YT YouTube2Text

YouTube2Text → Transcripts → Build a Trading Algo with AI | No Coding Required

Build a Trading Algo with AI | No Coding Required — Transcript

by Zerodha Varsity · 9,115 words · 1,464 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00I have been trading and investing since
0:022006.
0:04It's been a while and I've done all
0:06sorts of things in the market be it
0:08swing trading, intraday trading,
0:10investing, even mutual fund SIPs. But
0:13there's one thing that I've always
0:15wished I could do and that is algo
0:17trading. Although I understand what an
0:20algo is, I was never able to implement
0:23it on my own due to my inability to
0:26code. for anyone to develop an algo. To
0:28help me understand how I can develop my
0:31very first algo, I have my colleague
0:33Akshai here and he has promised me that
0:36he will teach me how to develop my very
0:38first algo using AI and of course using
0:41the APIs offered by your broker. So if
0:44you're like me wanting to learn how to
0:46develop an algo without knowing how to
0:49code, then go ahead, grab a cup of
0:51coffee, get your notebook and pen and
0:53we'll get started.
0:56All right, Akshai. Uh, so let's get
0:59started. Uh, but before we deep dive,
1:02tell me what is an ALGO?
1:05>> So, Karthik, an ALGO is basically a set
1:07of instructions, uh, trading
1:09instructions where you give it an input,
1:11which could be your data that you get
1:12from your broker and you get an output
1:15as a signal or you could automate it and
1:17it also places an order for you.
1:18>> Okay. Uh so depending on your use case
1:21an algo could be a fully automated algo
1:23or a semi-automated algo where you the
1:26trading is in basically in your control
1:27and you make a decision on the signal
1:30that you get.
1:30>> Can you very quickly give me an example
1:33of what an algo could be in terms of
1:36instructions? An algo can be something
1:38as simple as uh say buying like 200
1:41shares of Infosys if it if it uh if
1:44there's a moving average crossover
1:45happening or and even an RSI if there is
1:48an overbought and oversold level. Uh you
1:51could go long and short on any
1:53particular stock. So, not only can you
1:55take a decision on one stock, you could
1:57also use it to probably scan multiple
1:59stocks at the same time. And uh yeah,
2:01and basically refine your trading based
2:04on
2:05>> your correct me if I'm wrong. So, let's
2:07say I define my algo right now.
2:09>> I want to buy 100 shares of Reliance
2:13whenever HDFC moves above its 200 day
2:16moving average for whatever reason.
2:18>> Okay,
2:18>> that does that qualify as an algo? That
2:21is also technically possible
2:22[clears throat] if you have your rule
2:23set written in such a way that you know
2:26HDFC has to cross a particular level and
2:28you buy reliance for that.
2:29>> Okay. Because you said it's a bunch of
2:31instructions. It can it could be as
2:33wacky as this.
2:34>> But technically this is an algo also,
2:36right?
2:36>> Yes. Technically this would be an algo
2:38as well. So it just needs a set of
2:40instructions.
2:41>> Okay.
2:41>> Uh and if those instructions are met, if
2:43the rule set are met, it has a decision
2:45then to either go long or short on
2:47whatever decision you have made.
2:48>> Okay. All right. Aai. Now before we move
2:51ahead, I want you to reassure me once
2:53again that I don't have to know how to
2:56write a single line of code and yet
2:58build my first algo. Uh is that a
3:01definitive yes or a no?
3:02>> So I give a 100% promise that neither
3:06you nor I will be writing a single line
3:08of code.
3:08>> Okay.
3:09>> Uh we would orchestrate the entire algo
3:11using uh uh AI coding assistant that
3:14we'll be [clears throat] installing.
3:15>> Akshai. Now before we deep dive you also
3:17mentioned that there is something called
3:19as a semi-automated and fully automated
3:22algo. Now quickly can you please explain
3:25to me what do you mean by this and also
3:27tell me what is it that we are going to
3:29build today?
3:30>> Uh so an algo is basically your set of
3:32instructions that could give you a
3:34signal based on those instructions being
3:36met. Now in an automated algo uh the
3:39signals will be acted upon by the algo
3:41itself and to place the order but in a
3:43semi-automated uh case you would get the
3:45signals displayed to you as an alert and
3:47you would have to log into a trading
3:49terminal and then place the order
3:51manually yourself.
3:52>> All right. So basically in a
3:53semi-automated uh algo the end decision
3:56is upon me.
3:57>> Yes.
3:58>> So I the algo can suggest a signal but
4:00whether to take that signal and place
4:03the order the decision making is left to
4:05me.
4:05>> Yes. Whereas in a fully automated
4:07system, the algo takes that decision on
4:10on its own.
4:11>> Yeah.
4:11>> Right.
4:12>> Yeah.
4:12>> Super. And what are we going to build
4:14today?
4:15>> Uh so we would uh since we are doing
4:17this for the first time, I would suggest
4:19we start with a simple semi-automated
4:21system where
4:22>> uh we work on a daily time frame and
4:24generate a list of signals on list of
4:26stocks that are showing the signals for
4:28your trading idea. By the way, if you
4:31guys want us to demonstrate how you can
4:33generate a fully automated algo, then
4:35please do comment and maybe we can
4:38consider doing a separate video for
4:39that. Okay, Aai, now that you've given
4:42me some sense of what an algo is and how
4:44we are going to develop our first algo,
4:46uh can I define the algo that we want to
4:49develop and it's a fairly simple one.
4:52The idea is to scan all the top 100
4:54stocks of Nifty, basically the Nifty
4:56100, and generate a signal, bullish or a
5:00bearish signal, whenever the 6-day
5:02simple moving average crosses over a
5:0530-day simple moving average. Right? And
5:08when a signal is generated, I want the
5:10ALGO to list the signal in terms of its
5:14recency. Meaning if a signal if the
5:17crossover has happened yesterday in a
5:19particular stock I want the algo to give
5:22it the highest weightage
5:24>> as opposed to a signal that has happened
5:26let's say 2 weeks ago right so basically
5:29the most recent signal the algo assigns
5:33a higher weight so a this is the algo I
5:35have in mind can this be codified and we
5:39can put this in a form of algo and not
5:41just that I know we are using an AI tool
5:45Can we also do it in such a way that the
5:47entire algo is visually represented in a
5:51nice UI?
5:52>> Yeah. So that is actually a perfectly
5:54good example of uh first AL uh trading
5:57idea as well as how to visualize uh this
5:59trading idea in particular.
6:01>> Uh so yeah we can do it and I think we
6:03can get started with the process.
