WEBVTT 00:00.120 --> 00:02.680 So here we are in the edit fields node. 00:02.680 --> 00:05.640 And and this is such a core node. 00:05.640 --> 00:11.320 This node used to be called the set uh node is this is where a data engineer lives. 00:11.320 --> 00:14.040 This is all about mapping data from one format to another. 00:14.080 --> 00:17.960 It's got nothing to do with AI, although of course working with data is all about AI. 00:18.120 --> 00:23.800 But, uh, but it's absolutely crucial for this kind of, of data engineering work. 00:24.000 --> 00:27.920 And you can see that there are some, some special, clever ways to do it with JSON. 00:27.920 --> 00:32.440 But we're going to do manual mapping, which I think is most most common. 00:32.440 --> 00:39.680 And I'm going to to explain that what we're going to want to do is create a data set which has two fields, 00:39.880 --> 00:45.120 a field called content, which is going to contain text, which is going to be best to go in our knowledge 00:45.120 --> 00:47.760 base language, which can be vectorized. 00:47.760 --> 00:52.160 And then another field called category which will say, what category of thing is this? 00:52.160 --> 00:56.440 And that's because that feels like that's useful metadata that we'd like to tag everything with. 00:56.440 --> 00:57.920 However it gets vectorized. 00:58.040 --> 01:02.320 Um, and that's just we're not we're probably not going to make huge use of that, but it's a great 01:02.320 --> 01:07.880 practice to have that kind of metadata, and it could be used to filter on the results in different 01:07.880 --> 01:08.320 ways. 01:08.640 --> 01:12.760 So add field I press and this allows me to to come up with with some fields. 01:12.760 --> 01:15.360 And I said we're going to have one field that's called content. 01:15.360 --> 01:18.880 And we're going to have another field that's called category. 01:18.880 --> 01:25.080 So that everything that comes out, every every record, every record in the inputs, every, every 01:25.080 --> 01:31.760 one of these items is going to end up with a content and a category being built on the right hand side. 01:32.000 --> 01:32.280 Okay. 01:32.280 --> 01:33.680 Let's do content first. 01:33.720 --> 01:34.160 Okay. 01:34.200 --> 01:39.560 So again we're writing here the what should go in a field called content which is going to come over 01:39.560 --> 01:40.000 here. 01:40.000 --> 01:41.840 And we can choose what type it is. 01:41.880 --> 01:48.720 Is it a string a number true or false, an array a list, an object or just some some something else, 01:48.760 --> 01:51.120 some something like a PDF document or something. 01:51.320 --> 01:53.760 And we are going to say string. 01:53.760 --> 01:58.240 And what's going to go in here is going to be the text associated with this information. 01:58.520 --> 02:00.160 And, and really this is up to us. 02:00.160 --> 02:05.750 We want to describe something that if this ends up going to an MLM, it will be it will be useful. 02:05.750 --> 02:10.870 So we might as well just try and turn this information here into something which will be handy. 02:10.870 --> 02:14.710 So we will we will say something like a product name colon. 02:14.870 --> 02:18.790 And now we want and sorry I want this to be an expression not not fixed. 02:19.030 --> 02:20.470 So product name. 02:20.590 --> 02:22.790 And then we will have the name of the product. 02:22.790 --> 02:24.470 So remember how we're going to do this. 02:24.510 --> 02:28.630 It's dollar JSON dot name. 02:28.950 --> 02:33.030 And that you can see in the result below that says product name colon. 02:33.030 --> 02:34.990 And then the the keyboard. 02:35.030 --> 02:38.910 All right let's put an empty line let's say category. 02:40.350 --> 02:42.830 And let's put in the category dollar Jason. 