15:04] Yeah, please.
[00:15:05] Okay.
[00:15:06] And if we are creating a mobile application using the same
[00:15:14] tools, which tools we actually use if we are going for a
[00:15:19] mobile application.
[00:15:20] Is it practical?
[00:15:23] Is it practically?
[00:15:25] Is it possible?
[00:15:26] Yeah, it is.
[00:15:27] See, the thing is how a mobile app is just the front end to an
[00:15:35] app that is joining that does not have an interface.
[00:15:37] Okay.
[00:15:38] So the interface can be a website.
[00:15:40] It can be a mobile app.
[00:15:42] It can be a desktop app.
[00:15:43] Okay.
[00:15:44] So usually if you want to build it, I would suggest using
[00:15:49] cloud code to build it.
[00:15:50] Okay.
[00:15:51] Back end is what we will explore today.
[00:15:54] So these are replaceable components, but without further
[00:15:57] ado, it's seven or five and there are about 120 of us here.
[00:16:01] Okay.
[00:16:02] So let's start today's session.
[00:16:04] Okay.
[00:16:04] So today's session is about an eight and okay, let me share
[00:16:08] my screen and let's proceed.
[00:16:19] Okay.
[00:16:20] So the problem and eight and solves.
[00:16:24] Let's explore any done on the outside first.
[00:16:32] This is any done.
[00:16:33] Okay.
[00:16:34] It's a project.
[00:16:35] It says AI agents and workflows that you can see and
[00:16:38] control.
[00:16:39] Okay.
[00:16:40] And it has a nice looking website.
[00:16:43] The thing is the problem it solves is before let's say
[00:16:47] back in 2015, okay.
[00:16:49] Or 2020.
[00:16:52] If you had two apps, okay, let's assume this was
[00:16:57] sorry.
[00:16:59] This was Google Drive and this was canva and for any, let's
[00:17:16] say design that you made inside canva, you wanted to
[00:17:19] export it to Google Drive.
[00:17:21] Okay.
[00:17:22] This would have been a problem because you would have
[00:17:25] to download this and then upload it in Google Drive.
[00:17:30] And the same interconnectivity problem existed across
[00:17:35] other apps as well.
[00:17:36] Now imagine Claude not having access to other apps.
[00:17:42] Okay.
[00:17:42] Like Google Drive or canva or something like this or
[00:17:45] imagine you getting your leads in a business scenario
[00:17:51] from one app.
[00:17:52] Okay.
[00:17:53] Like Apollo and then it not getting into something like
[00:17:57] air table.
[00:17:59] Okay.
[00:18:00] The problem, the basic problem was interconnectivity
[00:18:03] between apps.
[00:18:04] Reason was even though all of these apps had APIs
[00:18:08] themselves, these APIs did not support the communication
[00:18:13] with external software.
[00:18:15] Okay.
[00:18:16] And as in when the technological landscape changed,
[00:18:20] these APIs for let's say canva, Google Drive, etc, etc,
[00:18:25] etc, they became open.
[00:18:27] Okay.
[00:18:28] And that is how you have something known as Google SDK.
[00:18:32] Okay.
[00:18:33] You can look into these Google Drive SDK, for example.
[00:18:39] Okay.
[00:18:40] And you will find a lot of relevant documentation.
[00:18:43] These search results are not relevant because this is
[00:18:46] Android SDK, but you do have APIs, Google Drive API, for
[00:18:51] example.
[00:18:52] Okay.
[00:18:53] These APIs became open source.
[00:18:55] And this is what N8N helps us connect.
[00:18:58] So N8N acts like a glue.
[00:19:01] Okay.
[00:19:01] That is its basic functionality.
[00:19:03] That was the main purpose.
[00:19:04] It was invented for to act like a glue, but over time and
[00:19:09] with the advent of AI and everything, it has also amassed
[00:19:13] a lot of AI workflow orchestration capabilities.
[00:19:18] What does these terms mean?
[00:19:21] It means that you can spin up an AI agent.
[00:19:24] That does something and its results get passed on to
[00:19:28] another agent and that does something.
[00:19:31] And along with the connectivity that it provides with
[00:19:33] other apps, it actually ends up becoming a very useful
[00:19:37] component that we can make a lot of automations with.
[00:19:41] Am I making sense?
[00:19:46] Okay.
[00:19:47] So we will go ahead and explore N8N.
[00:19:51] Okay.
[00:19:52] On the outside, then we will explore what the
[00:19:56] interface, etc.
[00:19:57] looks like.
[00:19:58] We will create an account and then we have a few demo
[00:20:01] automations that we will create for today's session.
[00:20:05] Okay.
[00:20:06] So, yes, sir.
[00:20:09] Yeah.
[00:20:11] Does this connect agents to agent?
[00:20:14] Yeah.
[00:20:14] You can connect agents to agents.
[00:20:17] Okay.
[00:20:18] Thank you.
[00:20:19] I have sent everyone the link.
[00:20:22] Okay.
[00:20:23] Go and open it up and create an account.
[00:20:28] By the way, when you're on the website, you will
[00:20:31] notice this icon, which is a GitHub icon.
[00:20:34] This means that the code base for N8N is also
[00:20:39] publicly available, which means, see N8N has a pricing.
[00:20:44] Okay.
[00:20:44] It costs like 2000 rupees a month.
[00:20:47] Okay.
[00:20:47] On its smallest plan.
[00:20:49] But if you want to, then you can download a copy
[00:20:53] of N8N, run it on your own computer or your
[00:20:56] offices servers and also work on it for free.
[00:21:02] Now, obviously there are a few, let's say license
[00:21:04] requests or requirements.
[00:21:07] Okay.
[00:21:08] But those are for decent enough, decently big
[00:21:11] organizations.
[00:21:12] It's not the most relevant.
[00:21:14] Okay.
[00:21:15] But the point is it is open source.
[00:21:18] Now, if one was to ask me, Hey, do you know about
[00:21:21] make and Zapier and how is it better than those?
[00:21:26] I would say make and Zapier and N8N, they do very
[00:21:30] similar things, but I would personally prefer N8N.
[00:21:34] Reason being it is open source.
[00:21:36] Plus I can own a copy of N8N.
[00:21:39] So even if let's say they change this 20 euro pricing
[00:21:42] to 2000 euros all of a sudden, I still have a copy
[00:21:46] of N8N that I can run myself.
[00:21:48] Plus it is more easy to use.
[00:21:50] Okay.
[00:21:51] And we will explore why it is easy to use.
[00:21:55] Okay.
[00:21:56] So I would request everyone to take two minutes.
[00:21:59] Okay.
[00:22:00] And do the sign up at this point because for everything
[00:22:04] from now we will require the N8N thing.
[00:22:07] Okay.
[00:22:08] Also while signing up, it will tell you that you need
[00:22:12] to sign up with a business email that is a false
[00:22:16] requirement.
[00:22:17] You can sign up with any email.
[00:22:24] Once done.
[00:22:26] Yeah.
[00:22:27] I tried that but the magic link is not coming to the mail.
[00:22:33] Well then use a different mail, check your spam.
[00:22:37] Use your temporary mail.
[00:22:40] Okay.
[00:22:41] Another mail.
[00:22:42] Yeah.
[00:22:43] Try another mail.
[00:22:45] Okay.
[00:23:04] Once done.
[00:23:05] Okay.
[00:23:05] Just drop like a done in the group because it helps
[00:23:09] me understand how to progress with the rest of the class.
[00:23:16] It's the get started the free trial one.
[00:23:18] Right.
[00:23:19] That's how we have to go for the free trial.
[00:23:22] It lasts you 14 days.
[00:23:24] It has enough credits to get you through.
[00:23:26] Okay.
[00:23:27] And yeah, we will explain the rest as in when it comes.
[00:23:52] Yes.
[00:23:52] Locate tell me.
[00:23:55] Yeah.
[00:23:55] Roy, can you also explain like how to work in open source
[00:23:58] like in GitHub because I am getting rid of free
[00:24:02] trial by my free till getting over by today.
[00:24:05] That's why I'm asking.
[00:24:06] Okay.
[00:24:07] If you are running out of the free trial, then what we will do is I will at
[00:24:13] least show you how to set it up in Docker.
[00:24:16] Okay.
[00:24:18] Okay.
[00:24:18] Yeah, but that would be at the last because this is like an introductory
[00:24:22] class jumping right into how to install things with Docker isn't going to suit
[00:24:27] the rest of the audience as well.
[00:24:29] Yeah.
[00:24:29] Sure.
[00:24:29] Thank you for that.
[00:24:31] Okay.
[00:24:31] Okay.
[00:24:31] And if anyone else also has the same requirement just like Nithil.
[00:24:36] Okay.
[00:24:36] Do drop that in the comment because then I'll understand like how what is the
[00:24:43] understanding of the class and how does it want to see things.
[00:24:48] That's it.
[00:24:51] Okay.
[00:24:52] Yeah.
[00:24:52] How deep we are planning to go for the decision.
[00:24:55] Oh, very likely how deep we are planning to go.
[00:24:57] I mean, what could we be connected with Google Drive, Google Docs, Google
[00:25:02] Sheets and what see if we will try connecting it with a landing page.
[00:25:08] And if we have time, then we will try doing it with telegram as well.
[00:25:13] Okay.
[00:25:14] But it all depends on how the class proceeds.
[00:25:18] And let's say the amount of doubts let's say I get.
[00:25:20] Okay.
[00:25:21] And try to connect LinkedIn here.
[00:25:25] I'm sorry.
[00:25:27] Yeah.
[00:25:27] Try to connect LinkedIn.
[00:25:34] Yes, we will.
[00:25:36] We are not exactly trying to connect Gmail, but it's not difficult as well.
[00:25:42] After like what we go over it will not be difficult either.
[00:25:49] Hello.
[00:25:50] Yes, tell me.
[00:25:52] Roy, I have I'm getting some NPN how much you want like the only me at 20,
[00:25:58] 20 to 99 or something like that.
[00:26:01] So what to choose here?
[00:26:03] Like members, you're getting options.
[00:26:09] The options are getting ABCD, EFG up to and it is asking how many members are
[00:26:15] there in your company, something like that.
[00:26:17] It's not matter.
[00:26:18] Just fill up anything.
[00:26:19] Anything.
[00:26:20] Select only me.
[00:26:21] Select only me.
[00:26:22] Okay.
[00:26:23] That's what I selected.
[00:26:25] Yeah.
[00:26:26] Okay.
[00:26:26] Thank you.
[00:26:26] So Roy, I have a little and your workspace is ready.
[00:26:31] I think that is the page we're expecting.
[00:26:32] Yeah.
[00:26:33] Yeah.
[00:26:34] So essentially one, once your workspace is ready, you should have either this.
[00:26:41] Okay.
[00:26:41] And if you go to AI assistant preview, you should have this.
[00:26:46] So once you get to this screen, just stop.
[00:26:50] Okay.
[00:26:51] Okay.
[00:26:52] That's all.
[00:26:54] I'm a completely pressure.
[00:26:55] Am I using any company email ID?
[00:26:58] Is it fine?
[00:27:00] Yes, you are.
[00:27:01] You can use any email.
[00:27:04] Okay.
[00:27:04] So let's start with today's content.
[00:27:07] So what I have thought, okay, is we will essentially automate the ordinary,
[00:27:14] repetitive operations, any and every company faces.
[00:27:18] Okay.
[00:27:18] So in this, we will have three, four automations.
[00:27:22] One is a company HR policy bot.
[00:27:25] Okay.
[00:27:26] So if you're an employee and you want to know, hey, how many leaves can I paid
[00:27:30] leaves, can I take in a year?
[00:27:32] How many casual leaves or sick leaves can I take in a year or what might let's
[00:27:35] say companies, maternal leave policy is, I can search that and it will answer
[00:27:41] according to my company's HR docs.
[00:27:44] And I can get to set those documents in like the way I want to.
[00:27:49] Okay.
[00:27:50] Second is, we will also build a finance tracker.
[00:27:56] Okay.
[00:27:56] So what happens is for most organizations, a lot of times employees make expenses
[00:28:02] and those expenses get reimbursed.
[00:28:05] Okay.
[00:28:05] By the company.
[00:28:06] So instead of you having to run to the HR or the finance guy and tell him,
[00:28:11] Hey, here's my $20 slip.
[00:28:13] Please pay me back.
[00:28:14] Why don't you simply upload the bill to a place and it gets
[00:28:20] collected in an Excel.
[00:28:22] It gets approved by a bunch of rules and then the finance guy gets access to it.
[00:28:28] Okay.
[00:28:29] The third thing is we can also build a recruitment scanner.
[00:28:33] Okay.
[00:28:34] So what will happen is we will, we can connect it, let's say to your, let's
[00:28:40] assume you have a company and you have, let's say an apply page.
[00:28:43] Okay.
[00:28:44] There you will have a bunch of requirements internally in the company.
[00:28:48] So if any candidate uploads a resume, the resume gets scanned.
[00:28:53] Okay.
[00:28:54] Oh, it does a matching.
[00:28:56] Okay.
[00:28:57] Whether the resume or the job is even relevant to the job role or not.
[00:29:01] If it is, let's say there's an 80% match, then we at least book an initial
[00:29:07] call with the candidate.
[00:29:08] Now, obviously these are very generic use cases and from the reason we are
[00:29:14] doing this is so that these can be transposed over to other practical use cases.
[00:29:20] Just like one member of the audience was saying, Hey, can we build, let's say a
[00:29:25] chatbot for my construction related knowledge?
[00:29:29] Yes, you can replace the knowledge in the HR policy bot and at least get
[00:29:33] our app starting to work with.
[00:29:36] Okay.
[00:29:37] So let's proceed unless anyone has any questions.