6:05>> Cool. So I'm fairly excited. Aai uh you
6:08please tell me what is the very first
6:10step that I need to take.
6:12>> Yeah. Uh so for the very first step we
6:14will install uh uh codeex which is
6:16CHGP's very own coding agent and we'll
6:19showcase its uh strengths in the sense
6:20that we would use it to install all the
6:23other tools that we need uh to build
6:25this algo. Yeah. So everything end to
6:27end would be done by the coding agent
6:28and we would uh in natural language uh
6:31prompt it to build our algo.
6:34>> All right. So as step number one we are
6:36installing codeex right?
6:37>> Yeah. Yeah. So let's start with
6:39installing uh codeex. Uh since it's a
6:41chatg product, we could just uh type
6:43chatgpt and codex on Google search and
6:46the first link is where we would have
6:49our download for our codeex. Yeah. So
6:51let's install it for Windows. Codex is
6:53also available for Mac OS as well. So
6:55since I'm using a Windows system, I'm
6:57just going to install it for Windows for
6:59now. Uh I'll run the Codex installer as
7:02well.
7:04And the installation should just take a
7:06couple of minutes.
7:10Now that we have Codex Karthik, I'll
7:12just open it and launch the app. A
7:14>> I'm looking at Codex for the very first
7:16time and to me it looks very similar to
7:18Chad GPT's uh user interface.
7:21>> Yeah, Karthik. So Codeex is basically a
7:25tool by the same uh company. Uh
7:28essentially it sits on your system and
7:30is able to interact and build files uh
7:32on your system directly. So it is a far
7:35more powerful tool than the chat GPD
7:37that we use on the browser.
7:39>> So Akai uh now that we've got codeex uh
7:42you've helped me understand the
7:44difference between what codeex is and
7:46chat GPT is uh so what is the next step?
7:50What should what should we do?
7:52>> Yeah. So with regards to building a
7:53semi-automated algo dashboard uh we have
7:56two prerequisites that we uh need to
7:58install. Uh the first tool that we'd
8:00like to install is Python which is the
8:02programming language
8:03>> on which you would write our rule set
8:05on.
8:06>> Uh the second tool that we'd like to
8:08install would be NodeJS.
8:10>> Okay.
8:10>> That would wrap around the algo and help
8:13visualize it in a cleaner and a more
8:15beautiful UI.
8:16>> Okay. So basically Python is our main
8:20programming language.
8:21>> Yeah.
8:21>> But just to clarify, we are not going to
8:23write a single line of code.
8:25>> Yeah. And uh Codex is going to do that
8:27for us. And we also have need this other
8:30tool called NodeJS to wrap a UI around
8:34whatever we are developing.
8:35>> Yes, that's exactly it.
8:36>> Okay, cool. Okay. So, how do we go about
8:39installing these uh utilities or tools?
8:42>> Yeah. So, let's use codeex itself to
8:46install everything that is required.
8:47It's a very simple prompt. It's like
8:48prompting how you would prompt in chart
8:50GPT. It's the same prompt that you would
8:52uh type here as well. Okay.
8:54>> So, let's start with Python first. So, a
8:56very simple prompt.
8:58>> Okay.
9:02[clears throat]
9:02>> A your prompt is can you install Python
9:05and add it to path?
9:07>> Yeah.
9:08>> What is this path and how would a
9:11non-programmer even know that this
9:13should be a specific instruction? So
9:16essentially adding it to path helps uh
9:19the operating system know that Python is
9:22available uh throughout the operating
9:24system. So any Python files we create uh
9:27if you were to open it it would open as
9:28a Python file and not as a text file.
9:31>> Okay. So for ease just add it to the
9:33path.
9:34>> Yeah.
9:34>> Right. Okay.
9:35>> Yeah. Okay.
9:37>> Uh yeah. So as you can see is launching.
9:40it has found the uh the installer itself
9:44>> and it has launched it by itself.
9:46>> Okay. So again it was a simple prompt
9:49that codeex is now taking over the
9:51system and doing
9:52>> I know I'm jumping ahead but when you
9:54said [clears throat] you need another
9:55utility called NodeJS
9:58>> is it going to be a similar path uh
10:00similar prompt wherein you say can you
10:02install NodeJS and add it to the path or
10:06or is it a different uh prompt
10:08>> here it would just be can you install
10:09NodeJS
10:10>> okay so no need of path
10:12>> no no need of path here Codex has
10:14finished installing Python let's go to
10:15the second utility ility which is
10:17NodeJS. So the same prompt again can
10:20install Node.
10:36Okay. So I see NodeJS is also being
10:38installed by Codeex. Uh as a quick
10:41summary so far we've first downloaded
10:43Codeex from Chad GPT. Then we asked
10:46Codex to download Python. And then after
10:50Python was installed, uh we asked Codex
10:53to install NodeJS.
10:55>> Yes.
10:55>> Right. Uh just for my uh understanding,
10:58this need not be in the same order. I
11:00can install NodeJS first and then do
11:02Python. Right.
11:03>> So there is no uh you know particular
11:06order for this. Uh so with this have we
11:10got all the prerequisites ready?
11:13>> Yeah. So we mostly have all the
11:14prerequisite that we need. Uh we have a
11:16little bit that we would also need to
11:18install which is uh your broker API.
11:21>> Okay.
11:21>> Uh so depending on which broker API you
11:23use, you could install the module for
11:25that API. Since we are using Zeroda
11:27Skite Connect here,
11:28>> okay,
11:28>> uh we would go ahead and prompt Codeex
11:30to install Zeroda Skite Connect module
11:32for us.
11:33>> Okay. Uh so this module that you're
11:35talking about Akai, uh is it again in
11:37the form of a prompt that we need to
11:39install?
11:40>> Yeah. So this module is for Python. Uh
11:42so very similar to how we installed
11:44Python and NodeJS, we can just prompt
11:47Codeex to install this module for us for
11:48Python.
11:49>> Okay.
11:50>> So I'll just go ahead and type that.
11:56A I see your prompt as can you install
11:59Kite Connect SDK for Python?
12:02>> Yeah.
12:02>> Uh I'm curious to know would codex know
12:05what kite connect is? Will it have the
12:07entire context uh for this particular
12:10installation?
12:11>> Yeah. So in this case since kite connect
12:13is a specific mod module uh built by
12:16zeroda itself uh codeex has context of
12:19what kite connect is uh and in this case
12:21is very easily able to install it by
12:23just this simple prompt. Yeah. Uh so
12:26yeah as you can see codex has installed
12:27kite connect python SDK.
12:30>> Okay. So now we have codeex we have
12:32python we have nodejs and we've also
12:34installed something called as kite
12:36connect python SDK.
