WEBVTT 00:00.120 --> 00:05.760 And with that, it gives me enormous pleasure to introduce you to this week's business challenge. 00:06.000 --> 00:10.880 This week, of course, all about accelerating your business or your client's business. 00:10.880 --> 00:17.000 And this is such a classic, classic use case, such as something that you can immediately go out there 00:17.000 --> 00:18.760 and implement for clients. 00:18.800 --> 00:24.680 The setup is we need to build an expert that is more than more than a chatbot. 00:24.760 --> 00:30.080 This has got to be something that can answer detailed, in-depth questions for your client about their 00:30.080 --> 00:36.480 business and the client in this case for this setup, has a lot of information about a lot of products 00:36.600 --> 00:43.240 in a big Google Sheet, huge document, and they want something that can quickly answer questions that 00:43.240 --> 00:45.600 pertain to any of this information. 00:45.760 --> 00:48.680 And of course, we'll be working with a fairly big but not massive massive. 00:48.720 --> 00:54.000 The idea is this needs to be scalable so that you could be working with so much data. 00:54.000 --> 00:59.600 You could be working with enterprise level data if you wanted, or if you're a startup, then your company, 00:59.600 --> 01:00.920 your clients data. 01:00.930 --> 01:04.810 so it could be able to work with huge quantities of data. 01:04.810 --> 01:07.410 Classic Rag business setup. 01:07.450 --> 01:14.050 Okay, so for our solution, as I mentioned yesterday, we're going to be using this managed database 01:14.050 --> 01:15.730 provider called Superbase. 01:15.930 --> 01:17.610 There are easier ways we could do this. 01:17.610 --> 01:22.930 There are built in databases and with with nodes that make life much easier for us. 01:22.970 --> 01:25.610 And I've made life a bit harder for us by choosing Superbase. 01:25.610 --> 01:29.010 And I had a reason for that, which is that Superbase is very popular. 01:29.010 --> 01:33.370 It's a great third party database, and it's something that many other people use. 01:33.370 --> 01:38.730 And so by giving you the skills to integrate with Superbase, it's like a big tick in the integrations 01:38.730 --> 01:41.010 list for you to have worked with Superbase. 01:41.210 --> 01:45.650 And it means that it's another set of credentials that will be set up, and it's something that you'll 01:45.650 --> 01:48.370 be able to use and extend for many other purposes. 01:48.370 --> 01:51.250 So I think it's an important infrastructure step. 01:51.250 --> 01:56.450 And that's why I picked a slightly harder challenge in saying we would integrate with a Postgres database 01:56.450 --> 01:57.330 on Superbase. 01:57.330 --> 02:02.420 It has one downside, which is there is going to be one point where we have to run some code, which 02:02.420 --> 02:08.660 I'm going to give you and which you will execute in supabase to set up the database, and you will be 02:08.660 --> 02:11.580 in your rights to think, ah, this is meant to be a low code. 02:11.580 --> 02:12.740 No code course. 02:12.900 --> 02:14.820 How would I be able to do this myself? 02:14.820 --> 02:16.020 And I will talk you through that. 02:16.020 --> 02:17.940 At the time, I'm not going to talk through the code. 02:17.940 --> 02:21.060 I'm going to explain how you could generate this yourself. 02:21.220 --> 02:25.340 And it's super easy and how you could handle things like problems with it. 02:25.340 --> 02:28.780 But that's so that's just going to be an obstacle for us to get through. 02:28.780 --> 02:31.940 But apart from that, Supabase is going to be fabulous. 02:31.940 --> 02:36.340 And it's certainly very much a heavyweight, scalable solution. 02:36.340 --> 02:41.100 And when I say heavyweight, I mean in the kind of like enterprise grade sense, it's actually quite 02:41.100 --> 02:43.020 a lightweight, easy to use platform. 02:43.020 --> 02:44.340 Supabase I think you're gonna like it. 02:44.580 --> 02:45.220 All right. 02:45.260 --> 02:48.420 Today is all about data ingest. 02:48.420 --> 02:53.060 And look, in one way, I'm going to dive in a bit deeper and build up some of your skills. 02:53.060 --> 03:00.180 We're going to do some data transformation using a node in an called edit fields used to be called set. 03:00.300 --> 03:00.580 Uh. 03:00.580 --> 03:04.900 And it's such a crucial node and this is really good core skills. 03:04.900 --> 03:08.140 So I've I've done well there in another way. 03:08.140 --> 03:14.820 I'm taking a bit of a shortcut in that the source of data we're going to use is going to be a Google 03:14.820 --> 03:15.460 sheet. 03:15.460 --> 03:17.980 And you know we've done this a few times now. 03:17.980 --> 03:20.060 We've used Google Sheets as our input data a lot. 03:20.060 --> 03:22.460 And you might be thinking, ah, that's a bit boring. 