[00:29:42] Can you just show the workspace again?
[00:29:43] Now, how it looks like.
[00:29:46] Oh, right.
[00:29:46] So she can, uh, okay.
[00:29:49] So, guys, I will ask you if you're not meaning to ask any questions, please
[00:29:55] keep yourself muted.
[00:29:57] Okay.
[00:29:59] Also, with regards to the workspace, what do you want to know?
[00:30:02] I just wanted to check.
[00:30:04] This is the screen.
[00:30:06] We are up to, okay.
[00:30:07] Fine.
[00:30:07] That's fine.
[00:30:07] You should land in this page.
[00:30:09] Okay.
[00:30:10] You can go here.
[00:30:12] I am seeing this one, the one which are showing.
[00:30:17] Okay.
[00:30:17] This one, right?
[00:30:18] Yeah.
[00:30:19] Come to this.
[00:30:19] Yeah.
[00:30:20] Fine.
[00:30:20] How we are landing to this approval.
[00:30:23] We just created the account and we are.
[00:30:25] You must have gotten a bunch of questions.
[00:30:28] How big is my team?
[00:30:29] What is my name?
[00:30:30] Etc.
[00:30:35] Yeah.
[00:30:35] Yes.
[00:30:36] That's correct.
[00:30:37] Yeah.
[00:30:37] So, uh, just fill up anything you will land on this page.
[00:30:41] This thing.
[00:30:43] Okay.
[00:30:44] From here.
[00:30:47] Hello, right?
[00:30:48] Sorry.
[00:30:49] Uh, I, instead of a assistant, I'm getting differently in eight in assistant.
[00:30:54] Is it okay or it's a different?
[00:30:58] I'm not sure what's there is a, there is a option.
[00:31:00] Yeah.
[00:31:01] Yeah.
[00:31:01] There is a options.
[00:31:03] It is same.
[00:31:04] Instead of AI, we are showing any time in, yeah, in a 10 in assistant.
[00:31:10] Just go to this.
[00:31:13] Yeah, it's same as a slash assistant.
[00:31:17] This is a, you are a, but even I do am saying any assistant on the left side.
[00:31:22] And now can someone share the screen and show me what they're talking about?
[00:31:27] Yeah, you are.
[00:31:29] Okay.
[00:31:30] I'm sharing it actually.
[00:31:32] Uh, right.
[00:31:33] This is the screen we are.
[00:31:36] Yeah.
[00:31:37] Correct.
[00:31:37] I am adding is an eight assistant instead of 10.
[00:31:42] Yeah.
[00:31:43] Okay.
[00:31:44] I mean, yeah, it's the same thing.
[00:31:46] Yeah.
[00:31:46] It's the same thing.
[00:31:48] Okay.
[00:31:48] Fine.
[00:31:49] Thank you.
[00:31:49] Thank you.
[00:31:50] Thank you.
[00:31:50] Thank you.
[00:31:51] So what happens is a lot of times, okay, these companies do something known as AB
[00:31:57] testing.
[00:31:57] Okay.
[00:31:58] They will, let's just show you two copies of the same ad and see which one gets
[00:32:02] more clicks.
[00:32:03] The same happens with a lot of such features as well.
[00:32:07] They will see whether calling it AI assistant gets more clicks or calling it any 10 assistant
[00:32:12] gets more clicks.
[00:32:14] Okay.
[00:32:14] So it's just an AB testing the feature that is getting AB tested.
[00:32:19] Okay.
[00:32:19] So the first thing you need to understand about this and I need you to set that
[00:32:24] mindset is this AI assistant is buggy by nature.
[00:32:29] Okay.
[00:32:30] What used to happen back in the days, and it had a very drag and drop
[00:32:36] interface.
[00:32:36] Okay.
[00:32:37] And I need you to understand this carefully because whatever we build.
[00:32:44] Okay.
[00:32:45] A lot of it might just not work properly in the start.
[00:32:49] Okay.
[00:32:50] And that is a part and parcel of building things with any done building
[00:32:55] and automation will take, let's say five ish minutes, fixing the automation
[00:33:01] to bring it to a point where it runs seamlessly is going to take another
[00:33:05] 20 minutes.
[00:33:07] So have that mindset ready.
[00:33:10] Okay.
[00:33:11] And don't get demotivated or don't feel apprehensive if it doesn't work
[00:33:18] in the first chance.
[00:33:19] Okay.
[00:33:20] On the same note, I wanted to, I wanted to show you something.
[00:33:26] Let me find where are the templates?
[00:33:31] Integrations templates.
[00:33:32] So if you go to the main page, okay, under product, you have an option
[00:33:37] called templates.
[00:33:38] Okay.
[00:33:39] So sure you get a lot of ready made workflows.
[00:33:43] Okay.
[00:33:44] Workflows can be things like, Hey, posting to LinkedIn post making an article
[00:33:48] and making an article.
[00:33:54] Guys, I would request if you're not intending to speak, please keep yourself
[00:33:58] muted.
[00:33:59] Okay.
[00:34:00] So it has a lot of pre made templates like, Hey, create a content piece of
[00:34:06] content and post it to LinkedIn or create an image posted to Instagram or
[00:34:10] let's say a bunch of these things.
[00:34:12] Okay.
[00:34:12] You can click each of these and understand what they look like.
[00:34:19] Now you will notice that each of these has a bunch of nodes.
[00:34:26] These boxes is what we are also going to create these boxes.
[00:34:31] Okay.
[00:34:31] These are called nodes.
[00:34:33] The reason I am showing you this is because sometimes the NIT and AI uses the wrong
[00:34:40] node or connects things a bit in a messy fashion.
[00:34:44] Okay.
[00:34:47] Okay.
[00:34:47] Coming back to it, let's take our first idea.
[00:34:52] Okay.
[00:34:52] The HR policy bot and let's type it out to the best detail that we possibly can.
[00:34:59] I want to build a HR policy bot for my company.
[00:35:12] I want to build it like a, I want to build it like a chat bot, but essentially
[00:35:22] what I will do is I'll connect it with a landing page.
[00:35:24] Okay.
[00:35:25] So I will create a form.
[00:35:28] Okay.
[00:35:29] A form just takes a question.
[00:35:33] Okay.
[00:35:34] And it gives an answer.
[00:35:35] I want to build it like a form.
[00:35:38] It should take a question from the user.
[00:35:46] Look up the information relevant to answer the query from a Google doc and answer
[00:36:04] it back to the user.
[00:36:08] Build me this workflow.
[00:36:13] Okay.
[00:36:14] This should be enough for starters.
[00:36:16] If we need to, then we will also fix things.
[00:36:19] Okay.
[00:36:20] Once you click enter, it's going to take 30 seconds.
[00:36:25] It will start spinning up the entire automation.
[00:36:34] Also, I will copy and paste these prompts.
[00:36:37] Okay.
[00:36:38] In the doc and I'll share you the link to it.
[00:36:43] But guys, beware of one thing that my prompt can get me an error.
[00:36:50] Okay.
[00:36:51] Your prompt might not get you an error.
[00:36:54] It is very unpredictable.
[00:36:56] Okay.
[00:36:57] I will, I will prefer if you guys are to simply write it by yourself.
[00:37:05] Supremu, what is the difference between power like an 8n power
[00:37:11] automate and UI path?
[00:37:15] UI path.
[00:37:16] See, I have not used power automate or UI path, but I know that UI path is used
[00:37:22] for what they call as an RPA.
[00:37:26] Okay.
[00:37:27] So the way I have built our PR software is basically an automation for more
[00:37:37] technical use cases.
[00:37:39] UI path is costlier.
[00:37:40] I have, I would prefer using any 10 for light use cases.
[00:37:47] If I had to use UI path, I would use it for very complex RPA processes.
[00:37:52] For example, for a machining use cases and factories.
[00:37:58] Okay.
[00:37:58] A lot of RPA use use cases exist for those as well.
[00:38:03] But because I have not used those two, I will not be able to give you
[00:38:09] the best answer possible.
[00:38:11] I mean, isn't it like a AI agent is like RPA plus LLM?
[00:38:16] I mean, that's it, right?
[00:38:17] I mean, we are just designing a workflow and it's just LLM at the
[00:38:21] back end giving the response.
[00:38:23] See, I am sure that there's more than that.
[00:38:25] Okay.
[00:38:26] In the sense that there are, there might be one primary goal that one tool
[00:38:32] helps you achieve, but there also might be 10,000 other supporting
[00:38:37] functions that make it a completely different product.
[00:38:40] So because I have not used UI path, I might, I will not be able to give you
[00:38:45] the perfect comparison between any 10 and UI path.
[00:38:50] Oh, one more question.
[00:38:52] See how different from the agent work flows.
[00:38:56] For example, I'm sorry, I just joined bit late probably I may have
[00:39:01] skipped a few things, but still I just want to understand like how this
[00:39:05] N18 is different from the agent work flows.
[00:39:09] We are creating agent work flows using N18.
[00:39:13] It's not different.
[00:39:14] It's one of the ways you can create an agent.
[00:39:16] Got it.
[00:39:17] Got it.
[00:39:18] Thank you.
[00:39:20] Okay.
[00:39:22] So yes, Dhruv, tell me.
[00:39:27] Hello.
[00:39:27] Just one more thing.
[00:39:28] Like yesterday you showed us like a proper website.
[00:39:33] Okay.
[00:39:34] So like how will this be useful there?
[00:39:38] Okay.
[00:39:38] So imagine that you have a company website where the people can log in
[00:39:44] and file their expenses or they can ask their queries or if you have
[00:39:49] an AI chatbot on your website.
[00:39:51] So this will help you create the backend for the chatbot while the
[00:39:57] website remains the front end.
[00:39:59] Okay.
[00:40:00] You will and N18 will basically act as the glue between the front end,
[00:40:04] the website and the backend, which is N18 itself.
[00:40:09] All right.
[00:40:10] Okay.
[00:40:11] So what we have here is this graph.
[00:40:15] Okay.
[00:40:15] Well, it has listed all of these nodes.
[00:40:18] Okay.
[00:40:18] Just like we saw in the templates as well.
[00:40:21] Okay.
[00:40:22] Now it has a form node.
[00:40:24] It has a Google doc node.
[00:40:26] It has an AI node.
[00:40:28] Okay.
[00:40:29] And it has an show answer node.
[00:40:31] Okay.
[00:40:32] Inside the AI node, it has a LLM model.
[00:40:36] It is invoking open AI.
[00:40:40] Okay.
[00:40:40] Now let's fill everything up to get it up and running.
[00:40:44] So on the left, it is waiting for us to configure a Google account with it.
[00:40:50] Okay.
[00:40:51] So what we need to do is obviously it needs to connect to your account
[00:40:55] because your account is what has your company's policy.
[00:41:00] We will sign in.
[00:41:03] Let's sign up.
[00:41:07] I will just give it all access because this Gmail is empty.
[00:41:11] Okay.
[00:41:11] I don't mind it seeing my files, but obviously because this is an open source
[00:41:15] software, et cetera, et cetera, I know what it will access and whatnot.
[00:41:20] Okay.
[00:41:20] So I don't mind giving it all access either.
[00:41:23] In production use cases, you can choose to give it less access.
[00:41:27] Okay.
[00:41:28] So now that I have signed up, okay, I can select my account, which is in this case
[00:41:34] already selected.
[00:41:35] Now I need to select the entire ID of the doc.
[00:41:39] It will refer for my HR policies for that.
[00:41:44] Let me see if I already have a doc like that connected or not.
[00:41:49] If I don't then I will create a new one.
[00:41:55] Looks like I don't hear.
[00:41:57] I just might have your policy company HR policy.
[00:42:03] Yes.
[00:42:05] Let me just copy this.
[00:42:12] Okay.
[00:42:13] So this is the HR demo HR policy of my company.
[00:42:17] I own a company called mine graph.
[00:42:20] So I'll call this HR policy doc and I will request you to make
[00:42:30] a copy of this document.
[00:42:35] Okay.
[00:42:35] I'll paste the link in the chat.
[00:42:45] Inside this, I will paste the URL of this document.
[00:42:54] Apply.
[00:43:17] So once you're done with up to here, okay, let me know.
[00:43:22] I am testing the workflow in the meanwhile just to see if things
[00:43:25] are working or not.
[00:43:27] Then we will come back to it.
[00:43:32] Test it for now.
[00:43:33] I am just doing my own testing to make sure things are working.
[00:44:12] Once done, let me know.
[00:44:15] That Google Drive file.
[00:44:18] We have to upload that link in this N8n.
[00:44:25] You need to make a copy of this file.
[00:44:27] Okay.
[00:44:28] To the Google Drive.
[00:44:30] I mean, if you make a copy, it will come to your Google Drive
[00:44:32] only by default.
[00:44:33] Oh, yes, yes, yes.
[00:44:35] Yes.
[00:44:36] Then you need to copy the link of the new document, not my
[00:44:40] document.
[00:44:41] Okay.
[00:44:41] That will result in an error.
[00:44:43] The link to your document inside this prompt that had come
[00:44:47] up inside this prompt.
[00:44:53] Well, I got a Google account prompt, right?
[00:44:57] Where I had to sign in.
[00:44:59] Yes.
[00:45:00] Yeah.
[00:45:01] So under that, you also have a place where you have to
[00:45:04] enter the link to this document.
[00:45:08] Okay.
[00:45:12] Hello.
[00:45:13] Yes, the rope.
[00:45:13] Tell me.
[00:45:14] Yeah.
[00:45:14] Can you send the prompt as well?
[00:45:17] Yeah, I have sent you the prompt.
[00:45:19] Google Doc is not updated.
[00:45:21] Actually, I tried refreshing it.
[00:45:25] I'm not getting your comments.