12:38>> Yeah. uh we've still not yet developed
12:40our algo. What is the next step?
12:43>> Yeah. So uh now that we have installed
12:44kite connect successfully uh the next
12:47step is to uh create an application on
12:50zeroda's kite developer portal.
12:52>> Okay.
12:52>> Uh the application is something that
12:54zeroda would recognize and would then
12:57basically we create a connection between
12:59both.
13:00>> Okay. and we receive an incoming stream
13:02of data and also if we have to post
13:04requests for order placement or any
13:07other requests that we have uh we could
13:09do using this application. So Ashai the
13:11way I understand this is so on one side
13:14you've instructed codeex to install
13:17Python install NodeJS and also install
13:21the kite connect Python SDK so you've
13:24set up some sort of uh infrastructure
13:26there on the other side you spoke about
13:30uh setting up an application within the
13:32Zeroda environment
13:34>> right so I'm guessing once this
13:36application is set up you're going to
13:38link the two and start developing ing
13:40the algo is my thinking correct?
13:42>> Uh your so your thinking is called
13:44Karthik. Uh we just have to create an
13:47account on Zeroda Skype developer
13:49portal.
13:49>> Okay.
13:50>> Uh post which we'll be creating an
13:52application within it.
13:53>> Okay.
13:53>> This application is what Zeroda would
13:55recognize and uh there would be a stream
13:58of messages that would be sent back and
13:59forth between Zeroda and this ALGO that
14:01you'll be creating.
14:02>> Okay. Uh and this application is the
14:05algo that we would be creating, right?
14:06>> Yes, the application is the algo.
14:08>> Okay. All right. But uh that sounds
14:10exciting Akai. Although I've been using
14:12Zerodha for such a long time,
14:14>> I have never really logged into this
14:16developer account. Although I keep
14:18hearing developer account uh every now
14:20and then in the office. So yeah, so a
14:24little embarrassing but I I've never
14:25used it. So please please do show me how
14:28to set it up.
14:29>> Yeah, Karthik. So let's navigate to the
14:31Kite developer account.
14:32>> Okay.
14:32>> So as you can see here, since I already
14:34have an account, uh it's already
14:36prefilled my information. But uh if you
14:38are creating an account for the first
14:39time,
14:40>> I'm guessing it's a simple signup uh
14:42page.
14:43>> Exactly.
14:44>> Okay.
14:44>> Uh so for the first time, we could avoid
14:47the IP white list since it's not starred
14:49as well.
14:50>> Okay.
14:50>> We can avoid this section
14:52>> and fill the rest of the sections here.
14:54>> Okay. Because I don't know what IP
14:56whitelist means. Uh
14:58>> uh as you may have realized, I'm a
15:00complete tech idiot. So So I don't know
15:03what IP whitelist is. But you're saying
15:04I can ignore this for now. Yeah, you can
15:06ignore this for now and towards the end
15:08we'll probably discuss what the IP
15:10whitelisting bit is.
15:11>> Okay, go ahead.
15:12>> So yeah, uh after logging into our
15:14developer account, this is how the page
15:16looks. Okay,
15:16>> so you have a tab for your apps,
15:18billing, a profile, forum, uh where uh
15:22there are discussions happening
15:23regarding APIs.
15:24>> Okay.
15:25>> Uh for now, let's start with creating
15:26our app first.
15:27>> Okay.
15:28>> So let's create an app.
15:29>> Okay.
15:30>> Uh so we see a couple of types here. For
15:32now, we can ignore personal and
15:33publisher.
15:34>> Okay. Since our algo requires historical
15:36data, we would have to go with the
15:39connect type. So connect requires 500
15:41credits which is equivalent to 500
15:42rupees and this gives us access uh to
15:45historical data as well as live market
15:47data.
15:48>> Okay.
15:48>> Uh this is charged per month. So that's
15:51500 credits for 30 days.
15:53>> So this is 500 rupees uh equivalent to
15:55500 rupees per month. Okay. This is
15:57>> so yeah a so basically what you're
15:59saying is we need to use connect because
16:03uh it has access to historical data
16:05which is exactly what we need for this
16:08particular algo.
16:09>> Yeah
16:10>> because we are scanning uh the crossover
16:12for the nifty 100 stocks the 6 day
16:16simple moving average versus 30-day
16:17moving average and how back in history
16:20are we going back for this? Uh so for
16:23this pit for our current algo we are not
16:25going too far back in history. Okay.
16:27>> Uh but the available data that we have
16:29here can go scan as far back as the
16:32stock's inception.
16:33>> Okay.
16:34>> So um
16:35>> uh so it is going to give me all the
16:37historical data. Uh if I'm thinking
16:40right this will be particularly useful
16:43if I have to back test strategies.
16:45>> Yeah. So this would be particularly
16:47useful if you want to back test
16:48strategies at a larger time frame.
16:50>> Okay. So let's start with an application
16:52name. Is there any name that you'd want
16:54to give for this algo?
16:56>> I'll be super imaginative. I'll call it
16:58my first algo.
17:00>> All right, that works.
17:02>> So my first algo.
17:04>> Okay.
17:05>> Uh and
17:07I'll add the zeroda client ID. So this
17:11is your zeroda's client ID that you
17:14have.
17:14>> Okay.
17:15>> So I'll add that as well.
17:16>> Okay.
17:17>> Um so now before Yeah. So now the next
17:20field that is required is the redirect
17:22URL.
17:22>> One one second. There is an app icon. Is
17:25it uh needed?
17:27>> An app icon is technically not needed.
17:29>> It's uh depending on whether you want to
17:32add an application icon as well.
17:34>> You could do it.
17:35>> Okay.
17:35>> Uh but for now we could ignore the app
17:37icon and also the postback URL.
17:39>> Okay.
17:40>> Uh the next field that we'd have to fill
17:41up here is the redirect URL. Okay. So
17:43the redirect URL is basically when you
17:45log to your application uh you would
17:49need a request token. Uh we would come
17:52to this in the authentication bit. Uh
17:54but for now all you need to know is that
17:56the redirect URL can be anything simple.
17:58It could be uh in our case we could also
18:01put google.com.
18:03>> Okay.
18:04>> So a very simple redirect URL. So I'll
18:06add that here.
18:07>> Okay.
18:08>> Uh and then a description for this algo.
18:10So maybe I can add your uh trading idea
18:14itself. So maybe u moving
18:18average
18:20cross.
18:23So a very simple description here.