03:22.620 --> 03:24.740 Like I was originally thinking. 03:24.740 --> 03:26.340 It would be pretty cool to have it. 03:26.340 --> 03:32.420 So there'd be a Google Drive folder, and every time you drop documents into that Google Drive folder, 03:32.420 --> 03:35.780 it would automatically kick off our data ingest pipeline. 03:35.780 --> 03:37.940 And so that would indeed be cooler. 03:38.020 --> 03:43.620 Uh, but it turns out there is a little bit of lift associated with that particularly there's some credentials 03:43.700 --> 03:44.780 authentication stuff. 03:44.780 --> 03:49.780 It's a big OAuth two thing which which I think would be too much of a distraction right now. 03:49.820 --> 03:51.700 We're going to do all of that next week. 03:51.700 --> 03:54.300 We're going to set it up so that you can do that. 03:54.300 --> 04:00.860 And so you could absolutely then come back and beef this up so that it's just got that extra step that 04:00.860 --> 04:06.430 you don't just update a Google sheet, but rather you can just drop a document in a Google Drive in 04:06.430 --> 04:10.270 a folder and that would automatically kick off your workflow. 04:10.310 --> 04:11.870 And I do think that would be quite cool. 04:11.870 --> 04:16.950 So so I would encourage you, after we've, we've built up those skills next week, that you could come 04:16.950 --> 04:22.070 back and further improve this, but for now, it seemed to me like that was it was too much of a distraction. 04:22.070 --> 04:27.950 It would be like a one hour sidebar to get all of that set up, and I would rather we focus on Rag and 04:27.950 --> 04:29.030 get this thing done. 04:29.030 --> 04:33.270 So the sort of shortcut is that our source of data is going to be Google Sheet, which we already know 04:33.270 --> 04:33.910 how to do. 04:33.910 --> 04:35.390 So that part's going to be easy. 04:35.390 --> 04:40.830 And we're going to be spending our main time focusing on the Agentic rag build out. 04:40.830 --> 04:46.230 And tomorrow we will also, of course, add in the question answering part of this. 04:46.230 --> 04:51.070 So that we've done the data in yesterday, we've added the question answering tomorrow and that will 04:51.070 --> 04:54.510 complete a genetic rag, a commercial build. 04:54.830 --> 04:57.790 And for sure we're going to add in a voice agent tomorrow. 04:57.790 --> 04:59.430 And then it's also going to connect with you. 04:59.430 --> 05:06.000 I think it will land as to why I've talked about building the the business logic in N810 and then having 05:06.000 --> 05:10.240 11 labs be built as the voice agent that drives the workflow. 05:10.280 --> 05:11.800 I think that's going to make a lot of sense. 05:12.040 --> 05:16.600 We're doing that that second approach for how we integrate 11 labs with N810. 05:16.680 --> 05:20.040 The N810 workflow focuses on the business logic. 05:20.400 --> 05:23.240 All right, enough with the talk. 05:23.560 --> 05:24.720 Time for the doing. 05:24.920 --> 05:28.560 Let's go and set up our data ingest. 05:28.600 --> 05:29.080 Okay. 05:29.080 --> 05:33.280 Let me start by introducing you to some data. 05:33.600 --> 05:36.920 The data that is sitting right here this. 05:37.120 --> 05:40.320 This is a Google sheet that your client has shared with you. 05:40.440 --> 05:41.520 In our example. 05:41.520 --> 05:48.720 And it's a sheet full of the products that they sell in their online store selling computer accessories. 05:48.920 --> 05:55.960 And you can see the names of the products, their their categories, their SKUs, uh, and the price 05:55.960 --> 05:59.120 and what they, what they are description. 05:59.360 --> 06:00.800 And there's a bunch of them. 06:00.800 --> 06:02.290 There is, in fact, not that many. 06:02.330 --> 06:06.810 There's 60 because I don't want to, like, like, uh, do something that's super. 06:07.050 --> 06:08.970 That's expensive in any way. 06:09.210 --> 06:11.050 I want to give this as an example. 06:11.050 --> 06:16.450 The point I want to show you is that this is 60, but it could easily be 60,000. 06:16.490 --> 06:19.930 What we're going to do is going to be incredibly scalable. 06:19.930 --> 06:21.210 And that's the point. 06:21.210 --> 06:25.690 And should you wish to, to generate more data or use your own, I would love that. 06:25.730 --> 06:27.370 That's very much to be encouraged. 06:27.410 --> 06:31.930 We're using Google Sheets for this, as I say, because it's something we've already integrated with. 06:31.930 --> 06:34.210 So it's super easy for us to pull in this data. 06:34.570 --> 06:41.010 There would of course, be more, uh, definitely a more advanced mission for you would be to try an 06:41.010 --> 06:46.690 already solve that that next step problem of have it be any document dropped in a shared Google Drive. 