[00:45:31] Google Doc is not updated.
[00:45:33] Yes, I got, I got.
[00:45:42] Just to extend this use case, right?
[00:45:45] Let's assume that I have a SharePoint or I have a web pages
[00:45:51] where the policies are there.
[00:45:52] HR policies are there.
[00:45:54] Okay.
[00:45:54] Now, two things I want to do is one is this is a very
[00:45:58] simple straightforward use case where you have one doc,
[00:46:02] right?
[00:46:02] Yeah.
[00:46:03] Now, and my policy is between multiple pages or maybe
[00:46:08] be multiple folders within a SharePoint.
[00:46:11] Yeah.
[00:46:11] Now, I want to create a knowledge base because that
[00:46:14] I mean, it will automatically should create a knowledge base
[00:46:17] and then my rendering process where I can specify the template
[00:46:21] that this is how the rendering should occur.
[00:46:25] So if I am doing that kind of a use case, how should I go
[00:46:28] about it using NA, NA, NA, NA from a prompt point of view?
[00:46:35] See, if you're going to index multiple documents,
[00:46:38] there are two elementary ways you can go about it.
[00:46:41] Okay.
[00:46:41] First is you can simply instead of fetching from one document,
[00:46:45] you can fan out and connect to multiple documents feed all
[00:46:49] of the data.
[00:46:51] Okay.
[00:46:51] To this AI LLM.
[00:46:54] Okay.
[00:46:54] And then ask it and the answer to our question.
[00:46:58] But again, as you might guess that is a very token
[00:47:01] draining way of going about things because all of the
[00:47:04] con all of the content of the documents needed needs to be
[00:47:08] uploaded to the context, which consumes a lot of tokens.
[00:47:11] The right way to go about this is to essentially create
[00:47:16] a data warehouse.
[00:47:17] Okay.
[00:47:18] Now, when I say data warehouse, I'm talking about documents
[00:47:22] that span over 200 to 50 maybe 500 gigabytes of data.
[00:47:26] Okay.
[00:47:27] You create these data pipelines that get updated every time
[00:47:32] a document gets updated.
[00:47:34] Okay.
[00:47:35] That triggers the ingestion of the newly made document or the
[00:47:42] changes changed chunks into a rag.
[00:47:45] Okay.
[00:47:46] Now the rag will only refer to the textual data, not any
[00:47:52] compute and it will not do any math.
[00:47:54] Okay.
[00:47:55] So if you're uploading excels, then you need a proper SQL
[00:47:59] and you need like a math engine behind it.
[00:48:02] Along with that, you also need to store structured data
[00:48:06] like Excel sheets in a SQL DB.
[00:48:08] Okay.
[00:48:09] All of these multiple components need different agents
[00:48:12] that will manage them on top of it needs to be a massive
[00:48:16] agent that does the parsing and understanding of the
[00:48:21] type of question it is.
[00:48:22] Then you get to give the answer.
[00:48:25] No, no, so my my thing is I'm using a SharePoint
[00:48:28] as a backend.
[00:48:29] Okay.
[00:48:30] And I'm rendering a I'm using a SharePoint portal as a front-end.
[00:48:34] Okay.
[00:48:36] Now my data is already sitting in a structured data model
[00:48:40] within the SharePoint.
[00:48:41] Right.
[00:48:42] Why I would need to create any kind of data warehouse,
[00:48:47] right?
[00:48:47] I can directly use the SharePoint as a backend and create
[00:48:51] a knowledge layer, knowledge brain layer on top of it where
[00:48:54] I do a vectorization and indexing and that can be
[00:48:58] used with the context.
[00:48:59] I can use that vector and index to get the right content
[00:49:05] and pass it to the LLM.
[00:49:07] We are talking about the same thing.
[00:49:08] We are talking about creating the same thing.
[00:49:12] Okay.
[00:49:12] So now if I am doing a let's say I have a share data
[00:49:15] sitting in the SharePoint how I can leverage the prompt
[00:49:19] to spin up this application.
[00:49:23] I know my data source, right?
[00:49:25] What I'm saying is I know in this context that as you
[00:49:27] said that the HR policy is sitting in a document,
[00:49:29] right?
[00:49:30] That is a kind of a knowledge base you have given in a workflow.
[00:49:35] Correct?
[00:49:36] Yeah.
[00:49:37] And that means set up after doing a workflow.
[00:49:39] After setting up this workflow, you are just setting up
[00:49:41] that from where this HR policy doc is coming, right?
[00:49:44] That's what we have given a URL and then that URL
[00:49:47] has been taken and then it going through LLM process.
[00:49:50] Correct?
[00:49:51] Yeah.
[00:49:53] In this step, I'm just adding one more step where
[00:49:55] I say that before I can even do it, I need to create
[00:49:59] a vector of my data, right?
[00:50:02] So if I have to do that kind of implementation, how I can go
[00:50:05] about it using it.
[00:50:09] See, I kind of wait.
[00:50:14] Sorry to enable.
[00:50:18] Well, I think you are miswriting the URL.
[00:50:21] I will simply share the URL again because this should
[00:50:26] not happen.
[00:50:27] See, everyone else is accessing it.
[00:50:29] You should be able to click it as well.
[00:50:34] See, I think I am not understanding your like the difference
[00:50:38] between what you said and what I said.
[00:50:40] Maybe I can take the question later.
[00:50:43] What I can say is see your data might be sitting in SharePoint,
[00:50:48] but it still needs to be index and vectorized like you said.
[00:50:51] Okay.
[00:50:52] That is why I was suggesting key user rag, which is necessary.
[00:50:56] Okay.
[00:50:58] Also, the reason we need a data pipeline is because
[00:51:01] your data might just get updated every once in a while
[00:51:05] and it might be coming from different sources.
[00:51:08] Plus all of the data will not be just usable out of books.
[00:51:12] Okay.
[00:51:12] It needs some sort of parsing as well.
[00:51:15] So that is where we extract the useful content from the raw
[00:51:19] data.
[00:51:19] That is why we use a pipeline.
[00:51:22] Maybe I can cater to this specific question later
[00:51:26] because it does not pertain to the exact requirement
[00:51:28] of the class.
[00:51:29] Okay.
[00:51:32] So, please come to me.
[00:51:36] I'm sorry, Samrat.
[00:51:37] I could not hear you.
[00:51:39] I'm sorry.
[00:51:39] I just wanted to say that let's please start in because
[00:51:42] yeah, yeah, yeah.
[00:51:43] Professor relatively new to NADN and maybe you can take
[00:51:46] advance and switch it up.
[00:51:47] Yeah.
[00:51:49] Okay.
[00:51:49] So again, before we...
[00:51:51] Right.
[00:51:52] Yes, Karthik.
[00:51:54] I'm getting a problem loading form.
[00:51:56] This is usually occur with the NADN book flow serving.
[00:51:58] This form is deactivated or no longer exist.
[00:52:05] Okay.
[00:52:05] Let me come back to you.
[00:52:06] I need to show you guys a couple of things before we
[00:52:10] come back to this.
[00:52:12] Okay.
[00:52:12] So the first thing is you have a bunch of executions.
[00:52:17] Okay.
[00:52:18] Here zero out of thousand executions.
[00:52:21] This means that on a on any NADN account, you get to use
[00:52:25] this app or any app for let's say 10 a thousand times.
[00:52:30] Okay.
[00:52:31] A thousand times invoked from an external source.
[00:52:35] We are not counting the testing that we are doing internally.
[00:52:39] The second thing is if you go to settings, you will see there
[00:52:43] is something known as gateway credits.
[00:52:45] Okay.
[00:52:45] And it says $2 remaining.
[00:52:47] Okay.
[00:52:48] Which means see this open AI node that was being invoked.
[00:52:52] Okay.
[00:52:53] Obviously it needs to have some AI credits.
[00:52:56] Okay.
[00:52:57] And this free free account comes with $2 worth of AI credits.
[00:53:02] Okay.
[00:53:03] The third thing you should notice is here's a dollar sign.
[00:53:08] Okay.
[00:53:08] This dollar sign says the amount of credits.
[00:53:13] This particular chat has taken in so much time.
[00:53:19] This chat has taken 74 credits out of about what 750 credits
[00:53:26] on an average from what I can tell.
[00:53:28] Okay.
[00:53:28] These credits are usually enough.
[00:53:31] Okay.
[00:53:32] But yeah, essentially even this AI chat takes some credits.
[00:53:37] Now Dhruv come again.
[00:53:43] Hello.
[00:53:44] So actually I was not able to load my document.
[00:53:46] Can you help me out?
[00:53:50] What do you mean not load your document?
[00:53:52] Like I copied the file and then I pasted the URL in the
[00:53:57] in this in it in it is not able to read it.
[00:54:02] Share your screen.
[00:54:04] Okay.
[00:54:09] Anyone else who's facing the same issue?
[00:54:12] Yeah, I just shared a messenger.
[00:54:16] So that is another I've been led up to.
[00:54:20] Okay.
[00:54:20] Let's fix this for everyone.
[00:54:22] Share your screen.
[00:54:25] Is my to execute the workflow now or just need to link the
[00:54:35] document.
[00:54:37] Execute the workflow.
[00:54:38] Okay.
[00:54:39] Tested whether it's working.
[00:54:44] So is my screen visible?
[00:54:45] Hello.
[00:54:45] Yeah.
[00:54:46] Copy the link.
[00:54:47] Yeah, I did copy the link.
[00:54:50] Okay.
[00:54:52] Is this Google Doc in the same Google account as the one
[00:54:59] that you signed up here with?
[00:55:02] Yes, yes, it is same.
[00:55:04] Okay.
[00:55:09] No, when I try executing this step it says error.
[00:55:14] Why don't you cross this?
[00:55:24] This is a get HR policy doc panel.
[00:55:31] This square that has come up.
[00:55:33] Yes.
[00:55:34] Click on.
[00:55:35] Yeah, click on.
[00:55:36] Did you try fix with AI?
[00:55:38] No, no, I did not.
[00:55:40] Okay.
[00:55:40] Try that because a lot of times it takes, let's say two,
[00:55:44] three times tries like I said, okay.
[00:55:47] And then it automatically fixes things.
[00:55:50] All right.
[00:55:51] I will do that.
[00:55:52] If it still doesn't work, then what you can do is you can try
[00:55:56] connecting a different Google account.
[00:55:58] All right.
[00:56:00] Okay.
[00:56:01] I'll thank you.
[00:56:02] Yes, Sujata tell me.
[00:56:04] Yes, perpropagates work for me, but it's a different
[00:56:06] question for example, I could see we can achieve the same thing
[00:56:11] in our cloud, right?
[00:56:12] Using the NIT and the workflow, we can achieve the same thing
[00:56:15] in cloud, right?
[00:56:17] So is just the brand is different or tool is different?
[00:56:22] Like, I just, I just Googled it.
[00:56:26] I just checked what is the difference between why can't we
[00:56:28] go ahead with cloud?
[00:56:29] I could see there is an AI reasoning can be done in cloud
[00:56:32] which can't be done in anything.
[00:56:35] So what makes it think so?
[00:56:36] We can connect a cloud node or we can connect a general
[00:56:40] LLM endpoint, okay?
[00:56:42] And we can use cloud inside of it.
[00:56:45] The advantage NIT and will give us over, let's say using
[00:56:48] cloud is it will come with so many different connectors.
[00:56:54] That's it.
[00:56:54] The connectivity is the advantage.
[00:56:56] Nothing else.
[00:56:58] Actually for a lot of use case depending on the use case
[00:57:01] or depending on how heavy weight the use case is, I would
[00:57:05] rather, you know, stick to cloud.
[00:57:08] But it is good for let's say wide coverage rather than
[00:57:11] in-depth coverage.
[00:57:13] Okay.
[00:57:14] Okay.
[00:57:15] Let's proceed with the rest because otherwise we will have
[00:57:19] a time crunch at the end.
[00:57:20] So, it's a problem, but if I am executing the workload
[00:57:24] flow and I am writing the question topics from the
[00:57:27] document as well, like casual leave and personal leave, I'm
[00:57:30] getting this document does not cover these features.
[00:57:36] Okay.
[00:57:36] Have you seen what content is being fetched?
[00:57:41] Let me show you how you can debug.
[00:57:43] Trying to fetch the content, but I'm not getting that.
[00:57:47] Okay.
[00:57:47] So let me show you something about the how you can debug
[00:57:51] which part is failing.
[00:57:53] Okay.
[00:57:53] So first you can click on any of these nodes.
[00:57:57] It opens up this panel.
[00:57:58] Okay.
[00:57:59] Now inside it, you get to see what is the input it
[00:58:03] takes on the left, how it works or how it, you know,
[00:58:07] computes or manipulates or things in the middle and the
[00:58:11] output on the right.
[00:58:12] Okay.
[00:58:13] So I tested a demo question asking how many maternity
[00:58:17] leaves can I take in a year?
[00:58:19] This was the question.
[00:58:20] It gave it a timestamp and the form was a test.
[00:58:25] Okay.
[00:58:25] Because obviously this is not being hosted yet.
[00:58:28] If you click on the second node, you will see that
[00:58:31] the same question I passed actually came here.
[00:58:35] Okay.
[00:58:36] And obviously it did some connection.
[00:58:39] Okay.
[00:58:39] With the tools and here is the content that the document
[00:58:44] fit that fetch that got fetched from the document.
[00:58:47] Okay.
[00:58:48] So if your chatbot is not being able to answer a query,
[00:58:53] there can be three things happening.
[00:58:55] One is the information is actually not present or the
[00:59:00] document is not being loaded properly.
[00:59:03] Okay, which you can verify verify from fear or thirdly
[00:59:08] the document might just be so long.
[00:59:11] Okay.