18:25>> Okay. So redirect URL you've said
18:27google.com. So u you will have to
18:30explain
18:32maybe at a later stage uh why this is
18:35needed and what what actually should go
18:38in there. But we'll pick it up later.
18:40>> Okay.
18:40>> Um All right. So, so far it's pretty
18:44straightforward.
18:45>> Yeah. So, let's go ahead and create our
18:47application.
18:49>> Uh you would have to enter the I
18:51understand prompt here.
18:55>> Okay.
18:56>> And once you do so it there is a prompt
18:59here saying insufficient balance.
19:01>> Okay.
19:01>> Uh which is where we'd have to go to a
19:03billing account.
19:04>> Okay.
19:04>> Uh let's just go to our my apps and
19:06confirm. So you can see here your my
19:08first ALGO is created
19:09>> but the status says here as pending.
19:11>> Pending. Okay.
19:12>> So let's quickly go to the billing
19:14section
19:15>> and add 500 credit. So there are a
19:17couple of ways that you could add
19:18credits.
19:18>> Okay. It's a straightforward payment. Uh
19:21>> exactly. So you have a straightforward
19:23payment uh section or you could directly
19:25link it to your kite account that you
19:27already have
19:28>> and uh depending on whether you have 500
19:31rupees available in that account or not.
19:32Okay.
19:32>> You could directly add credits from
19:34there as well.
19:34>> Okay. Uh so for this example, I'll just
19:36use a payment gateway and add credits.
19:39>> Okay.
19:43>> Yeah. So as you can see, payment is
19:45successful. Our credits have been added
19:46here.
19:47>> Uh let's quickly go back to my apps. Uh
19:50go to our first algo here and we can
19:53activate it right now. So I'll just
19:54click activate now.
19:55>> Okay.
19:56>> And similarly, I'll just put
20:01the I understand prompt.
20:03>> Okay.
20:04Uh yeah.
20:08So yes, we have activated our first Algo
20:10application.
20:11>> Okay.
20:11>> Uh it tells us when it expires, so 30
20:14days from now.
20:14>> Okay. [clears throat]
20:15>> And these are the app details here
20:18>> that we filled in earlier.
20:20>> Okay.
20:20>> Now this is our application and our
20:22application has uh a couple of
20:25credentials that are unique only to this
20:28application and these are credentials
20:30that technically you should not share
20:31with anyone else. Uh so on one side
20:34you've set up the uh codeex and python
20:36and nodejs infra. So that was let's look
20:39at it as part a part b you've come to
20:42the zeroda environment and you've set up
20:44a application.
20:46>> Yeah.
20:46>> Uh you've we've called it my first algo.
20:49>> Uh after successfully creating it uh we
20:52are looking at something called as an
20:54API key and I know that there is
20:57something else called as API secret.
21:00>> Yes. Right. So
21:03is this uh is the API key and the API
21:06secret like my login ID and password uh
21:10is that correct? Is my understanding
21:11correct?
21:12>> Yeah. So the API key and API secret
21:15basically allows Zerodha's ecosystem to
21:17recognize that this application is
21:19trying to connect to Zeroda's APIs.
21:22>> Okay.
21:22>> So your API key is like your username
21:25>> and your API secret is like your
21:26password.
21:27>> Okay. uh apart from these two
21:29credentials you have one more credential
21:31that you will request from Zeroda itself
21:34and that is your session that that is
21:36called an access token. So that access
21:37token keeps your session alive for 24
21:40hours.
21:41>> Okay. Okay.
21:42>> So for now for just this process uh
21:45let's quickly save the API key since we
21:47don't want to keep coming back and forth
21:49on the developer account.
21:50>> We can save the API key and the API
21:52secret in a text file and save it in a
21:54folder which where we'll be working on.
21:57>> Okay. So I'll quickly create a new
21:59folder.
22:02So I'm creating a new folder. Is there a
22:04name that you'd want to call this
22:05folder?
22:06>> Uh let's call this folder my first algo
22:09itself. Again very imaginative.
22:13>> Yeah.
22:13>> Right. So this folder is created on the
22:15desktop. Uh I'll just quickly open it.
22:18>> Uh the folder is completely empty right
22:20now.
22:20>> Okay. Okay. Uh what I'll do is I'll
22:22create a new text file
22:26and call it credentials.
22:29>> Yeah, since we are doing it at a
22:30beginner uh state.
22:32>> Okay.
22:32>> We'll save our credentials in a
22:33[clears throat] text file. Ideally, a
22:35more robust way of saving it would be an
22:37env file where it's not technically
22:39visible directly.
22:42>> Yeah. So, let's save the credentials in
22:43the cred the text file. So,
22:45[clears throat] our API key here
22:55and our API secret as well.
23:07So, I'm saving both and
23:09I [clears throat] have saved them to
23:10this text file in the my algo folder. My
23:13first Algo folder. Yeah, just to
23:15reconfirm
23:18for a non-programmer like me, the way I
23:20understand this is the API key and the
23:22API secret uh is as sacred as your
23:26bank's user ID password. So, utmost
23:29caution has to be exercised while uh
23:32saving it on system and um uh and even
23:36sharing it with others. Right.
23:37>> 100%. Yeah.
23:38>> Okay.
23:40>> So now I have saved the API key and the
23:42API secret in a text file.
23:44Now we are good to start developing our
23:46algo.
23:46>> So for the first step what we do is we'd
23:48go back to codeex again.
23:50>> Okay.
23:51>> Uh in the same chat or you could create
23:53a new chat.
23:54>> You could give it the location of where
23:56your folder is. So I'll just copy the
23:58location here.
23:59>> Okay.
24:00>> And
24:02uh and again a simple prompt I'll ask
24:04can
24:05>> can we start a fresh chat just to uh
24:07>> yes
24:08>> uh start from like a clean slate.
24:11>> Sure.
24:11>> Okay. So we could start a new chat
24:13directly. So in the same uh chat
24:16>> I could I'll just put a prompt saying
24:19can you
24:21in this folder?
24:31Yeah. So I'm prompting codeex to now
24:33work in the folder that we just created.
24:36>> Okay. [clears throat]
24:46All right. So as you can see codeex is
24:48now going to work in this folder and
24:50it's also told us that it currently
24:52contains a credentials.ext.
24:54>> Okay.
24:54>> Uh now for the final bit about of the
24:57application is uh our login for this
25:00application will have a separate URL.
25:02>> Okay.
25:02>> And that URL will remain constant
25:04throughout.
25:04>> Okay. So we would first ask Codex to
25:07generate that URL for us and and we'd go
25:10ahead and bookmark that URL for future
25:12use as well.
25:13>> Okay.