06:46.690 --> 06:47.770 That would be really cool. 06:47.810 --> 06:51.130 But it's also great just to imagine you've been given this sheet. 06:51.130 --> 06:52.970 It's the kind of thing that really happens. 06:52.970 --> 06:54.810 And so that's what we're working with. 06:54.850 --> 06:56.530 This is the data we've got. 06:56.530 --> 07:02.420 And we should get used to it, because we're going to be reading it in and analyzing it and doing stuff 07:02.420 --> 07:03.140 with it. 07:03.580 --> 07:05.260 Let's go and get started with that right now. 07:05.380 --> 07:06.020 Okay. 07:06.060 --> 07:08.740 So we're going over to Nate and I. 07:09.340 --> 07:11.340 And we're going to press the sign in button. 07:11.340 --> 07:13.820 Come into our workspace here. 07:14.460 --> 07:17.420 And here we are in the home screen. 07:17.620 --> 07:21.660 And we're going to create a brand new workflow. 07:22.220 --> 07:22.940 Okay. 07:23.300 --> 07:26.100 So first of all what's going to trigger this workflow. 07:26.140 --> 07:29.380 Well there are a lot of interesting things we could do to trigger this workflow. 07:29.420 --> 07:34.420 Like when a sheet changes when a file is dropped in the Google Drive, I'm going to keep it quite boring 07:34.420 --> 07:35.620 and trigger it manually. 07:35.660 --> 07:40.300 We're going to have plenty of time to add more interesting triggers in the future, but we're focusing 07:40.300 --> 07:43.860 on Rag today, so we're just going to focus on the data pipeline. 07:43.860 --> 07:44.460 But feel free. 07:44.460 --> 07:48.580 You can make it so that when you send a slack message or a telegram message, that's what wakes it up. 07:48.580 --> 07:54.500 Whatever we want, this is going to kick off our data, ingest pipeline, load in the data from that 07:54.500 --> 07:58.460 Google sheet and put it in super base in a vector store. 07:58.580 --> 07:59.980 That's what we're building right now. 08:00.020 --> 08:05.940 And at this point I'm, I'm going to to roll up my sleeves because we're about to get deep into some 08:05.940 --> 08:08.100 data, which I always enjoy doing. 08:08.100 --> 08:09.020 And so should you. 08:09.180 --> 08:13.260 Uh, if you're wearing a short sleeve t shirt, then you can just imagine you can just go through the 08:13.260 --> 08:17.260 motion as we get into our data set. 08:17.260 --> 08:22.340 And we're going to start by having a Google Sheets in here. 08:22.340 --> 08:25.420 We are going to want to read our rows. 08:25.660 --> 08:28.380 Uh, so that is get rows in sheets. 08:28.700 --> 08:29.660 If you remember this. 08:29.660 --> 08:31.380 Well we're going to get rows. 08:31.380 --> 08:37.620 I'll, I'll fill this in so that we can pull in all of the stuff in that product sheet. 08:37.660 --> 08:38.060 Here we go. 08:38.060 --> 08:39.980 So I've selected products from here. 08:39.980 --> 08:42.260 And now it's obviously there is only one sheet. 08:42.260 --> 08:43.220 It's sheet one. 08:43.460 --> 08:46.340 And this all sounds great okay. 08:46.380 --> 08:48.820 That is going to get the rows from our Google Sheet. 08:48.820 --> 08:50.620 Let's press Execute Workflow. 08:50.620 --> 08:55.540 And if you've got this this manual when clicking on Execute workflow this manual trigger you just press 08:55.540 --> 08:56.780 the button and it runs. 08:56.780 --> 08:57.820 And we've just loaded this in. 08:57.820 --> 09:01.380 And so if we double click here you can see that it's run. 09:01.380 --> 09:05.190 And you see here an example the 60 items that were indeed 60 items. 09:05.190 --> 09:06.830 And this would be one of them. 09:06.990 --> 09:10.590 Uh, the Nova key tactile keyboard coming in like this. 09:10.790 --> 09:13.550 Uh, so, uh, great. 09:13.590 --> 09:15.030 We've brought in our data. 09:15.070 --> 09:15.910 What next? 09:15.950 --> 09:22.470 Well, this, of course, is the extract stage of ETL extract transform load, which means the next 09:22.470 --> 09:24.310 step is transform. 09:24.310 --> 09:29.190 And so we want to just massage this data to be in the format that's going to make most sense for what 09:29.190 --> 09:30.150 we have to do. 09:30.350 --> 09:33.150 And to do that we need to look up a node. 09:33.150 --> 09:36.750 In the previous version of this used to be called set. 09:36.950 --> 09:42.230 In the new version it is called edit Fields and look they put set in brackets afterwards because that's 09:42.230 --> 09:42.910 what it used to be called. 09:42.910 --> 09:44.270 This used to be the set node. 09:44.310 --> 09:46.990 Now it's the edit fields node. 09:47.310 --> 09:52.030 And so what we see here is that on the left we've got the data coming in. 09:52.190 --> 09:58.590 Uh and on the right is whatever we're going to massage it to uh, which would be the best possible format 09:58.590 --> 09:59.910 for our knowledge base. 09:59.910 --> 10:02.790 And so that that mapping is what we're going to do now.