[00:59:12] That it is overflowing the context of the LLM, which is a very
[00:59:16] rare case, but obviously something that can happen.
[00:59:19] Okay.
[00:59:20] So you can take your time and debug that, but the standard
[00:59:25] procedure of debugging is either go with fix with AI button
[00:59:28] click it or just look into the input outputs after executing
[00:59:35] the test workflow.
[00:59:36] Okay.
[00:59:41] Can you come again once on the cloud or I need to write
[00:59:46] one of the parts will just ask right now.
[00:59:48] So only sorry to interrupt.
[00:59:50] So Prabha, is it okay to finish what we plan and then
[00:59:53] we can get the questions.
[00:59:55] I would actually prefer that.
[00:59:57] Yeah, I would actually prefer that because it's minutes.
[01:00:00] Let me finish it off that one and then we can have it
[01:00:03] to end it.
[01:00:03] The downside to that is let's say if someone is stuck
[01:00:07] at some point and cannot move forward, then he misses out.
[01:00:11] So what I would suggest is because we are 50 minutes in keep
[01:00:15] this automation on the side.
[01:00:17] Let's focus on the other ones.
[01:00:18] Okay.
[01:00:19] See if you have similar problems occurring again, if
[01:00:22] not, then do the ones we are currently doing then we will
[01:00:25] debug.
[01:00:26] Okay.
[01:00:26] Because debugging does take time.
[01:00:29] Okay.
[01:00:31] So this automation work properly and how I verified it
[01:00:35] was I executed and I got a test answer and a test output.
[01:00:40] We will connect it to pages later for now.
[01:00:44] Just one question for you and showing as an open AI chat
[01:00:48] module for me showing is Germany only.
[01:00:50] Is that okay for me?
[01:00:52] Yeah, that is that is okay because like I said, AI is the
[01:00:56] AI that is running in the background is non-deterministic
[01:00:58] like any other AI right?
[01:01:00] It feels like different nodes for different people.
[01:01:03] Okay.
[01:01:04] Okay.
[01:01:05] So let's go to the plus icon on the top left.
[01:01:09] Okay.
[01:01:09] Let's click on new workflow and let's go with build with AI.
[01:01:20] We are again greeted with the empty space.
[01:01:23] Okay.
[01:01:24] This time what we will do is we will create a finance scanner.
[01:01:28] Okay.
[01:01:29] This will take two things.
[01:01:31] One is a list of rules.
[01:01:33] Okay.
[01:01:34] And there we will list all of the type of expenses.
[01:01:37] We will allow the employees to get reimbursement for for
[01:01:42] example, a Claude subscription and the things that we will
[01:01:45] not allow the employees to get reimbursed for for example,
[01:01:48] their personal party expenses.
[01:01:50] Okay.
[01:01:51] Also these expenses, the verified ones will get stored in
[01:01:55] an Excel sheet.
[01:01:56] So let's make that I want to build an employee re
[01:02:05] reimbursement scanner.
[01:02:12] I want employees to upload their expense PDFs onto a form.
[01:02:25] Again, we will go with the form and then connected with a
[01:02:27] website or something.
[01:02:29] This form will be scanned.
[01:02:37] Okay.
[01:02:40] And the following data will be taken out of it.
[01:02:48] We will find out the expense name, the expense date and the
[01:02:56] expense amount obviously.
[01:03:01] Okay.
[01:03:02] Then we have a Google Doc that has a list of rules for expenses
[01:03:15] that are allowed and expenses that are not allowed to be
[01:03:26] reimbursed for.
[01:03:34] Then we will store each expense in a Google sheet with the
[01:03:46] following columns.
[01:03:49] One is expense name, expense date, expense amount and status
[01:04:05] status and reason.
[01:04:11] Okay.
[01:04:13] We will also mention what status and reason has status
[01:04:18] volume will have the will have weather the expense was approved
[01:04:30] or not reason will have why it was approved or denied.
[01:04:41] Okay.
[01:04:42] Now let it run.
[01:04:44] I will copy the same prompt and paste it in the Google Doc.
[01:04:53] Let's give it a couple of minutes.
[01:04:55] Okay.
[01:04:56] Then what we will do is it will be rather fun to connect
[01:05:01] it with a live webpage and see whether it works with the
[01:05:06] webpage or not.
[01:05:07] Okay.
[01:05:54] So from one suggestion is there one diagram which you have
[01:05:59] prepared.
[01:05:59] Can you copy paste that into prompt section also?
[01:06:02] It will be easier to screenshot of this understanding the
[01:06:07] design.
[01:06:08] Okay.
[01:06:09] Okay.
[01:06:09] Sure.
[01:06:09] Why not?
[01:06:11] Let me take a screenshot of this.
[01:06:14] Then paste this.
[01:06:22] Can you please share this URL of prompt as well?
[01:06:27] URL of what prompt?
[01:06:29] Yeah.
[01:06:30] This is already right in the meeting chart.
[01:06:32] I am not able to find that.
[01:06:34] Okay.
[01:06:35] No issues.
[01:06:36] There you go.
[01:06:38] Thank you.
[01:06:56] I will have one more question also like the last session I
[01:07:00] asked about SAP automations with NA8.
[01:07:05] So does it handle the complex automations also or it is the
[01:07:10] lightweight only automations related to workflows in SAP?
[01:07:17] Can you suggest me?
[01:07:18] I know you haven't like you are like complexity, how
[01:07:26] complex it is NA8?
[01:07:28] Just want to understand.
[01:07:30] I have not used NA8 with SAP and I do not have an idea of how
[01:07:37] complex the operations of SAP it supports.
[01:07:43] Okay.
[01:07:44] Because anyways SAP is such a difficult thing to use.
[01:07:50] I know how difficult things can get with SAP because
[01:07:54] Yeah, I know.
[01:07:54] Yeah.
[01:07:55] So I do not, I am not like enough place to answer that.
[01:08:01] If it was like an elementary app or something, I would have
[01:08:04] said, okay, you know what?
[01:08:05] NA10 seems powerful enough to handle this or that, but SAP
[01:08:10] is a rabbit hole.
[01:08:11] Okay.
[01:08:14] Okay.
[01:08:20] So right this is following a new workflow.
[01:08:22] So earlier when we did connected Google Form and able
[01:08:26] to get the response and now we are trying a new workflow
[01:08:29] altogether.
[01:08:29] Is that right?
[01:08:30] This is a new workflow.
[01:08:31] Yes.
[01:08:32] Okay.
[01:08:34] Dog file in this case would remain same.
[01:08:38] I'm sorry dog file for the would remain same in this
[01:08:42] case and dog files of what?
[01:08:46] For the PDF file, HR policy document.
[01:08:50] We are not going to use the HR policy dog for this.
[01:08:55] We will use like a different dog.
[01:08:58] Okay.
[01:08:59] Let, let me proceed with the rest because otherwise we will
[01:09:02] run out of time at the end.
[01:09:04] Okay.
[01:09:05] So the first thing here is you have these rectangular boxes.
[01:09:10] This is nothing but an abstraction to show you like
[01:09:13] a simpler interface.
[01:09:15] If you just click on it, you will see that it falls
[01:09:18] into multiple nodes.
[01:09:19] Okay.
[01:09:20] So one has an extract PDF text thing.
[01:09:23] Okay.
[01:09:24] Then it hits the reimbursement rules.
[01:09:27] Then it analyzes some text.
[01:09:30] Then it saves it to Google sheet.
[01:09:32] Now I have a problem with this particular arrangement because
[01:09:39] this node that you see that is supposed to extract data from
[01:09:43] the PDF might actually not work properly.
[01:09:48] Okay.
[01:09:48] Because this is an innate and inbuilt node.
[01:09:51] I would rather pass it past the PDF file entirely to
[01:09:57] open AI.
[01:09:58] Okay.
[01:09:59] But let me not fidget with it right now.
[01:10:02] Let me see it works by default.
[01:10:04] If it does, then I will not change it.
[01:10:06] Okay.
[01:10:06] I'll not fix what is not broken.
[01:10:09] So first of all, my account is already created.
[01:10:12] Okay.
[01:10:13] And I will find the reimbursement rules.
[01:10:18] I think I have a document for that already created.
[01:10:25] I'll copy this.
[01:10:27] I'll create another Google Doc.
[01:10:29] I'll share you the link.
[01:10:44] So Roy, you have this link and you're, uh, you're
[01:10:47] patching this link and that, uh, and the questionnaire, right?
[01:10:53] Yeah, I am going to paste this link in any 10.
[01:10:59] Yeah.
[01:11:00] So what about like if we have a multiple documents, it's
[01:11:03] supposed to validate before you can find out you can
[01:11:08] find out and you can have multiple inputs in the form or
[01:11:11] you can connect it with multiple Google Docs.
[01:11:16] You simply have to ask this.
[01:11:17] Yeah, it will lay out all the notes for you.
[01:11:20] Okay.
[01:11:20] Unless you have like, like I was talking about a hundred
[01:11:24] gigabytes of data, this arrangement will work for it.
[01:11:29] Okay.
[01:11:30] Is it expecting a PDF and we are giving a document?
[01:11:35] Word doc?
[01:11:37] See the bill will be a PDF.
[01:11:40] The reimbursement rules will be a word doc.
[01:11:44] Okay.
[01:11:44] So these are reimbursement rules.
[01:11:47] So let me paste the link for this.
[01:11:50] Roy can, uh, is it is work on a live doc file?
[01:11:55] Means we can change any, uh, uh, reimbursement in a doc.
[01:12:01] You can add or remove rules and it will get updated
[01:12:05] automatically.
[01:12:06] So now this is where I need to configure Google sheets.
[01:12:11] So again, I will sign in.
[01:12:13] Okay.
[01:12:13] It will ask me for access.
[01:12:21] It has connected my Google sheet account as well.
[01:12:24] Now what I need to do is save the Google connected with
[01:12:27] the Google sheet.
[01:12:29] So let me go to sheets.
[01:12:33] Uh, Google doc.
[01:12:34] Sorry.
[01:12:38] Let me then travel to Google sheets.
[01:12:43] Then I have a sheet called reimbursements already prepared.
[01:12:48] Okay.
[01:12:49] Let me delete this.
[01:12:51] Let me delete.
[01:12:52] Okay.
[01:12:52] Let's let me use a new sheet actually.
[01:13:10] Am I having a network issue?
[01:13:12] Uh, can you guys hear me?
[01:13:14] Yes.
[01:13:15] Yeah.
[01:13:16] Yes.
[01:13:16] Yes.
[01:13:17] Not sure why it was not loading.
[01:13:20] Nevertheless, let's call it expense sheet inside it.
[01:13:36] Let's give it the same columns that we were talking about.
[01:13:39] First is expense name.
[01:13:43] Next is expense date.
[01:13:47] Now this step is slightly tricky because you need to enter
[01:13:52] the same columns, column names in the same order.
[01:13:56] Otherwise the AI is going to be very finicky about this
[01:14:00] and cause a lot of errors.
[01:14:02] Okay.
[01:14:05] I know this because I have just used it for a bunch of times.
[01:14:11] So I pasted the exact column names in the same order called
[01:14:20] it expense sheet.
[01:14:22] Now let's go back and select it from the list.
[01:14:26] So there is a error like he is missing in column B expense
[01:14:32] sheet.
[01:14:34] Okay.
[01:14:35] Let me fix that.
[01:14:37] Uh, will it match the column name or the column sequence?
[01:14:43] It will match the column name.
[01:14:45] Okay.
[01:14:46] And if it doesn't match the column name now it creates a
[01:14:50] bunch of problems like for example after connecting if I
[01:14:54] change the column names here and I tell the AI hey, you
[01:14:57] know what I change the column names.
[01:14:59] It's going to fumble like five times before it gets it
[01:15:02] properly.
[01:15:04] Okay.
[01:15:05] So I got a document form list and sheet form list in which one
[01:15:11] I need to basically add this sheet.
[01:15:15] Document form list and sheet form list.
[01:15:18] Yeah.
[01:15:19] What do you mean?
[01:15:20] You share your screen.
[01:15:21] Okay.
[01:15:22] Let me share.
[01:15:23] Mm hmm.
[01:15:24] Is it visible?
[01:15:33] Please give that Excel sheet even here.
[01:15:36] Here I have two options document form list and sheet list.
[01:15:40] Oh, okay.
[01:15:41] It's easy.
[01:15:42] Uh, basically see one Google sheet can have multiple sheets
[01:15:47] inside it.
[01:15:48] One Google sheet document can have multiple sheets.
[01:15:51] Right.
[01:15:52] Sheet one sheet two sheet three.
[01:15:54] So if you you just get to select sheet one, that's it.
[01:15:58] Yeah.
[01:16:00] Okay.
[01:16:01] I yeah.
[01:16:08] That's a by the way.
[01:16:10] Don't copy my sheet.
[01:16:12] Okay.
[01:16:12] For the person who was asking me for the sheet link because
[01:16:16] then your automation will send data to my sheet.
[01:16:20] Okay.
[01:16:21] Plus obviously you will not be able to because you don't
[01:16:25] have access to this.
[01:16:26] So going back, let me try whether this entire automation
[01:16:37] works and for that I need a bill.
[01:16:40] Okay.
[01:16:40] Let me see if I have a bill.
[01:16:42] I should have a Claude bill or and throw pick.
[01:16:53] Uh, as a operation, we have to select append and update
[01:17:21] rules.
[01:17:21] Correct.
[01:17:22] Yeah.
[01:17:23] Go with append.
[01:17:24] Append looks like I can't find a bill.
[01:17:34] So let me just download one.
[01:17:36] Okay.
[01:17:54] I will download this and share it with you.
[01:17:57] Okay.
[01:18:29] So this is my Claude bill.
