WEBVTT 00:00.080 --> 00:03.360 Okay, so here we have our full workflow. 00:03.360 --> 00:04.720 Let's tidy it up. 00:05.200 --> 00:07.680 Uh, this thing has a red box here. 00:07.720 --> 00:09.760 Parameter table name is required. 00:09.760 --> 00:11.600 That does sound sensible, doesn't it? 00:11.800 --> 00:13.120 Uh, back we come. 00:13.160 --> 00:15.680 We need to tell it which table this is going in. 00:16.240 --> 00:18.680 You probably spotted that intentional uh, miss there. 00:18.800 --> 00:24.280 Uh, so we come back here to the super base vector store thing, and of course, it's asking, uh, 00:24.320 --> 00:26.560 what what table name should we pick here? 00:26.560 --> 00:32.040 We right click here and we'll see that knowledge base is the table that we created. 00:32.040 --> 00:35.200 That's the thing that our SQL script created called Knowledge Base. 00:35.200 --> 00:36.720 That's exactly what we want. 00:36.840 --> 00:37.440 Very good. 00:37.440 --> 00:39.040 Thank you for prompting me with that. 00:39.360 --> 00:45.800 And here we have our complete, uh, our workflow that's ready for data ingest. 00:45.840 --> 00:50.680 We should probably rename it to something that sounds important, like data ingest. 00:52.600 --> 00:55.360 And we should be ready to, uh. 00:55.360 --> 00:57.720 Now, give this a try. 00:57.760 --> 01:02.600 Let's just go back quickly to super base, uh, and go to our database table. 01:02.840 --> 01:07.240 Uh, and so we go to database over on the left here loading tables. 01:07.880 --> 01:08.600 Here it comes. 01:08.600 --> 01:09.440 Here's the knowledge base. 01:09.440 --> 01:12.600 Let's open that by clicking on that button on the top right here. 01:13.040 --> 01:14.720 And the table is empty. 01:14.760 --> 01:15.800 There's no data. 01:15.840 --> 01:19.360 Our knowledge base is empty and so should yours be. 01:19.400 --> 01:20.320 Let's go back over here. 01:20.320 --> 01:23.200 Let's hide the logs, give us a nice clean screen. 01:23.200 --> 01:27.960 And I'm simply going to press the Execute workflow button to kick this whole thing off. 01:27.960 --> 01:29.040 All right here we go. 01:29.080 --> 01:30.240 Execute workflow. 01:30.680 --> 01:32.480 It's loading in the 60 items. 01:32.480 --> 01:33.560 It's embedding them. 01:33.560 --> 01:38.560 It's working with super bass and tick workflow executed successfully. 01:38.600 --> 01:43.800 See the number 60 down there telling us that 60 documents were loaded. 01:43.840 --> 01:45.480 They've been vectorized. 01:45.520 --> 01:46.320 They've been chunked. 01:46.320 --> 01:48.480 Although we didn't actually need to chunk anything. 01:48.560 --> 01:53.920 Uh, each row has turned into a vector, has been put in super base. 01:54.200 --> 01:55.560 Everything is green. 01:55.840 --> 02:00.920 This would have been potentially a week of work or something if you were trying to code it, trying 02:00.920 --> 02:02.440 to stitch together all the pieces. 02:02.440 --> 02:07.840 But thanks to N810 and Super Base, we did this in a matter of minutes. 02:08.360 --> 02:09.080 Or did we? 02:09.600 --> 02:12.000 The way to find out is to go and look at the database. 02:12.000 --> 02:13.110 Okay, here we are. 02:13.150 --> 02:15.910 Our rag project that it says the table is empty. 02:15.910 --> 02:16.870 That's because it doesn't refresh. 02:16.990 --> 02:18.790 Oh, it did refresh automatically. 02:19.070 --> 02:20.750 Ah, here we go. 02:20.910 --> 02:21.830 Here it is. 02:22.070 --> 02:23.870 This is our knowledge base. 02:23.870 --> 02:25.350 What have we got here, everybody? 02:25.390 --> 02:26.270 Let's see. 02:26.470 --> 02:30.030 So we have, uh, we have 60 items. 02:30.150 --> 02:37.190 I think you'll find, uh, and, uh, in these 60 items, each one of them has an ID, it has content, 02:37.190 --> 02:43.430 which is text, it has metadata, and it has embedding, which is a bunch of numbers. 02:43.630 --> 02:45.390 If I double click on this, let's have a look. 02:45.430 --> 02:47.150 This is one of the contents. 02:47.150 --> 02:48.790 It is the stuff that you remember. 02:48.790 --> 02:50.310 We wrote this very nicely. 02:50.310 --> 02:55.990 This is stuff which would be useful to give an LLM should there be a question that is related to it. 02:56.230 --> 03:06.230 Uh, and uh, over here is, uh, um, here is the, the, uh, the details of the metadata, which 03:06.230 --> 03:09.070 has a, a category in it. 03:09.470 --> 03:11.270 There you see it, there's the category. 03:11.310 --> 03:13.670 It got added into the metadata just as we wanted. 03:13.670 --> 03:18.430 There's also some some metadata that is automatically added in there. 03:18.430 --> 03:20.190 And over here is the vector. 