25:13>> So a simple prompt here again. Can you
25:17give me the guide? Okay.
25:21Um based on
25:27yeah so I'm asking Codeex to give me the
25:29login URL here.
25:30>> Okay.
25:39So is this step basically
25:42uh asking codeex to connect to kite?
25:44>> Yes. So this is a step that's basically
25:47asking codeex to connect to our
25:49application uh URL.
25:52>> So as you can see here uh much of the
25:55URL is the same. The only bit that will
25:58change across uh different applications
26:00is your API key here. Okay. So this is
26:02the API key that was there for my
26:03application.
26:04>> Okay.
26:05>> So I'll just copy this URL.
26:08I'll go back to my browser. Paste it
26:11here.
26:12>> Now this URL you could bookmark since
26:14this URL doesn't change.
26:16>> Okay.
26:17>> Uh yeah. So you have to authorize the
26:20application.
26:22Yeah. So now that I've logged in,
26:25>> uh it has taken me to google.com.
26:28>> Uh now I'll probably explain the
26:30redirect URL. So the redirect URL of
26:32this application takes me to google.com.
26:35>> Along with that URL, there is certain
26:37extra information that is given here.
26:39>> Okay.
26:39>> As you can see,
26:40>> right?
26:41>> Uh the information that we require is
26:42the request token here.
26:44>> Okay. Okay.
26:44>> Uh so the request token is this block of
26:47text here
26:48>> that is between the request token and
26:51the and that is there here.
26:53>> Okay. Okay.
26:53>> So let's just quickly that is the
26:56request token that is there.
26:58>> Okay. Now that we have reached the step,
27:00we can now finally start building the
27:01algo.
27:02>> Okay. Uh
27:03>> but what are you going to do with that
27:05request token? Do we have to paste it in
27:07the credential file or
27:09>> No, no, no. We would not be pasting it
27:10in the credential file.
27:11>> Now that we're building our UI uh in the
27:14UI itself, we'll add our login uh
27:16portal.
27:17>> Okay. A so what we've done so far is
27:20we've built the uh infrastructure, the
27:23railroads required to develop our first
27:26algo. All the things that we've done
27:27with codeex, all the things that we've
27:29done on the zeroda kite connect
27:31environment, they're all the basic uh
27:34you know background work
27:36>> but we've still not developed a single
27:38line of code or rather we've still not
27:40asked codeex to you know write a single
27:43line of code uh and start developing the
27:46algo. Uh so have we reached that point
27:49or is there something more that needs to
27:51be done?
27:52>> No, we have reached that point now. We
27:54have done with everything that was
27:55required. uh we can start building our
27:57algo right now.
27:58>> Okay.
27:58>> So let's go back to codeex and prompt it
28:00to start building our algo.
28:01>> Okay.
28:02>> So now we can start building our algo.
28:04>> Okay.
28:04>> So I already have a prompt to build the
28:07algo.
28:08>> Okay.
28:08>> So I'm just going to paste that here in
28:09chat and ask codeex to build the algo.
28:12>> Okay. Now Akai, now before you hit enter
28:15and uh CEX starts doing its thing, uh
28:19please take me through this prompt line
28:21by line. What is it uh that you've
28:24you're asking Codex to build? I'm
28:26guessing this is the uh the heart of the
28:29algo engine that we are building, right?
28:31>> Yes. Yes.
28:33>> Yeah. So, please uh take me through it
28:35line by line.
28:36>> So, the prompt is basically asking Codex
28:38to build a single Python file.
28:40>> Okay.
28:41>> Which would run which would be our algo
28:42that we'd run.
28:43>> Okay.
28:44>> Uh so, the first thing it would do is it
28:45would au authenticate
28:47>> using the API key and the API secret.
28:49>> Okay. I will also be giving it the
28:51request token that we had from the URL
28:53earlier
28:54>> uh to get the access token.
28:56>> Uh post the authentication it would
28:58download the nifty 100 constituents.
29:00>> Okay.
29:01>> And get the instrument symbols from uh
29:04zeroda's API uh for instruments.
29:07>> Okay.
29:07>> Uh post that the next line is that it
29:09fetches the daily historical data. It
29:11then calculates the SMA 6 and SMA 30 for
29:14the all the stocks that we are
29:16>> compiling. uh it would then detect a
29:18bullish and bearish crossover and then
29:20rank them by the most recent crossover
29:22dates.
29:23>> Okay.
29:23>> So this is what it's doing at the
29:24background. It's output that it will
29:27have is it would rank the stocks.
29:29>> Okay.
29:29>> It would show us the symbol. Uh it would
29:32show the signal type the crossover date
29:34based on recency.
29:35>> Okay.
29:36>> Uh and the close price SMA 6 and SMA 30
29:39as well.
29:39>> Yeah. So just to get my understanding
29:42correct
29:44you are essentially [clears throat] now
29:46connecting to the zeroda kite ecosystem
29:50uh and now I understand uh why you said
29:52we need kite connect because you're
29:54asking it to download historical data.
29:57>> Yeah.
29:57>> So on one hand it is fetching all the
30:00data then it's also doing the math
30:03required to identify the crossover.
30:05>> Yes.
30:05>> Right. And you're also specifying a
30:08certain format in which you want the
30:10output in.
30:12>> So, so yeah. So, this is the meat of the
30:14algo, right?
30:15>> Yes.
30:16>> Okay. So, yeah, super curious. Why don't
30:18you hit that enter button?
30:20>> Uh, so right before we hit it, I'll just
30:22give it the request token that is
30:23required.
30:24>> Okay.
30:25>> So, going back to the redirect URL that
30:26was there earlier, let's just copy the
30:28request token as well.
30:30>> Okay.
30:31>> And I'll copy this bit.
30:34let codeex know that this is the request
30:36token.
30:37>> Okay.
30:42>> And let's hit enter.
30:45>> Okay.
30:46>> While Codex does whatever it's supposed
30:48to do, Ashai, uh what if I have a
30:50different algo in mind? What if the user
30:52has a different algo?
30:53>> Yeah.
30:54>> Uh it could be based on uh technical
30:57analysis, it could be based on
30:59>> uh I don't know standard deviations or
31:01whatever. So how can they go about
31:05building an algo uh using whatever setup
31:09you've shown us?
31:10>> Yeah. So the the beauty of Python and
31:13along with Codex is that uh almost all
31:16of the technical indicators this
31:18includes even candlestick patterns uh
31:20and the entire vastness of the technical
31:22indicators everything is available as
31:24modules in Python.
31:25>> Okay.