02:44.190 --> 02:49.190 And then category spelt with a capital C because that's how it is in the table that's coming in. 02:49.750 --> 02:51.030 And then we will. 02:51.070 --> 02:51.390 Why not. 02:51.430 --> 02:57.630 We'll just we'll just put everything basically we'll say this the SKU which is dollar Jason dot SKU 02:59.510 --> 03:01.670 and then we'll give it the price. 03:05.020 --> 03:08.660 Dollar Jason Price. 03:08.900 --> 03:10.100 Now let's see how that comes out. 03:10.140 --> 03:14.860 Do you see if you look, if you look down at the at the result, you'll see the prices come there just 03:14.860 --> 03:17.020 as it came in without a dollar sign. 03:17.020 --> 03:18.980 Because that's how it is in the table. 03:18.980 --> 03:25.180 And maybe it will be helpful for us to put a dollar sign there so that that's what goes to the model. 03:25.180 --> 03:30.180 And it knows in this content that that it's getting something that is that is a dollar value. 03:30.500 --> 03:31.700 Uh, okay. 03:31.860 --> 03:37.500 And then finally and most importantly description, uh, spell it right. 03:37.500 --> 03:38.700 Description. 03:38.860 --> 03:46.980 And this is where I'm going to put in the, um, the description of it, of course coming in from the 03:46.980 --> 03:47.660 input. 03:47.820 --> 03:48.780 There it goes. 03:49.060 --> 03:49.700 Okay. 03:49.820 --> 03:52.300 And so this is a field called content. 03:52.300 --> 03:57.100 And this is what, what what it evaluates to it comes to this bunch of text down here. 03:57.380 --> 04:01.380 Now to someone that's used to data engineering and mapping things carefully, this might seem a bit 04:01.380 --> 04:02.020 janky. 04:02.100 --> 04:07.300 We're just sort of describe some text here, and I came up with product name because I thought it sounded 04:07.300 --> 04:08.700 good for an LLM. 04:08.860 --> 04:14.900 And and this is really the sort of where data data engineering meets data science, where it becomes 04:14.940 --> 04:16.220 more of an AI concern. 04:16.220 --> 04:18.020 We want to try and come up with a string here. 04:18.020 --> 04:21.620 That's going to be most reasonable if that's sent to an LLM. 04:21.700 --> 04:26.380 So it gives it the most relevant context to answer a question about this product. 04:26.380 --> 04:29.340 And so, you know, there's no right or wrong way of doing it. 04:29.340 --> 04:30.660 You want to experiment. 04:30.700 --> 04:36.300 You want to try different ways that you can make this content be rich for the LLM and get that right. 04:36.300 --> 04:41.900 And the best way to find out is to experiment, trial and error until you get good outcomes. 04:41.900 --> 04:45.460 But this seems like a perfectly reasonable starting point to me. 04:45.500 --> 04:45.900 Okay. 04:45.940 --> 04:51.340 And now next up we've got this category field that I have a feeling is going to be useful. 04:51.340 --> 04:56.780 And we want that to be basically the contents of this field right here, which is also called category, 04:56.780 --> 04:59.500 but with a capital C, because that's how it is in the table. 04:59.780 --> 05:00.340 Okay. 05:00.540 --> 05:03.140 So uh, we should be able to do this quite easily. 05:03.140 --> 05:06.620 Now we click on expression and we could drag and drop from here. 05:06.620 --> 05:08.420 But now we're getting more familiar with this. 05:08.460 --> 05:12.660 We know it's JSON dot category with a capital C. 05:12.940 --> 05:13.660 There it is. 05:13.660 --> 05:16.620 Keyboard that seems to work perfectly. 05:16.820 --> 05:22.820 We've got ourselves a mapping we have mapped from this stuff here on the left to something new, which 05:22.820 --> 05:25.620 is going to have two fields content and category. 05:25.620 --> 05:30.980 Content is going to be some text that we feel is going to be useful for an LLM category is going to 05:30.980 --> 05:34.140 be whether it's a keyboard, whether it's a monitor or whatever it happens to be. 05:34.460 --> 05:39.180 All right, let's give this a try so we can simply press the execute step button right here. 05:39.180 --> 05:42.300 Just press that and look what happens on the right. 