[01:18:30] Let me try if this works.
[01:18:42] I'll submit Dean voice.
[01:18:52] I'll see if it works and see it said it has been
[01:19:02] denied personal expenses are not reimbursable as this
[01:19:06] subscription does not serve a legitimate business purpose.
[01:19:10] Now this means that something with the rules is wrong.
[01:19:14] Okay.
[01:19:15] Personal expenses will not be reimbursed, but let's
[01:19:22] say AI subscriptions are okay.
[01:19:29] Now let me try again.
[01:19:37] Let me delete this row of data.
[01:19:39] Let me run this again.
[01:19:57] This time it got approved and it detected the expense
[01:20:00] name, the date, the amount, the status and the reason.
[01:20:05] Okay.
[01:20:07] So this is how we got it connected with a form.
[01:20:11] Have you updated the document?
[01:20:14] I mean, which document?
[01:20:17] I mean that expenses was not coming right because
[01:20:22] while it's shuffling the screen I missed like how come the
[01:20:25] expense came correctly now?
[01:20:27] What is that?
[01:20:29] I just added this but AI subscriptions are okay.
[01:20:35] Okay.
[01:20:37] Yeah, because it was getting denied.
[01:20:39] So it was being nitpicky about personal expenses.
[01:20:44] Thank you.
[01:20:49] Yes, sir.
[01:20:51] Yes.
[01:20:53] I mean, I'm saying
[01:21:03] I'm saying
[01:21:09] the approval gets auto generated by the LLM.
[01:21:12] Okay.
[01:21:16] Yes.
[01:21:19] Can you show me one second?
[01:21:23] Yeah, we already did.
[01:21:26] Here you got on the left side of prompt and if you missed out
[01:21:32] on the prompt you can perhaps click on this.
[01:21:35] You can see this Google Sheets account and under that you can
[01:21:40] click an account and configure it with it.
[01:21:44] Okay.
[01:21:49] Also just for your, for everyone's information, look at this
[01:21:52] panel. Okay.
[01:21:54] It gets
[01:21:57] I'm sorry.
[01:22:00] Was the meeting manager saying something I could not hear you.
[01:22:13] Professor, we know that we have all the LLM's back so I'll just
[01:22:18] give them 10 minutes to go message.
[01:22:21] Thank you.
[01:22:29] I did not understand the thing of what you said.
[01:22:32] I think he is in two sessions maybe.
[01:22:35] Okay.
[01:22:37] Okay. Nevertheless, you can click on this Google Sheets icon
[01:22:43] and you can select an account.
[01:22:46] Also, what I do have is a test case for the other situation.
[01:22:52] This one.
[01:22:55] Are you saying something because we thought we heard something.
[01:23:00] I think.
[01:23:08] I think we thought 10 minutes to go message and share with them.
[01:23:12] If you need anything, Professor, I'll be your test.
[01:23:15] If I need anything, just type in a message I'll be able to see.
[01:23:19] And I will reach out to you.
[01:23:21] Professor.
[01:23:23] Thank you so much, sir.
[01:23:25] Okay. No issues.
[01:23:35] So Roy actually for me, what manager back now?
[01:23:40] Roy is Nishant.
[01:23:42] Everything runs well.
[01:23:45] But the thing is that the Excel is not affected with the columns.
[01:23:49] It got a JSON.
[01:23:51] So actually in the mapper I picked auto column mapping.
[01:23:55] So I need to give the name of the columns.
[01:23:58] What I would suggest is you must be getting some error, right?
[01:24:02] No, it is no error.
[01:24:04] I mean it worked very well.
[01:24:06] But the only thing is that in Excel instead of splitting the AI
[01:24:09] that I would put JSON into columns,
[01:24:12] I just pasted the whole JSON there in my seat.
[01:24:15] So I think the splitting didn't happen.
[01:24:17] So is there any mapping you've done in your Excel?
[01:24:20] I mean in your workflow?
[01:24:22] I didn't do any explicit mapping.
[01:24:25] The AI that set the nodes did what it did.
[01:24:29] I would suggest take a screenshot.
[01:24:32] Okay, share your screen.
[01:24:34] Take a screenshot, paste it in N8N.
[01:24:37] And let it run. It will auto fix things.
[01:24:39] Okay, let me do that.
[01:24:41] Thank you. Please proceed.
[01:24:45] Yes. Okay.
[01:24:52] Now there are a couple of more things that one needs to know.
[01:24:56] First is if you this zoom bar is very annoying.
[01:25:02] Okay, nevertheless,
[01:25:06] first thing you need to see here is there is a test.
[01:25:12] Okay, in the form.
[01:25:14] Okay, there's a test URL.
[01:25:16] And there's a production URL,
[01:25:18] which means that for testing this URL will be active for actual use cases.
[01:25:26] This URL will be active.
[01:25:28] Now if I publish this,
[01:25:31] it goes live on the internet.
[01:25:34] Okay.
[01:25:36] And the place it is live at is this URL.
[01:25:40] So what I can do is I can paste this for now.
[01:25:44] Okay.
[01:25:46] Then I can let's say go enter.
[01:25:49] And I will see that this app of mine,
[01:25:52] this mini expense tracker app of mine is live now.
[01:25:55] Yes, Abhishek, tell me.
[01:25:57] I'm not related to what you're saying.
[01:26:00] So just finish it then I'll ask.
[01:26:02] Sorry.
[01:26:05] I just had a question if I have to change the link to the documents that I've provided already.
[01:26:10] Now I want to add a new document or replace it.
[01:26:13] How can I do that in this view?
[01:26:16] Or just click on the note.
[01:26:18] You will find a place where the link for the document has been set.
[01:26:21] Just have to change it there.
[01:26:23] Click on the where node.
[01:26:25] Click on this node.
[01:26:27] This node will open the link for the Google Doc.
[01:26:30] This will open the one for the Google sheet.
[01:26:32] Okay.
[01:26:34] Thanks.
[01:26:36] So what we saw is this published button makes things available for the public internet.
[01:26:41] So you guys can share the link and see if each other's automations are also working.
[01:26:46] Next, what we will do is see I have a small app that we that I had built on lovable.
[01:26:56] Okay.
[01:26:58] And this can be anything.
[01:27:00] Okay.
[01:27:02] This can be let's say your company's website.
[01:27:04] Okay.
[01:27:06] It can be let's say an AI assistant.
[01:27:08] It can be a small chatbot or something that collects forms.
[01:27:14] Okay.
[01:27:15] Collects people's emails.
[01:27:17] So I have something very similar designed here.
[01:27:21] Which room were you in?
[01:27:23] I'm sorry.
[01:27:27] I believe you just trying or you don't have to know.
[01:27:30] Do you want me to add you to any room?
[01:27:33] Whoever is the manager.
[01:27:35] Can you mute yourself?
[01:27:37] I think you're talking about two sessions.
[01:27:39] He's troubling us repeatedly.
[01:27:44] He's just not aware that his mic is unmuted.
[01:27:47] No issues.
[01:27:50] Okay.
[01:27:51] Guys, come back to this.
[01:27:53] I'll share you the link to this project so that you can copy this entire thing.
[01:27:57] Okay.
[01:27:58] What we will do is.
[01:28:01] Yes.
[01:28:03] Salman.
[01:28:05] Salman, your mic is unmuted.
[01:28:07] Okay.
[01:28:08] Someone just.
[01:28:10] The manager.
[01:28:12] Could you check whether you have the access for that?
[01:28:15] Yes.
[01:28:18] Okay.
[01:28:24] Nevertheless, let's come back.
[01:28:26] Okay.
[01:28:27] So guys, what I have here is a lovable page.
[01:28:30] Okay.
[01:28:31] I hope everyone knows by now how to create sites on lovable.
[01:28:35] If not, let's not worry about this.
[01:28:37] The point is your company, your personal webpage, whatever it is,
[01:28:42] will have some kind of a landing page.
[01:28:44] Right.
[01:28:45] Let me share this with everyone.
[01:28:48] Copy in white link.
[01:28:51] And there you go.
[01:29:01] Hopefully you can see this.
[01:29:03] If you come here, you can copy this.
[01:29:06] And what we will do is we will integrate this form.
[01:29:13] With this expense sheet.
[01:29:15] Okay.
[01:29:16] This is a dummy use case.
[01:29:18] We will simply modify this for our use case.
[01:29:21] Or let me see what we can do is this, this page also has an HR policy bot.
[01:29:27] Okay.
[01:29:28] That I had created last time.
[01:29:31] This does not work right now.
[01:29:35] Okay.
[01:29:38] This is a static page.
[01:29:39] We will connect our HR policy bot with this page.
[01:29:44] So what did we interrupt?
[01:29:46] Actually the link you shared it shows that the link that project is private.
[01:29:51] So is it like we need to sign into the lovable?
[01:29:55] How can it be private on a free account?
[01:29:58] For me it worked.
[01:30:00] It says access.
[01:30:01] You still need to publish it.
[01:30:03] You are not supposed to sign in.
[01:30:07] Who is this speaking?
[01:30:11] For me also it is saying private.
[01:30:16] For me it's nothing.
[01:30:20] Chani are you?
[01:30:25] Yeah.
[01:30:26] Shall I share my screen and show you?
[01:30:29] Yeah sure.
[01:30:31] Actually we need to sign in to Google account and then accept the invitation.
[01:30:41] If you have already used the lovable then only you will be able to say.
[01:30:47] Sign in or go to lovable and you can from Google account you can log in.
[01:30:54] Okay.
[01:30:55] So guys that is resolved.
[01:30:58] Okay.
[01:30:59] Take two minutes.
[01:31:00] Sign in.
[01:31:01] Okay.
[01:31:02] Okay.
[01:31:03] Okay.
[01:31:04] Okay.
[01:31:05] I will.
[01:31:06] Sign in.
[01:31:07] Okay.
[01:31:13] Superb.
[01:31:14] Just a quick question.
[01:31:16] So the N8N workflow that you taught us right now.
[01:31:20] So how much concurrency can it handle?
[01:31:23] I mean if there are like for like internal company like HR policy like there are 1000 users
[01:31:29] or 2000 users.
[01:31:31] But if you want to roll it out for a wider audience, probably one million people and executing, you know, this workflow simultaneously, will it be able to handle it?
[01:31:42] If you're talking about a million people, it requires very different kind of an architecture. Okay.
[01:31:50] So let's go about it in this way. On a free N8 in account, you can assume that it will serve, let's say 10-15 users.
[01:31:57] Okay, not more than that. If you have a pro account, it will probably support, let's say 50-60 users.
[01:32:03] For use cases that span over anything beyond a hundred users, I will not suggest to use something like N8N.
[01:32:12] Okay, not because I hate it, but because it is very fragile by the nature of it.
[01:32:17] Okay. And for use cases where it gets bombarded by a thousand users, I would say, hey, you know what, just, I mean, build it on your own.
[01:32:27] Yeah.
[01:32:28] Our use like third party tools like N8N the least. Okay.
[01:32:32] Okay.
[01:32:33] For a million users, you are going to that Uber level where even the database needs to be custom made. Okay, so.
[01:32:39] Correct.
[01:32:40] Yeah, so obviously this is not going to support anywhere close to that.
[01:32:45] Oh, okay.
[01:32:47] So what is the alternative for that? If you say that it's more than 100 users, no, you should not go for N8N.
[01:32:55] So what's an alternative?
[01:32:57] See, after a hundred users or after, let's say 500 concurrent users, the problems that appear are very of various types.
[01:33:08] You have, let's say cashing issues, you have race conditions that you have net that you had never predicted before.
[01:33:15] Okay, you have database concurrency issues, you have right issues.
[01:33:22] Okay, with let's say data getting written to the disk.
[01:33:25] You have a bottle data outflow bottleneck issues.
[01:33:30] You have, let's say if you're using any form of parallelization like Kubernetes that might create issues.
[01:33:36] Okay.
[01:33:37] So giving you a one stop solution is not possible.
[01:33:42] It's more of a fix it as an as in when it comes kind of an approach that you need to take.
[01:33:50] Okay.
[01:33:54] I hope that everyone has copied this project.
[01:33:57] We need this HR page only.
[01:33:59] Okay.
[01:34:00] There are two pages.
[01:34:01] One is the homepage.
[01:34:02] One is the HR bot page.
[01:34:03] Imagine this is like a protein bar company and they have a FAQ page.
[01:34:09] Okay.
[01:34:10] Or they have an open HR page where employees can come and ask questions.
[01:34:14] What we will do is we will go back to our HR bot.
[01:34:19] We will tweak it a bit so that it can take responses.
[01:34:23] And then we will make the same changes.
[01:34:27] We will make lovable aware of these changes.
[01:34:31] Okay.
[01:34:32] So that it gets connected from it.
[01:34:34] So let's go back.
[01:34:37] Let's go to overview.
[01:34:40] We have an HR policy bot.
[01:34:43] Okay.
[01:34:45] Let's close this panel.
[01:34:47] Let's open up the AI assistant.
[01:34:49] Now what we will do is I will say.
[01:34:55] I want to connect it with my company's landing page so that employees can ask questions directly.
[01:35:17] Sanjit, unless you're intending to ask me something, please mute yourself.
[01:35:27] Subra, I have one quick doubt.
[01:35:29] Okay.
[01:35:30] Let me just finish typing this from the then I'll tend to you.
[01:35:34] Okay.
[01:35:35] Sure.
[01:35:36] Okay.
[01:35:37] Can ask questions that directly replace the form that you have created with an API post
[01:35:54] to get so that you receive a text question and give out a text response.
[01:36:12] Also give me an open API spec of the, what is it called URL open API spec of the function
[01:36:32] sorry endpoint.
[01:36:37] So what I will do is I will open this workflow.