31:26>> Uh and Codex is smart enough to have
31:28known what your context is. So if you
31:31give it if your idea requires
31:34candlestick patterns,
31:36>> if your prompt involves candlestick
31:37patterns, it knows where which module to
31:40pull.
31:40>> Okay.
31:40>> So if the module isn't installed, it'll
31:42go ahead and install it.
31:43>> Okay.
31:44>> Uh but the entire vastness of technical
31:46analysis is possible using codeex and
31:48Python.
31:48>> Okay. Now for example, let's say I have
31:50a different algo in mind.
31:52>> Uh I want to download the historical
31:54data of um of Niftyund stocks. I want to
31:59calculate the returns of these stocks
32:02>> then based on the returns I want to plot
32:05the standard deviation
32:07>> and I want to identify the return which
32:10is at let's say the third or third
32:12standard deviation
32:14>> plus three standard deviation short
32:16those
32:17>> and go long on the minus three standard
32:19deviations something as quantitative as
32:22this can also be codified and an algo
32:25can be developed right
32:26>> yes it can very easily be developed In
32:28fact, all of this can be done in the
32:29same natural language that you're
32:31speaking
32:32>> if you were to prompt it in the same
32:33way. Codex would be able to build it
32:35completely uh end to end.
32:38>> So technically this uh I I I remember
32:41reading this tweet somewhere.
32:44>> The most powerful language today,
32:45programming language today is English.
32:48>> Yeah.
32:48>> Right. So as far as I can creatively
32:51express my algo
32:53>> in plain simple English natural language
32:56I should be able to develop any algo.
32:58>> Yes. In [clears throat] fact the more
33:00detailed you are in your uh in in our
33:02earlier prompt that we had
33:04>> uh not only did we tell it what to do we
33:06also gave it how the output should look.
33:08>> So the more detail you add to your
33:10prompt in what you want the algo to do
33:12uh the better it is.
33:13>> Okay. So so far from whatever I've seen
33:17uh I think the uh the difficult part I I
33:21wouldn't even call it difficult but uh
33:23for most common people
33:25>> uh including uh like someone like me who
33:28has zero tech idea
33:30>> I think the the confusing or the
33:32difficult part is to set up the initial
33:35infrastructure needed uh this is right
33:38from in installing codeex
33:40>> python uh NodeJS and on the the side
33:44setting up the entire kite
33:46>> uh you know kite connect ecosystem once
33:49the infrastructure is set
33:51>> I think the other things are fairly easy
33:53to do right the the defining of the algo
33:56itself is fairly easy
33:57>> yes so once your infra is set up the
34:00rest of use the usage of API is actually
34:02very easy
34:02>> the API the documentation is already
34:04available everything is built around the
34:06ease of usage
34:08>> okay
34:09>> yeah so as you can see here the the
34:12codeex has built the script here.
34:13>> Okay,
34:14>> I'll quickly just open the script.
34:16>> Wait, so are you telling me this is my
34:18algo?
34:19>> Yeah. So, this file here
34:21>> is now your algo.
34:23>> Okay.
34:23>> So, it's an SMA scanner.
34:27>> Yeah. So, this is how your algo looks.
34:29>> Okay.
34:30>> It's a lot of lines.
34:32>> This is uh super scary to me. Uh you're
34:35just showing me
34:37>> lines and lines of code that C codeex
34:40has beautifully written. I have I have
34:42zero technical idea to even verify
34:46whether this is correct or wrong
34:48>> but a to me
34:51>> while this is fancy this is not usable
34:54to me uh I can't I can't I'm not capable
34:57enough to figure out if these lines are
35:00correct code you know I can't do it so
35:04>> unless and until you put this in a nice
35:06UI
35:09>> this is I I can't use this. So, can we
35:12wrap up uh wrap around it a nice UI?
35:15>> Yeah. So, now that we have built this,
35:17we could build the same thing
35:20>> in a more useful.
35:22>> Why don't you close this? It is scaring
35:23me.
35:23>> Yes. So, I'll start by doing that.
35:26>> So, yeah, you don't have to see these
35:27scary lines of code. Uh let's wrap the
35:30entire algo in a nice good UI.
35:33>> Okay.
35:33>> Uh so, we'll break down the UI in three
35:35prompts and we'll slowly build the
35:37entire dashboard that we are going for.
35:39>> Okay. Uh so let's start with the first
35:42prompt. Okay.
35:42>> I already have the prompts uh
35:44pre-written.
35:45>> So I'll just paste them here.
35:49>> Yeah. So this is the prompt. Uh as you
35:51can see here in the prompt, there are a
35:53couple of tools that we did not install
35:56uh initially.
35:57>> Okay.
35:58>> Um which is react and beat.
36:00>> Uh in this case, Codex would recognize
36:02that these tools are not available
36:04>> and it would install it by itself.
36:06>> One second. So uh so your prompt says
36:10build a React plus wheat front end. So
36:14for non-programmers like me, how would I
36:16even know that you need something like
36:19React and Wheat to build a UI
36:22>> or now that you've said it, is it a good
36:25practice to always prompt saying that
36:28build a UI using React and Wheat? Oh
36:31>> yeah. So the reason why I've chosen
36:33react and wheat is the front end looks
36:35really beautiful when you use uh
36:37JavaScript.
36:38>> Okay.
36:38>> Uh if you hadn't prompted Codeex uh to
36:41build it in using React and it would
36:44probably use a module from Python and do
36:46the same as well.
36:47>> Okay.
36:47>> Uh just that the front end would not
36:48look as beautiful as it would with this.
36:51>> Okay. Can I generalize it this way that
36:53whenever I'm trying to build a UI,
36:56>> I need to use React and Wheat to make it
37:00look beautiful. Is it a fair
37:02generalization?
37:02>> That is a fair generalization to use
37:05React and V to make the front [snorts]
37:06end more beautiful
37:08>> and as we build this app you will
37:10realize how uh goodlooking the UI can
37:13be.
37:13>> Right. It better be good-looking.
37:14>> Yeah. [laughter]
37:16>> Okay.
37:16>> So yeah so the front end which is the UI
37:19bit is using React and we
37:22>> for the back end we are using Python
37:23itself. Uh particularly a module called
37:26fast API. Again, uh since fast API
37:28wasn't installed at the start, uh Codex
37:30will recognize the lack of this module
37:32and install it by itself.
37:33>> Yeah, by the way, for all the people
37:36watching this, all the prompts we've
37:39used, we will put it up in the
37:40description. So, it's it will be easily
37:42available for you all. So, don't get
37:45stressed uh you know about not seeing
37:47the prompt fully clearly at this point.
37:50You can always check on the description.