05:42.300 --> 05:45.740 It shows us what's the result of mapping this. 05:45.740 --> 05:47.940 Through this we get this. 05:47.980 --> 05:50.260 We get something that's perfectly civilized. 05:50.420 --> 05:53.580 Uh, product name, category, keyboard, price, description. 05:53.580 --> 05:56.700 That's what we'll be providing to the LLM as context. 05:56.700 --> 06:00.580 And we also have this useful extra piece of information category is keyboard. 06:00.900 --> 06:01.860 Excellent. 06:01.900 --> 06:04.060 We have just done some data mapping. 06:04.100 --> 06:07.530 And if I now go back to the prior screen and I press Execute Workflow. 06:07.570 --> 06:08.810 It does it for the whole lot. 06:08.850 --> 06:16.650 All 60 rows have just been mapped to the new data format, and that is the beginning of some data pipes. 06:16.690 --> 06:17.370 Okay. 06:17.410 --> 06:20.970 At this point I'd like you please to go to Super Bass. 06:21.250 --> 06:27.130 This is Super bass.com, and if you haven't already, please set up an account. 06:27.130 --> 06:29.410 You need to to create an organization. 06:29.410 --> 06:32.450 As I said yesterday my mine is called Donna Research. 06:32.490 --> 06:34.010 Feel free to call it whatever you want. 06:34.050 --> 06:35.490 Don't call it Donna research though. 06:36.930 --> 06:41.410 Uh, and then once you've set it up, you should find yourself pretty quickly at like a projects page. 06:41.410 --> 06:42.770 You won't have any projects yet. 06:42.890 --> 06:47.410 If not, then click on this projects in the left nav or click on the name of your organization and come 06:47.410 --> 06:47.770 here. 06:48.250 --> 06:52.690 You should be able to find your way to this screen, where there should be a new project button that 06:52.690 --> 06:58.330 you can press and come on in and let's call the project right as the name of the project. 06:58.370 --> 07:00.090 The compute size can be tiny. 07:00.090 --> 07:02.090 You should be able to be on the free plan. 07:02.090 --> 07:04.810 If you're on a paid plan, then this costs a cent an hour. 07:04.810 --> 07:06.530 And of course you can bring this down. 07:06.650 --> 07:09.730 Delete it after we're done with this activity. 07:10.050 --> 07:11.450 Um, and I will. 07:11.490 --> 07:14.490 Then I will come up with a database password region. 07:14.490 --> 07:15.730 You should pick the region closest to you. 07:15.770 --> 07:18.730 Mine would normally be in the Americas, but I am in the UK today. 07:19.170 --> 07:21.610 So Europe is, I guess, right for me right now. 07:21.850 --> 07:28.170 Uh, but I think I'm going to go Americas anyway because I probably want to be closest to where my An 07:28.210 --> 07:28.770 is running. 07:28.890 --> 07:33.250 Um, and, uh, once you're ready with that, we'll press create new project. 07:33.290 --> 07:33.730 Okay. 07:33.770 --> 07:35.850 So I've just put in a strong password there. 07:35.850 --> 07:40.530 I've kept a record of my password, but safely in a password manager, as should you. 07:40.730 --> 07:43.210 And now I'm pressing Create New Project. 07:43.210 --> 07:46.090 And it's telling me that that because I'm on a paid plan, it will increase my costs. 07:46.090 --> 07:46.370 But. 07:46.370 --> 07:51.610 But I think this should be something that will be free for you as part of the the trial free account. 07:51.810 --> 07:54.370 And it's been set up and here we go. 07:54.570 --> 07:55.930 Uh, it's called rag. 07:55.970 --> 07:56.890 It's getting set up. 07:56.890 --> 08:01.090 It will take a few minutes while it while it gets things together, including your API endpoints. 08:01.090 --> 08:03.450 This is terminology that you're so familiar with. 08:03.610 --> 08:08.610 And then we will be the proud owners of our first super base project.