[01:36:41] Sorry.
[01:36:42] This workflow I will with its production and point sorry with its production URL because
[01:36:56] any 10 has two kinds of URLs.
[01:36:58] One is the testing URL.
[01:37:00] One is the production URL.
[01:37:02] I'll say, Hey, give me the open API spec with the production URL.
[01:37:05] And now if you're wondering what the open API spec is, see the front end and the back end
[01:37:12] are two completely different components.
[01:37:14] As you can see, this is one component.
[01:37:17] The bot here is another component and they need to speak the same language in order
[01:37:23] to be able to give send the data from the front end to the back end and take the
[01:37:30] response from the back end and send it back to the front end.
[01:37:33] Right.
[01:37:34] So that medium, okay, that common medium or that common language is called an open API spec.
[01:37:44] Okay.
[01:37:48] This is not open AI spec.
[01:37:50] Okay.
[01:37:51] It's not the company behind chat GPT.
[01:37:53] It's no, it's no way related.
[01:37:55] It's open API.
[01:37:57] Okay.
[01:37:59] And this is language agnostic.
[01:38:01] So it works on Python, Node.js, anything.
[01:38:04] So let's press enter.
[01:38:06] Let it make the changes and yeah, come back to me with that question.
[01:38:12] Yes.
[01:38:14] So actually, so is there any caching method in this N8N?
[01:38:19] So instead of hitting that chat GPT model again and again with the same, the same
[01:38:24] kind of questions.
[01:38:26] So is it possible?
[01:38:28] Maybe.
[01:38:31] Why don't you sort, why don't you ask the AI that and see what response it
[01:38:36] gets because it should have some kind of a caching node.
[01:38:43] Okay.
[01:38:44] A different way you can check for it is if you open a blank canvas, okay, on the right
[01:38:51] hand side, you have a panel on that.
[01:38:54] If you click on the panel, it opens up a list of all of the possible nodes you
[01:38:59] have inside that you can perhaps search for cash and see if it returns anything
[01:39:04] useful.
[01:39:05] Okay.
[01:39:06] Thanks for that.
[01:39:08] Super.
[01:39:09] Okay.
[01:39:10] We'll explore after we get done with this.
[01:39:13] Yeah, sure.
[01:39:20] Can you share the prompt?
[01:39:22] Yeah, sure.
[01:39:23] Let me paste it.
[01:39:32] So yeah, I've pasted it.
[01:39:43] Okay.
[01:39:44] So what do we have here?
[01:39:46] You see the prompt got replaced by what they call a webhook.
[01:39:51] A webhook is like this thing hanging on the internet that when clicked.
[01:39:59] Okay.
[01:40:00] It's like a switch in the internet that when clicked activates a workflow.
[01:40:04] Okay.
[01:40:08] It basically is a link that gets accessed.
[01:40:13] You can send some data or you can receive something from it.
[01:40:17] Okay.
[01:40:18] And that is what we replaced the form with.
[01:40:22] Now we have been giving the given the production URL as well.
[01:40:27] And what it has given us is the open API spec.
[01:40:30] So let's copy the open API spec here before copy and pasting it.
[01:40:45] Let's understand.
[01:40:47] It says there's a version number.
[01:40:50] Good.
[01:40:51] There's a title called HR policy bot API description is ask a natural language question about
[01:40:58] the company and get an answer reasonable.
[01:41:01] The URL is also mentioned.
[01:41:04] Okay.
[01:41:05] The URL also has a description.
[01:41:07] The path is slash HR policy bot.
[01:41:11] As you can see, it is the last part of the link.
[01:41:14] Okay.
[01:41:16] Operation ID has something description description.
[01:41:20] Okay.
[01:41:21] But the important part is content under content.
[01:41:26] It says send me data in the form of a JSON.
[01:41:30] Okay.
[01:41:32] Inside that you have a question which is the employees question as an example.
[01:41:40] How many vacation get do I get to take response has particular codes 200 for okay 500 for error.
[01:41:47] Okay.
[01:41:49] The description is answer in plain text.
[01:41:52] There's an example as well.
[01:41:54] So it communicates the exact format in which the data will be exchanged.
[01:42:00] So I will open the page.
[01:42:04] I will say I want to connect the page slash HR bot with its back end.
[01:42:24] Okay.
[01:42:25] The open API spec for the back ends as methods is this and I paste the entire specification.
[01:42:40] That's all.
[01:42:42] But I need to make sure that it does not make any other fancy changes because lovable tends to, you know, do things extra.
[01:42:51] Okay.
[01:42:52] Please don't change.
[01:42:56] Please don't make any other changes than just connecting with the HR bot via the URL.
[01:43:17] Okay.
[01:43:18] Let's paste it while lovable is working.
[01:43:30] Let's publish the latest changes.
[01:43:34] See if it works, then great.
[01:43:44] If it doesn't work, there's a tool called called okay, which is available across Linux windows and let's say Mac.
[01:43:54] Okay.
[01:43:55] We will use Linux.
[01:43:56] Sorry.
[01:43:57] We will use called to test whether any of these parts are working or broken.
[01:44:01] Okay.
[01:44:02] But we will only explore that once it's broken.
[01:44:06] Okay.
[01:44:07] It says it's connected now.
[01:44:09] Let's try asking a question.
[01:44:11] Sir, please reiterate from lovable and what's where you pasted lovable this call.
[01:44:18] Yeah, sure.
[01:44:20] So basically see what we did was we had a form input, right?
[01:44:26] We were entering the questions in a form.
[01:44:29] So we need to link that to our website.
[01:44:32] Yeah.
[01:44:33] So what we did was we removed the form and we took what is known as a webhook.
[01:44:38] Okay.
[01:44:39] A webhook is like a switch on the internet that gets triggered by anything that we wanted to.
[01:44:45] Okay.
[01:44:46] In this case, it's our landing page.
[01:44:47] Okay.
[01:44:48] Yes.
[01:44:49] So we got a description of the car of the format in which the switch receives all that
[01:44:55] data.
[01:44:56] Okay.
[01:44:57] And also returns its output.
[01:44:59] Now that language is called open API spec.
[01:45:03] So I copy the code.
[01:45:05] Correct.
[01:45:06] Yeah.
[01:45:07] So then I got the code.
[01:45:10] Yeah, how to go to this lovable website according to each.
[01:45:17] Have you created pages with lovable before?
[01:45:20] No.
[01:45:24] Has this entire cohort not done that?
[01:45:28] No, we don't.
[01:45:30] We can use your one right now, right?
[01:45:35] Because your one is working for most of us.
[01:45:38] The one you shared.
[01:45:40] It was not part of any of the sessions.
[01:45:43] Some of us we are aware about it.
[01:45:45] We know about it and some of us we have worked on that.
[01:45:48] Okay.
[01:45:49] Okay.
[01:45:50] I see.
[01:45:51] See on a project lovable converts text into websites.
[01:45:56] Okay.
[01:45:57] For your reference on the left hand side, you have a panel.
[01:46:00] You can connect.
[01:46:01] You can activate that by one of these icons here.
[01:46:04] Let me see.
[01:46:05] Yeah.
[01:46:06] This is perhaps the icon.
[01:46:08] This slider panel.
[01:46:09] This.
[01:46:10] Okay.
[01:46:11] You say, yeah.
[01:46:13] So we can use like a yeah.
[01:46:16] Okay.
[01:46:17] Let's see if it's working.
[01:46:20] How many casual leaves can I take in a year?
[01:46:30] It may work.
[01:46:32] It may not work.
[01:46:33] Okay.
[01:46:40] It says I can take up to six paid or personal casual leaves.
[01:46:44] Amazing.
[01:46:46] Sorry to cut you.
[01:46:49] Yes.
[01:46:50] Even I never used lovable.
[01:46:53] So how can because I'm slightly back.
[01:46:57] I'm sitting behind doing this activity.
[01:47:00] So if I go for the recording and do it, so how can I create a lovable page and then
[01:47:05] tag it or I can use the link which you have shared it.
[01:47:08] See for now.
[01:47:10] Okay.
[01:47:11] For now, just use the link.
[01:47:13] Okay.
[01:47:14] So if you want to change it just prompt, but prompt efficiently because I'm sure everyone
[01:47:20] has gone through prompt engineering or how to use prompts properly.
[01:47:25] So just give a casual, just give a decent prompt.
[01:47:28] It will work.
[01:47:30] You mean prompting the current page, right?
[01:47:33] Not going into the lovable website.
[01:47:35] Right.
[01:47:36] That's what you're seeing.
[01:47:37] You can also prompt this page to change this page.
[01:47:42] If you want to create a new page of your own, then just go to lovable login, then prompted
[01:47:48] here it will create your new page.
[01:47:51] You can do the same thing on a new page as well.
[01:47:55] Understood.
[01:47:56] Thank you.
[01:47:57] I just fed you up by giving you an automatic already ready page.
[01:48:02] Got it.
[01:48:03] Thank you.
[01:48:05] Yes, through viewer also asking me something I guess.
[01:48:09] Actually, after the step, when we have prompted it to give the URLs for get and that after that,
[01:48:17] can you give a summary of the steps that we have taken up till now after we have given
[01:48:24] the URL for what like we have given the prompt to change it to URL.
[01:48:30] Like API URLs.
[01:48:32] Like if you scroll up in the AI chat.
[01:48:38] Yeah.
[01:48:41] About a little bit up.
[01:48:44] Where we have copied the curl and things like that.
[01:48:48] We did not copy the curl.
[01:48:50] We just pasted this open API response.
[01:48:56] This was the prompt I gave.
[01:48:58] No, before this before this after that, like what all things we did after after giving
[01:49:06] the initial prompt, we did absolutely nothing.
[01:49:09] Okay.
[01:49:10] I have no idea why you guys are.
[01:49:13] You know, continuously giving me giving my application questions.
[01:49:20] Yeah, the project.
[01:49:22] My project.
[01:49:24] Yeah.
[01:49:26] I'm going to remove everyone's is amazing.
[01:49:32] Fine.
[01:49:34] Go ahead.
[01:49:36] I cannot even remove edit access.
[01:49:39] So quick question like actually I'm trying to hit through the postman.
[01:49:45] Yeah, I'm getting a 200 success, but I couldn't able to see any.
[01:49:50] Like a response from that.
[01:49:52] There goes someone else.
[01:49:57] Who's this guy's superb tech.
[01:50:00] I just said, yes, you tried hitting it with postman.
[01:50:07] What was the error?
[01:50:09] No, it's not giving any error.
[01:50:11] It shows as 200 success, but the thing is I couldn't able to get any response from that.
[01:50:17] Okay.
[01:50:18] If you're not getting any response, it means something's wrong with the backend.
[01:50:22] Yeah, actually I published that hook also sorry that web book, but I couldn't able to get any response.
[01:50:28] So do this make a simple call request.
[01:50:31] Okay.
[01:50:32] See whether you're still not getting the response or not.
[01:50:36] And if you're not getting then copy the entire response of this and paste it to the AI inside any 10.
[01:50:44] Let me do it along with you.
[01:50:46] So for example, this is the open API.
[01:50:51] Right.
[01:50:52] Let me see.
[01:50:53] Yes, this is the link.
[01:50:54] Let me go to charge.
[01:50:56] Let me paste the open AI open API as specification.
[01:51:01] And I will say give me a call command for this.
[01:51:10] So it gave me an example.
[01:51:15] Let's say question that I can ask.
[01:51:17] Right.
[01:51:18] I copied this.
[01:51:20] Now if I want to make sure that my backend is working properly, I will open a terminal.
[01:51:26] I'll paste it.
[01:51:27] I'll press enter.
[01:51:28] I got a response.
[01:51:36] Right.
[01:51:37] You are perhaps not getting a response at this point.
[01:51:41] So copy this entire thing.
[01:51:43] Okay.
[01:51:44] Then go to any 10, then paste it, then press enter and tell, Hey, you know what?
[01:51:52] I your backend is not working properly because call is not returning me the response.
[01:51:57] That's it.
[01:51:58] Okay.
[01:51:59] I will try that.
[01:52:00] Yeah, I will try with it.
[01:52:05] Okay.
[01:52:06] Who all would not connect lovable with this?
[01:52:13] Because I know this pro this step is slightly problematic.
[01:52:18] It will not work for everyone.
[01:52:20] We have 15, 20 minutes.
[01:52:23] Let's debug other people's issues now.
[01:52:26] Yes, through tell me.
[01:52:27] Can we get the recording?
[01:52:30] That's for the manager by default.
[01:52:32] You do get the recording.
[01:52:34] The courses coming classes are done means if someone is meeting new is going to happen or
[01:52:40] we are covered everything what you are targeting.
[01:52:43] See, we can get into the we can get into more automations.
[01:52:47] Okay.
[01:52:48] But it will only over complicate things.
[01:52:51] If you guys don't have any doubts, then I will proceed with other automations.
[01:52:57] Hello.
[01:52:58] So, yeah, so I was asking about the editing steps, not the website one before.
[01:53:04] When I asked, you know, can you just open anything again?
[01:53:09] Yeah, sure.
[01:53:10] Yeah.
[01:53:16] So here, like before, no, we were using a form.
[01:53:20] Then we gave it to, you know, get the URL from like using our APIs to get and give the input
[01:53:30] output in the chat.
[01:53:34] Can you just scroll up?
[01:53:36] Like where we have given the prompt.
[01:53:38] Yeah, yeah, this one.
[01:53:42] I pasted this prompt.
[01:53:44] I think in the dog.
[01:53:45] Yes, yes, you have already pasted it in the dog.
[01:53:48] But after that, the steps, I was not able to get through it.
[01:53:51] So can you explain them again?
[01:53:53] I did nothing after that it all of it happened automatically only I pasted this prompt.