37:52Why don't you hit that enter button and
37:54let's see what happens. Uh sure Karthik
37:58>> a so typically how long does it take CEX
38:01to build a nice UI?
38:03>> Um so depending on your tasks it could
38:06take uh a simple prawn could take a
38:08couple of seconds.
38:09>> Okay.
38:09>> Uh but if you're trying to build
38:11something large and has multiple moving
38:13parts then it could take uh anywhere
38:15from a couple of minutes to probably 10
38:18plus minutes as well.
38:19while this does whatever it's doing
38:23to I just wanted to touch upon the cost.
38:26So firstly we spent 500 rupees for
38:30against 500 credits for kite connect. So
38:32that's on the
38:33>> uh API side of things.
38:35>> Yes.
38:36>> Uh are you using I should have asked
38:38this before but are you using a free
38:40version or a paid version of Codex here?
38:42>> So CEX is available both to the free
38:44users and the paid user.
38:46>> Okay. The version I'm using here is the
38:48paid version. Okay.
38:49>> Uh so this is uh a pro version that I'm
38:52using.
38:52>> Okay.
38:53>> Uh but in the free version also you
38:56could use uh since uh codeex functions
38:59on the basis of credits which basically
39:02get replenished every week.
39:04>> Okay.
39:04>> Uh so on the basis of your usage if
39:06you're using it for something very
39:08heavy. Okay.
39:08>> Uh like building a UI as well,
39:10>> it would take a certain amount of
39:12credits. So if you do run out of
39:14credits, you'll probably have to wait
39:15for a particular duration for the
39:16credits to get replenished.
39:18>> Okay. So basically uh it makes sense to
39:21have a pro version of whichever LLM
39:23you're using, right?
39:24>> Yeah. In if your use case is going to be
39:27to build an algo and continuously
39:29develop on it.
39:31>> My suggestion is to have a paid version.
39:33>> Okay.
39:33>> As it helps in that.
39:35>> Okay. So also while we are doing this
39:37using codeex I think uh people can use
39:41any LLM be it anthropic be it
39:44>> gemini or uh whatever else right so it's
39:47not it's it is basically LLM agnostic
39:50>> yeah this is completely LLM agnostic you
39:52could use cloud work as well cloud code
39:54as well
39:55>> uh or any other LLM that you that you
39:58prefer to build the same
40:00>> okay but regardless of which LLM you use
40:03the framework is going to be the same.
40:05So you first install
40:07>> the prerequisites on one side and then
40:09do whatever is needed on the kite
40:12connect side and then start prompting it
40:14uh by by using natural language.
40:17>> Yeah, that is
40:18>> okay.
40:18>> Yeah, Karthik. So I think Codex has
40:20finished uh building our uh UI.
40:23>> Okay.
40:23>> So let's quickly launch it and see. So
40:26it says the app is running locally now
40:28at front end here.
40:29>> So I'll just copy this link and go to my
40:32browser and paste it. Okay.
40:37>> So, yeah, as you can see, it has built a
40:40login page with the API key, API secret,
40:43and request token here.
40:44>> Okay.
40:45>> Let's quickly fill those forms in and
40:47see how what happens post the login.
40:49>> Okay.
40:50>> Um, so I'll just go back to my
40:52credentials that I had saved.
40:58>> Yeah. So, I filled the API key and the
40:59API secret.
41:00>> Okay. Uh the request token has a shelf
41:03life of a couple of minutes.
41:05>> Okay.
41:05>> Uh so you have to use it in that time
41:07frame. So since we had our request token
41:10called a long while back,
41:12>> we'll call it again.
41:13>> Okay.
41:14>> So I'll just log in again.
41:19>> So you're logging into the developer
41:21account.
41:22>> Yes, I'm logging into my application uh
41:24which has my API key here.
41:26>> Yeah, basically your application is in
41:27the developer account here.
41:31So I'll just log in.
41:38Yeah. So it's again redirected us back
41:40to google.com.
41:42>> Uh but the request token is here. So
41:44I'll just copy the request token. So
41:46everything between the request token and
41:48>> the and here.
41:49>> Yeah, I remember that.
41:51>> Copy that.
41:52>> Okay.
41:53>> Um go back to our dashboard that is here
41:56and paste it here. Now let's see what
41:59happens when you click login.
42:02So as you can see [clears throat]
42:04uh we have logged in uh it has it in the
42:08prompt here that we had shared earlier.
42:10>> Okay.
42:11>> We had asked it to create a dashboard
42:13which wherein after login
42:16>> it creates a user tab for us
42:18>> and then it calls the API for our user
42:20profile.
42:20>> Okay. So it calls the username, user ID,
42:24the products that are there and the
42:25exchanges that are available for this
42:27account.
42:28>> So since this is my account, uh the
42:30details that it's showing here are my
42:32username, my user ID,
42:34>> the products that I have available and
42:36also the exchanges that I have available
42:38for me to trade on.
42:39>> Okay. But show me the signals, the
42:41bullish and the bearish signals. That's
42:43what we are working towards, right?
42:45>> Yes. Yes. So we'll reach there. So this
42:46was the first prom that has built this
42:48>> and now we know that the APIs are
42:50running. Yeah. So now let's go ahead and
42:52add a signals tab which will show us the
42:54signals that I've generated for your
42:57trading idea that you came up with.
42:58>> Okay.
42:59>> So I already have a prompt ready for it.
43:02I'll just share that with Codex and it
43:04should be able to build it.
43:09>> Okay. Uh can we go through this once?
43:11>> Yes. add a signals tab to the existing
43:14zeroda dashboard which basically is what
43:16you've built just now using codeex
43:19[clears throat]
43:19>> and um and basically the other bits is
43:24uh is to do the math right
43:26>> yeah is to do the math so the first part
43:28is to fetch the symbols that is the 100
43:31symbols
43:32>> uh then to get the historical candles
43:35for it
43:36>> then calculate the SMA 6 and the SMA 30
43:39>> uh the next part is it detects the
43:41bullish and bearish crossovers and then
43:43it ranks the stocks.
43:44>> This was done earlier for the Python
43:46file.
43:46>> Python. Yeah,
43:47>> a very similar prompt for this UI.
43:49>> Just that we are now mentioning that the
43:51UI it has to be built for a UI and not
43:53for a Python file.
43:54>> Okay. Okay.
43:55>> So now I'll just run this prompt as
43:56well.
44:10Uh so yeah, it looks like Codex has
44:13finished working through this prompt as
44:14well.
44:15>> Okay,
44:15>> let's check the signals it's generated.
44:17Let's go back to the UI. Um it's added a
44:21signals tab here to the side. Let's go
44:23there.