[01:53:59] It gave me the open API spec.
[01:54:01] I pasted the open API spec in loveable.
[01:54:04] It automatically configured everything.
[01:54:07] And then I tested it by giving it a question.
[01:54:09] That's it.
[01:54:10] Okay, so we went to loveable to after this.
[01:54:13] Okay.
[01:54:14] Yes.
[01:54:15] Yes, you're going to tell me.
[01:54:19] Yes.
[01:54:20] So in this and it, it chose the model.
[01:54:25] I think it chose open API.
[01:54:28] Sorry.
[01:54:29] The chat GPT model.
[01:54:33] But if say, for example, I have a cloud subscription and I wanted to use cloud subscription.
[01:54:39] Is there a way we can connect cloud to use?
[01:54:43] See, this is a programmatic way of accessing your AI model.
[01:54:49] It will not use your $20 cloud subscription.
[01:54:53] It was a cloud API key.
[01:54:56] Yeah, okay.
[01:54:58] So if you have an API key, you can prompt the AI to, you know,
[01:55:03] Oh, I see.
[01:55:05] Okay.
[01:55:06] Yeah, it will not use the subscription.
[01:55:09] Okay.
[01:55:11] Anyone else anything and guys we have what 15 minutes.
[01:55:16] Do you want to cover another automation or do you want to have a doubt clearing session?
[01:55:22] Yeah, go for other automation.
[01:55:25] Text in the group so that I can get like a pool.
[01:55:41] Sir, we need like this section slide section.
[01:55:44] Okay.
[01:55:48] Let's do this.
[01:55:52] What was the second automation we had created?
[01:55:55] Let's see if we employee reimbursement scanner.
[01:55:59] What we can try is let's say we tried the webpage HR policy recruitment screener is very similar.
[01:56:13] Okay.
[01:56:14] I don't want to repeat it because it will connect with calendar Google calendar.
[01:56:18] There will be one node difference.
[01:56:20] Okay.
[01:56:21] Finance we have tried.
[01:56:23] If we can do an hours we can connect.
[01:56:27] Let's say the HR policy bought to telegram.
[01:56:30] Do you want to do that?
[01:56:40] Okay.
[01:56:41] Let's let's connect it with telegram.
[01:56:42] It should not take a lot of time as well.
[01:56:48] Let me just open telegram once and see if I have telegram in the first place.
[01:56:57] Just give me one second.
[01:57:21] Okay.
[01:57:26] So what it was I logged into telegram web.
[01:57:29] Okay.
[01:57:30] You can also use the mobile telegram does not matter.
[01:57:33] It's the same thing.
[01:57:38] Inside this, you need to go to this account called bought father.
[01:57:43] Okay.
[01:57:44] This is mandatory and it is called at the rate bought father.
[01:57:49] Yes.
[01:57:57] Here is the username.
[01:58:00] This bought father account allows you to create new bots inside telegram or telegram bot is
[01:58:09] necessary for connecting it with telegram obviously.
[01:58:14] So once you have connected to telegram, once you are at this page, let me know and just
[01:58:23] drop like a done or something in the group so that I can keep a track of the progress.
[01:58:45] Can you please come again like what are the significance of bought father this bought father
[01:58:51] is an internal telegram tool that allows you to create bots.
[01:58:56] Okay.
[01:58:57] Thank you.
[01:58:58] I am coming to the end here.
[01:59:10] So why do we have to create and what is the purpose we have to create multiple bots using
[01:59:16] what for that?
[01:59:17] We need to we don't need to create multiple bots.
[01:59:20] We need to create one bot so that that bot becomes the connection to our and it and see and it
[01:59:31] and is the glue right but and it sorry telegram also needs to receive right so it needs the
[01:59:40] request and it needs to authorize the requests and it needs to check and verify whether an
[01:59:48] incoming request to your account or to my account is a valid request.
[01:59:53] Right.
[01:59:54] So the way we do that is by creating what is called a bot.
[01:59:58] A bot is nothing but something something like any 10 only and you can use it for multiple
[02:00:06] purposes.
[02:00:07] You can define custom logic inside it as well.
[02:00:11] Okay.
[02:00:12] But here we will simply use it as a head that receives an output state and it will receive
[02:00:21] data from our phone or let's say this web panel and it will send it over to any 10 and it
[02:00:28] will get the response from any 10 and paste it in telegram.
[02:00:31] That's it.
[02:00:32] Do we need to sign up?
[02:00:36] Obviously, the reason we are working with telegram as compared to what we have to create
[02:00:51] is that telegram as compared to WhatsApp is doing the same thing and WhatsApp is very
[02:00:57] difficult.
[02:00:58] Okay.
[02:00:59] You need to get yourself a business account.
[02:01:01] Then you need to verify your business.
[02:01:03] Then there's a KYC step.
[02:01:05] Then you need like a predefined message starters and there's like a lot of
[02:01:11] complication.
[02:01:12] Telegram is something that takes 15 minutes to get started.
[02:01:17] Yes, Aditya.
[02:01:18] Tell me instead of telegram.
[02:01:20] If I'm doing it from an organization perspective where you cannot use telegram or
[02:01:27] WhatsApp or any of these things, we can also use Slack or Teams also, right?
[02:01:34] Yeah, we can connect it as a Slack bot as well.
[02:01:38] Okay.
[02:01:39] Thank you.
[02:01:42] Yeah.
[02:01:43] Guys, once done, please like drop a response in the chat because otherwise I cannot
[02:01:58] progress.
[02:01:59] Doing that.
[02:02:06] Okay.
[02:02:07] Where is the bot father?
[02:02:33] I didn't get while login.
[02:02:37] I this option I didn't get.
[02:02:39] Yeah, you have to go to this account at the rate bot father.
[02:02:43] Is it?
[02:02:48] Yeah.
[02:02:49] By the way, telegram is a multi-billion dollar company that is operated completely
[02:03:17] remotely by a group of less than 30 people.
[02:03:24] It is something that the Russian government and other governments around the
[02:03:30] world have failed to ban.
[02:03:33] 26 employees, right?
[02:03:39] No, not six.
[02:03:40] I think 30 26 26.
[02:03:43] Okay.
[02:03:44] It's an amazing feat of engineering and I have massive respect for the guy.
[02:04:25] Guys, are you done?
[02:04:27] Because we are already kind of running out.
[02:04:44] Are you looking for any certain number of people to be ready?
[02:04:49] Really, but like I could see that about 40 45 guys were hopping on the lovable app.
[02:04:57] Similarly, there were about 15 20 people on the Google Doc, but there are only like
[02:05:02] five, seven, maybe 10 people on this.
[02:05:06] Because see, I can proceed.
[02:05:08] Okay.
[02:05:09] Without a problem, but then others will not be able to catch up and then the
[02:05:14] trainer comes and does it too fast.
[02:05:21] Okay.
[02:05:22] Let's let's proceed nevertheless.
[02:05:24] Okay.
[02:05:25] So the first thing you need to do is it what father has a new a bunch of commands.
[02:05:30] Okay.
[02:05:31] If you text any something like hello, how do I create a bot?
[02:05:38] It gives you an automatic message.
[02:05:42] It says you can control me by sending these commands slash new bot to create a new bot.
[02:05:48] Let's do that.
[02:05:49] All right.
[02:05:54] A new bot.
[02:05:55] What are we going to call it?
[02:05:57] Let's call it an eight and bought an eight and bought three.
[02:06:03] Okay.
[02:06:04] I have a few bots here.
[02:06:06] Good username.
[02:06:07] It must end in underscore bought.
[02:06:10] So let's call it an eight and bought three underscore bought.
[02:06:15] Okay.
[02:06:19] Let's call it any 10 32.
[02:06:24] Okay.
[02:06:31] The bot has been created.
[02:06:33] It also has a link for itself.
[02:06:35] You can add a description.
[02:06:37] Fine.
[02:06:38] But what is important here is use this token to access the HTTP API and this is what is
[02:06:45] necessary.
[02:06:46] Okay.
[02:06:47] So let's go back to any 10 now.
[02:06:52] Let's go back to HR policy bot.
[02:06:56] Okay.
[02:06:57] Let's pin up the AI again.
[02:07:01] Let's say I want to connect it with telegram as well.
[02:07:11] It should work.
[02:07:14] Okay.
[02:07:15] Do not remove access.
[02:07:17] Do not remove the existing web hook API access but add the telegram bot access using more nodes.
[02:07:45] By the way, Salman, if you're present, this is the end of the official session.
[02:07:50] So take the ratings poll.
[02:07:53] Okay.
[02:07:54] Sir, I have already launched the rating polls.
[02:07:57] Oh, okay.
[02:07:58] Sorry.
[02:07:59] I missed in that case.
[02:08:00] I want to connect it to telegram.
[02:08:02] Okay.
[02:08:03] Thank you, sir.
[02:08:05] Okay.
[02:08:06] Telegram bot access during more nodes.
[02:08:10] Fine.
[02:08:11] It should take a question from the user via telegram and give the user the
[02:08:23] response on telegram itself.
[02:08:28] The reason I'm being so stupidly explicit is so that the AI gets me properly.
[02:08:34] Okay.
[02:08:35] Because otherwise sometimes it doesn't and it takes way more time just to debug the
[02:08:41] error than what it would have if I had just in the explanation a bit more easy,
[02:08:46] in a more easy way.
[02:08:51] I have pasted the prompt in the doc as well.
[02:09:36] Allow AI assistant to edit workflow.
[02:09:39] Obviously.
[02:09:58] Does anyone have any doubts they want to clear while this AI is working?
[02:10:02] So what I understood so far, like you will be asking the similar question from telegram
[02:10:18] itself.
[02:10:19] Is it right?
[02:10:20] Yeah.
[02:10:21] How many actual days we can do it?
[02:10:23] Okay.
[02:10:24] So yeah.
[02:10:26] So basically we are providing it alternate heads to take inputs from.
[02:10:31] Okay.
[02:10:32] I'm just showing you how to connect different apps to the same workflow.
[02:10:37] So now that we know that it has done some processing and we might need to fix this.
[02:10:43] Okay.
[02:10:44] Because it is getting slightly redundant.
[02:10:46] We have this prompt where it is asking us to connect telegram.
[02:10:51] If we open connect telegram, it asks us for an access token.
[02:10:55] We were already given the access token by bought father.
[02:10:59] Okay.
[02:11:00] So let's just copy this.
[02:11:04] Let's see if it got copied properly because if there are extra spaces or letters, it becomes
[02:11:09] a problem.
[02:11:10] Let's paste the access token here and save and let's apply.
[02:11:33] Okay.
[02:11:34] It says it works.
[02:11:36] I don't know if it works.
[02:11:39] But before that, let me just publish this updated version and then check it from telegram.
[02:11:47] If I want to check it, what I will do is I'll go back to this account that it had given me
[02:11:52] for my bought.
[02:11:55] I'll start a conversation.
[02:11:57] I will ask it how many casual leaves can I take in a year.
[02:12:14] The policy states employees can take up to six personal slash casual leaves per calendar year.
[02:12:27] See, so it works.
[02:12:29] But right.
[02:12:35] It didn't end.
[02:12:36] And I think why it is showing me in a two different flows rather it is updating existing one.
[02:12:42] See, I was also thinking about the same thing that why did we simply not paste a telegram
[02:12:50] node here and another telegram node here and connected with the existing thing.
[02:12:54] Then we have less redundancy in the system.
[02:12:56] Isn't it?
[02:12:57] I think it's more of an architectural decision that AI took.
[02:13:01] I have seen I have seen in previous classes where I have given the same prompt and it had
[02:13:07] done this and it had done the previous thing.
[02:13:10] But in this in this case, it made two copies of the same.
[02:13:15] Let me ask the same question to this.
[02:13:17] Let me see if it can paste both of them to this.
[02:13:21] And then we if we can then we'll make it work.
[02:13:24] By the way, yes, Raghavan's hotel may.
[02:13:27] I believe the problem is if it just pastes the telegram hook at the end, it should also
[02:13:33] decide when to send a response to telegram and when to send a response to web hook, right?
[02:13:38] Will it have any condition or something?
[02:13:41] I mean, it does and it and inherently does support in if conditions.
[02:13:47] Let's see.
[02:13:48] Let's ask the end get to the root of this.
[02:13:50] Why did you clone the existing workflow and make a copy of it?
[02:14:01] Why didn't you attach the only the telegram nodes to the previous structure?
[02:14:17] Answer the question.
[02:14:21] Don't make changes to the to it because I have seen both of them, both of these ways working.
[02:14:31] Okay.
[02:14:34] Let's see if the input is from telegram, then the response should also to the telegram.
[02:14:42] And if it is from the web page, then it has to be in the web page.
[02:14:45] Yeah, but it can also have like a if node here and then branch out.
[02:14:51] See, an agent does support branching out.
[02:14:54] Okay.
[02:14:55] It could have just branched out.
[02:14:56] I did not make a separate copy.
[02:14:58] It's still one workflow.
[02:14:59] What I did do is I had parallel duplicate branch.
[02:15:03] Okay.
[02:15:04] Rather than feeding, why didn't I attach telegram response node is web hook specific.
[02:15:11] Okay.
[02:15:12] Agent reads the question from a specific trigger.
[02:15:15] So it's also trigger specific.
[02:15:17] I think what it means to say is it is trig source as in trigger specific to triggers into one linear chain.
[02:15:25] Do not cleanly branch back if both triggers fed the same middle nodes, the output would still flow into a single response with the answer node.
[02:15:36] Okay.
[02:15:37] Okay.
[02:15:38] To share the middle nodes, I would have needed to add if switch logic to detect the source.
[02:15:46] Now this is what I was talking about where it could have used an if statement as well and branched out, but it chose to know not do that because it feels telegram has its own clean parallel path.
[02:16:07] So if I'm not wrong.
[02:16:13] You can also have one structure using multiple if switch statements so as to remove these this extra copy.
[02:16:37] And I also mentioned don't make the changes.
[02:16:44] Okay.
[02:16:54] So see it also admits that it has an alternate design.
[02:17:02] Okay.
[02:17:03] So sometimes it takes that sometimes it takes this that kind of officially ends the, you know, things for today's session.
[02:17:21] I will take questions now.
[02:17:24] Yes, Raghavendra.
[02:17:26] Could you please suggest like if you have any list of different workflows which we can try over the week.
[02:17:33] I would suggest to go to the templates on the site.
[02:17:38] We were already exploring templates, I guess templates templates and explore running the templates that is the fastest way to actually get anything useful out of any 10.
[02:17:55] You have like 12,000.
[02:17:59] Now depending on what you work on.
[02:18:01] Okay.
[02:18:02] Let's say you want to create LinkedIn posts or you want to create AI generated videos for YouTube shots.
[02:18:09] Okay.
[02:18:10] You can search for them.
[02:18:11] Let's let's search YouTube 120 automations translate dubbed videos create upload ASMR shots news digest videos, YouTube video metadata,
[02:18:30] cinematic videos from text prompts.
[02:18:32] Okay.
[02:18:33] Now you can create let's say shorts using this.
[02:18:36] So yeah, try these.
[02:18:39] Okay.
[02:18:40] And even if my subscription ends the automations which I created will still work is what you're telling him initially.
[02:18:46] Right.
[02:18:47] See, if your subscription ends your account will stop working.
[02:18:52] You will be your workspace will be zipped and you will be made to download it.
[02:18:58] Then you can let's say upload it to a new account or you can run it on your personal computer.
[02:19:05] I guess someone asked also asked me to show how to run it on their personal locals.
[02:19:11] Let me show you guys that because that is also important.
[02:19:16] So guys, if you want to run a copy of any 10 that is not limited by the 14 day trial, you can do this.
[02:19:26] Go click on the GitHub icon.
[02:19:28] Okay.
[02:19:29] Go to this page on GitHub.
[02:19:31] Inside this, you have a quick start.
[02:19:36] Try instantly with this.
[02:19:38] So essentially all you need to do is paste this.
[02:19:43] So you copy this.
[02:19:45] Go to terminal.
[02:19:46] Okay.
[02:19:47] Paste it.
[02:19:48] Press entered.
[02:19:50] Everything else should be automatic.
[02:19:53] Okay.
[02:19:54] But this step requires a software called Docker and installing Docker will take about, let's say 10 ish minutes.
[02:20:04] You might need to restart your computer.
[02:20:06] Go into BIOS, but it's a one time thing and it's some it's, it's kind of important.
[02:20:12] You will use it across different pieces of software you will build throughout the course of anything soft very that you do.
[02:20:19] So install Docker.
[02:20:21] Then go to any 10.
[02:20:23] Copy this line.
[02:20:25] Paste it in the terminal.
[02:20:26] Everything else should be automatic.
[02:20:28] Now a couple of words of caution here.
[02:20:31] One is your local computer cannot receive requests.
[02:20:38] Okay.
[02:20:39] So imagine you have a lovable page like this.
[02:20:43] Okay.
[02:20:44] And you're hosting and you're running any 10 on your laptop.
[02:20:49] This means that the web page will not be able to access the instance of any 10 that is running inside your laptop.
[02:20:58] Okay.
[02:20:59] This is a result of what we called port filtering, I think, or basically we the way our Nats.
[02:21:13] Okay.
[02:21:14] Basically our laptops, okay, that are running on our home Wi-Fi network are not accessible directly by the outside internet.
[02:21:22] So you need what is known as a public IP address.
[02:21:25] So it will only work if you have a server and that server essentially needs to run on an interface on a on an infrastructure like AWS, GCP.
[02:21:39] Okay.
[02:21:40] Or you can also go for hosting or hosting or has good anything.
[02:21:43] Things with an eight and self hosted and even for 799 rupees a month.
[02:21:56] Okay.
[02:22:00] So even if you don't want to pay, that's it $20 every month, you can get it for as low as 599 a month.
[02:22:08] Okay.
[02:22:12] So I have one more question.
[02:22:14] Tell me.
[02:22:15] Yeah.
[02:22:16] Yeah.
[02:22:17] They actually we have next week the exam and the madam confirmed that she will share the MCQs.
[02:22:24] But today we got that it is like in the blanks and even they mentioned like it is the somewhat one online towards but it is not clear because the comments hold something MCQs and she will share the templates also how this MCQs will be there but we
[02:22:42] will not be able to do that.
[02:22:43] So I will give you the results.
[02:22:44] But next week we have the exam.
[02:22:45] Is there anyone from them they are aligning to the drum?
[02:22:48] M.
[02:22:49] M.
[02:22:50] M.
[02:22:51] M.
[02:22:52] M.
[02:22:53] M.
[02:22:54] M.
[02:22:55] M.
[02:22:56] M.
[02:22:57] M.
[02:22:58] M.
[02:22:59] M.
[02:23:00] M.
[02:23:01] M.
[02:23:02] M.
[02:23:03] M.
[02:23:04] M.
[02:23:05] M.
[02:23:06] M.
[02:23:07] M.
[02:23:08] M.
[02:23:09] M.
[02:23:10] M.
[02:23:11] M.
[02:23:12] M.
[02:23:13] M.
[02:23:14] M.
[02:23:15] M.
[02:23:16] M.
[02:23:17] M.
[02:23:18] M.
[02:23:19] M.
[02:23:20] M.
[02:23:21] M.
[02:23:22] M.
[02:23:23] M.
[02:23:24] M.
[02:23:25] M.
[02:23:26] M.
[02:23:27] M.
[02:23:28] M.
[02:23:29] M.
[02:23:30] M.
[02:23:31] M.
[02:23:32] M.
[02:23:33] M.
[02:23:34] M.
[02:23:35] M.
[02:23:36] M.
[02:23:37] M.
[02:23:38] M.
[02:23:39] M.
[02:23:40] I just also have another question to manager. Will we get the recording of this session?
[02:23:46] This recording would be there on your AI clinic part, on your AI clinic tab within 24 to 48 hours.
[02:23:57] Or we can find it on the AI clinic part?
[02:24:01] It would be on your courses, on your LMS.
[02:24:04] It will be available only for 24 hours.
[02:24:12] Guys, let's the professor finish. I will take if any questions are there towards the end.
[02:24:16] Sorry, sir. You can proceed.
[02:24:19] Thank you.
[02:24:19] Yes, Shree question.
[02:24:23] The thing is, I have shared the permission for my Google Doc.
[02:24:27] And the problem is, I still could not access the talk.
[02:24:32] So, share your screen?
[02:24:34] Guys, is there anything that I can access it?
[02:24:39] Share your screen?
[02:24:47] Guys, one thing. No offense to anyone.
[02:24:51] Okay, but you guys have been completely messing up my name. It's Shopro.
[02:24:58] Yes, show me a double click on it. What is the error? Resource could not be found.
[02:25:08] How did you log into this Google account?
[02:25:12] I think it's like I have nothing. It's like I have given the permission,
[02:25:18] but I don't know how to change it.
[02:25:20] Great. You have given what permission?
[02:25:24] To read permission to this doc.
[02:25:28] Where have you given the read permission?
[02:25:30] When I acted it first time, I was prompted to give a permission.
[02:25:36] There only I gave.
[02:25:38] Okay, then do this. Create a new credential. Double click on the Google Doc.
[02:25:43] Yeah. Do you see the pencil icon beside Google Docs account?
[02:25:48] This one?
[02:25:48] Yes.
[02:25:49] Okay.
[02:25:51] Connection.
[02:25:54] Click disconnect.
[02:25:59] Sign in.
[02:26:09] Now you can select whatever permissions you want to give.
[02:26:12] Yes, Shri Krishna, tell me.
[02:26:14] Yeah, I mean this one like already N8 input already has some access,
[02:26:18] but do we need to give full access or only?
[02:26:22] See, if you don't give it C access, then it does not get to see the files on your drives.
[02:26:27] If you don't give it edit access, then it cannot write to a file.
[02:26:31] If you do not give it create access, then we probably don't need it because we are not creating
[02:26:36] any files as of now, but obviously it's not going to delete.
[02:26:40] So I don't have a problem giving it all access.
[02:26:49] What you cannot do is give it access to just one file.
[02:26:54] Probably you're looking for that, but no, we cannot give it access to just one file.
[02:27:02] Shri Krishna, tell me anyone else?
[02:27:14] Hello.
[02:27:15] Yes, Pradesh, tell me.
[02:27:17] I want to try Telegram connect to my this of Indian file, but it's not coming connected.
[02:27:24] I'm sorry?
[02:27:25] That might my telegram token.
[02:27:27] I want to connect it, but it's not connect my Indian data.
[02:27:35] Let me show you what you mean by that question.
[02:27:38] Let me show you this.
[02:27:42] See, I want to connect this telegram account to this just one.
[02:27:48] By token, I want to add token here.
[02:27:52] Right.
[02:27:52] Is it to write?
[02:27:55] Delete that.
[02:27:57] Do you have the open your telegram show me the account, the long credential?
[02:28:04] Is that in my phone?
[02:28:07] Well, how did you paste it here?
[02:28:09] I pushed pasted from the my another.
[02:28:11] I want to send the current.
[02:28:13] Well, then open a new tab pasted there.
[02:28:17] Let me see the credential.
[02:28:24] This one.
[02:28:27] You're sharing just your browser.
[02:28:29] If there are any other windows, I can't see them.
[02:28:32] Okay.
[02:28:38] Let me.
[02:28:40] You have to stop sharing and share your entire screen.
[02:28:45] Okay.
[02:29:01] Pradesh, if you're not ready,
[02:29:03] you can perhaps come back to me.
[02:29:06] Okay.
[02:29:07] Later.
[02:29:08] Okay.
[02:29:09] Okay.
[02:29:10] Guys, I think that is all for this session.
[02:29:14] I'll stop dragging it for everyone.
[02:29:16] If you have any questions, I have dropped my LinkedIn.
[02:29:20] Feel free to connect me there.
[02:29:21] Connect with me there.
[02:29:22] Ask me questions.
[02:29:24] I will most likely reply late because, you know,
[02:29:28] I might have just be taken up in something.
[02:29:32] But I will definitely come back to you.
[02:29:34] Okay.
[02:29:35] Might take me a day or two, but I'll definitely come back to you.
[02:29:38] Okay.
[02:29:39] That is all for today's session.
[02:29:41] All right.
[02:29:42] Just one question.
[02:29:43] Is that fine to ask now?
[02:29:46] Yes, Karthik.
[02:29:47] Okay.
[02:29:48] So on the same architecture, which we are discussing.
[02:29:51] So I think we can even connect Telegram rather.
[02:29:54] I mean that we can connect multiple channels.
[02:29:56] So through Telegram through the website, a loveable website.
[02:30:00] So I mean, in that same architecture.
[02:30:03] So through this, we can have multiple apps.
[02:30:05] And so based on if I am asking through a liveable website,
[02:30:09] so it should be able to respond me that question.
[02:30:12] So if I'm asking to Telegram,
[02:30:13] it should be able to respond me that question.
[02:30:15] So even that level of architecture also it's possible.
[02:30:18] Right.
[02:30:20] That's what we did achieve today.
[02:30:21] I mean, you did see that we connected the HR board
[02:30:24] with a form with Telegram and with loveable, right?
[02:30:29] Yeah.
[02:30:29] So but we didn't separately, but even though if we combine all,
[02:30:33] but still this architecture should work.
[02:30:35] Yeah.
[02:30:36] We also saw that we can combine them and it will still work.
[02:30:40] The AI literally answered about that.
[02:30:44] Okay.
[02:30:44] So we don't need to have a logic or something like
[02:30:48] like we need the logic to not have those parallel structures.
[02:30:54] So they're not getting it.
[02:30:55] Can you be clear?
[02:30:57] Yeah.
[02:30:57] So if you notice.
[02:31:00] See, we have two copies.
[02:31:03] Right.
[02:31:03] For Telegram one for web book.
[02:31:06] Yeah.
[02:31:06] But they also asked that we can consolidate with two tricks merged in
[02:31:12] or shared middle and an if because this structure does not have an if.
[02:31:18] Okay.
[02:31:19] Okay.
[02:31:20] Okay.
[02:31:20] That's it.
[02:31:25] It will still work, right?
[02:31:26] I mean, I understand it is it don't have an if, but if like since we already implemented
[02:31:31] Telegram.
[02:31:32] So now if you go back to level that side still it will work.
[02:31:34] Is it right?
[02:31:36] Yeah, it will.
[02:31:36] It will.
[02:31:37] Okay.
[02:31:38] Okay.
[02:31:39] Thank you.
[02:31:39] That's it.
[02:31:41] That's it for that's all for today's session.
[02:31:43] Thank you guys.
[02:31:44] Hopefully it was I was able to inform you or teach you something.
[02:31:49] Thank you.
[02:31:50] Thank you so much.
[02:31:51] Thank you.
[02:31:51] Thank you.
[02:31:52] It was very nice session.
[02:31:54] Thank you.
[02:31:54] Thank you.
[02:31:54] Thank you.
[02:31:56] Thank you.
[02:31:57] Thank you Vikram.
[02:31:58] Thank you on everyone.
[02:31:59] Thank you.
[02:31:59] Thank you.
[02:32:02] Thank you.
[02:32:02] Thank you.
[02:32:03] Thank you.
[02:32:03] Bye.