44:24>> So yeah, it has built a short SMA, long
44:27SMA, loop back, and also max is 100.
44:30Okay.
44:31>> So, let's just see what happens when I
44:32click the button here.
44:35>> So, yeah, Karthik, you can see it has
44:37generated signals here.
44:38>> Okay.
44:39>> It has ranked it by the crossover date
44:41as well.
44:42>> Okay.
44:42>> So, this is the most recent date
44:45>> uh to the last date that is available.
44:47So, it's ranked it by the recency as
44:50well.
44:50>> Okay. A So, the algo has generated a
44:55bunch of signals.
44:57Can we physically verify the validity of
45:00any of the signals? Maybe we can pick
45:02Asian paints.
45:04>> Uh it's supposed to be a bearish signal
45:06as of today, 30th of uh June.
45:10>> Can we check it on the kite chart?
45:12>> Yeah. So, let's just navigate to our
45:14kite portal.
45:15>> Okay.
45:16>> Uh I'll just
45:19go to Asian PES.
45:22So, this is Asian PES on a daily time
45:24frame.
45:25>> Okay. I'll just add a moving average
45:27crossover indicator.
45:30So, I'll go for a moving average double.
45:32>> Okay.
45:33>> And then let's change the
45:36>> the settings
45:37>> settings. Yes.
45:41I'll make this a six
45:44and 30.
45:45>> Okay.
45:49>> Yes. So, as you can see there is a
45:51crossover.
45:52>> It has Yeah, it has got a crossover. So
45:55basically all we have to do is now that
45:57we've built the algo and put a nice UI
46:00around it,
46:02>> I just have to run this on a daily basis
46:04and it's going to scan
46:06>> and tell me where the signals are
46:09>> generated. Yeah,
46:10>> and this is exactly what you mean by
46:12semi-automated because based on the
46:14signal here,
46:16>> I still have to go back to Kite or
46:18whatever my whoever my broker is and
46:21place an order to buy or sell, right?
46:24>> So, if the [clears throat] algo was
46:27automatically placing the order bases
46:30the signal, that would have been a
46:31full-fledged uh fully automated algo.
46:34>> Yeah.
46:35>> Right. So, uh so actually this is uh
46:38super interesting.
46:39Um
46:41what else should I know at this point?
46:43>> This session we have built a signal
46:45which is an SMA crossover.
46:46>> Okay,
46:46>> you could apply this to any of your
46:49particular trading uh idea. So you could
46:52>> you could also try volume breakouts,
46:54range breakouts.
46:55>> Okay,
46:55>> any idea you basically have you could do
46:57it across uh not just on a daily time
47:00frame,
47:01>> you could also scan it across intraday
47:03time frames as well.
47:04>> Okay,
47:04>> so there you get the signals uh coming
47:06in while the market is open. Okay.
47:09>> So for traders who are intraday traders
47:11could also benefit from using APIs as
47:13well.
47:14>> Okay. Uh and I'm guessing it doesn't
47:15matter what the underlying asset is it I
47:19can pretty much use the same thing
47:20across commodities as well.
47:22>> Yes. So like [clears throat] I had
47:23showed initially in my user tab if
47:26commodities is enabled for you.
47:27>> Okay.
47:28>> Uh you could do this across commodities
47:29as well. You could do this across BC as
47:32well. So you have your NSE BC MCX.
47:35>> Yeah.
47:35>> A this was very interesting.
47:38Now let me quickly summarize. I know
47:40I've done it several times throughout
47:41the video but um but I would like that
47:44uh summary at every uh logical uh you
47:48know points. So basically I've used
47:52codeex it could be any LLM that I can
47:54use
47:55>> and the as a first step or as a part a I
48:00I set up the infrastructure needed to
48:03build my algo
48:05>> uh which was to install Python which was
48:08to install NodeJS
48:10and then as part B I connected uh I set
48:13up my kite connect API
48:15>> ensured both of them are linked uh and
48:18then started uh coding my algo using
48:22codeex uh through a prompt.
48:25>> Yeah.
48:25>> Uh but of course the the scary Python
48:28lines I didn't want to deal with. So I
48:30asked uh Codex to build a beautiful uh
48:35UI right and you said you promised me it
48:39would build a nice
48:40>> beautiful UI which it has. Uh I can I
48:43can beautify this further. I think it
48:45all depends on the user's creativity,
48:48right? And now that this algo is done,
48:51all I have to do is run this algo on a
48:53daily basis and my uh and I'll get my
48:56signals to uh track, right?
48:59>> Yeah. And and pretty much any trading
49:02idea that you have can be
49:04>> developed as an algo using natural
49:07language, right?
49:08>> Yeah.
49:09>> Yeah. And I think I [clears throat] find
49:10this idea super powerful because if
49:13because it enables non-programmers like
49:16me to actually aspire to build algos.
49:20>> Yeah.
49:20>> Yeah. So any closing thoughts?
49:23>> No, I mean uh it all it took was a
49:26natural language prompt to build
49:27something so unique.
49:29>> Yeah.
49:29>> And in particular to your trading style.
49:32>> Yeah.
49:33>> Yeah. It's I mean we have come a long
49:34way from the AGO journey
49:36>> from from having to have done all that
49:39manually by yourself.
49:40>> Yeah.
49:41>> Going through the whole entire not only
49:43did you have to be a trader earlier, you
49:44had to be a programmer as well.
49:46>> Yeah.
49:46>> Now you just need to have the idea
49:48>> and orchestrate the entire algo using
49:51codeex. So
49:53>> yeah, absolutely. This is fantastic. I
49:55hope you found this video useful. Uh for
49:58me personally, there was a lot of
49:59learning. I hope you guys also uh
50:02learned something new through this
50:04video. Uh if you want us to make more
50:07videos on ALGO, please do let us know
50:10comment and we will read your comments
50:13and take your feedback into
50:14consideration. Also, we are doing
50:17physical events on uh helping people
50:19understand how to develop ALGOS. uh
50:22please do keep a track of all the
50:25physical events uh that we are posting
50:26on our social media and if it's
50:29happening in your city please do
50:30register and attend these events. Uh
50:33hopefully we get to see you offline. By
50:36the way, we also have a Hindi and a
50:37Canada varsity channel. Please do
50:40subscribe uh and check out those
50:42channels if you prefer learning in your
50:44regional language. I'll see you guys in
50:47the next video.

About this transcript

This page contains the full transcript of Build a Trading Algo with AI | No Coding Required by Zerodha Varsity, generated from the public captions YouTube serves with the video. The transcript has 9,115 words across 1,464 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.