03:20.190 --> 03:23.150 I almost double click on this and escape out of that. 03:23.150 --> 03:27.150 This would have the the 1536. 03:27.190 --> 03:27.790 Is that right? 03:28.110 --> 03:37.590 1536 numbers, which represents a point in that 1536 dimensional space that best reflects the meaning 03:37.590 --> 03:38.870 of this content. 03:38.870 --> 03:42.110 But while I was showing you that, I realized I made a silly mistake. 03:42.110 --> 03:43.190 Just a typo. 03:43.390 --> 03:46.710 But it gives a chance for me to quickly show how easy it is to do this. 03:46.710 --> 03:48.550 So I'm going to embrace the typo. 03:48.710 --> 03:49.990 You may have noticed this yourself. 03:49.990 --> 03:51.750 I bet you didn't make this mistake in here. 03:51.750 --> 03:52.830 This looks all very nice. 03:52.830 --> 03:55.150 Fancy text that we constructed for the LM. 03:55.190 --> 04:00.270 Says the price is $1.59 .99, and somehow I put in an extra dollar sign there as well. 04:00.270 --> 04:02.470 The description is dollar vertical design. 04:02.470 --> 04:06.150 And if we look at another one here you'll see the description is dollar compact. 04:06.390 --> 04:06.990 Oops. 04:07.030 --> 04:14.510 Well luckily this is so simple just to rerun it thanks to the the how easy and streamlined these workflows 04:14.550 --> 04:14.950 are. 04:15.070 --> 04:19.790 So first up we need to, uh, probably delete what we've got here. 04:19.790 --> 04:20.670 Let's just start again. 04:20.670 --> 04:24.230 So I'm just going to click here and I'm going to go over here and press delete. 04:24.230 --> 04:26.070 60 rows can't be undone. 04:26.070 --> 04:26.710 Oh, well. 04:27.110 --> 04:28.150 Off it goes. 04:28.190 --> 04:29.150 The 60 rows are gone. 04:29.150 --> 04:31.390 We've now got an empty database again. 04:31.550 --> 04:34.470 Let's come back to Edit fields. 04:34.470 --> 04:38.030 If I double click on here, you will see here I'm an idiot. 04:38.070 --> 04:40.270 Look, there's an extra dollar sign in there. 04:40.270 --> 04:41.950 The description as a dollar sign. 04:41.990 --> 04:48.550 I don't know how I did that, but I will remove it, I will save, I will come back to here again and 04:48.550 --> 04:49.630 then we will kick this off. 04:49.670 --> 04:55.630 It's also a nice point to mention that yellow line and that triangle there is telling us that that this 04:55.630 --> 05:02.470 node has changed since we last ran it as the as the pop up tells you, uh, so that that the things 05:02.470 --> 05:03.750 aren't consistent right now. 05:03.750 --> 05:06.350 So I press this again to run it again. 05:06.630 --> 05:07.950 Everything runs. 05:08.310 --> 05:11.270 This is again putting it in the vector data store. 05:11.310 --> 05:12.910 60 things are in there. 05:12.910 --> 05:13.950 Fabulous. 05:14.070 --> 05:17.830 Let's go back here, see if it refreshes automatically like it did last time. 05:17.830 --> 05:18.830 Yes it does. 05:18.830 --> 05:19.630 Here it is. 05:19.670 --> 05:24.430 Let's double click on one of these and the dollar sign is gone. 05:24.430 --> 05:25.750 No more nonsense. 05:25.750 --> 05:27.430 We've got good data here. 05:27.590 --> 05:33.950 And again it's it's just terrific that it's so easy to rerun the pipeline with a small mistake like 05:33.950 --> 05:35.590 that fixed quickly. 05:35.630 --> 05:36.230 Excellent. 05:36.230 --> 05:37.590 We're in great shape. 05:37.590 --> 05:40.550 We've built our data, ingest pipes, and they work. 05:40.590 --> 05:46.590 And so to wrap up first, to put this in front of you again, we talked about the two different phases 05:46.590 --> 05:48.390 the data ingest that they're doing today. 05:48.390 --> 05:51.350 And the pipeline question answering will do tomorrow. 05:51.550 --> 05:56.390 And in the steps of that ingest, the data source represents the fact that we had a Google sheet with 05:56.390 --> 06:03.390 product data in it extract that was the node in N810 that that connected to Google and pulled in the 06:03.390 --> 06:05.630 Google Sheet transform. 06:05.790 --> 06:12.110 That was the the field, the node, the field settings node that did a field mapping and allowed us 06:12.110 --> 06:17.750 to create the new thing with the content and the category as the output, we transform the data. 06:17.830 --> 06:19.030 Fairly simple transformation. 06:19.030 --> 06:22.630 It can be a lot more involved than that, but it was a good way to learn about it. 06:22.630 --> 06:29.710 And then chunking and vectorizing that happened because of those two extra sub nodes that we had there. 06:29.750 --> 06:32.910 Did you notice that that whole thing was was another node cluster? 06:32.950 --> 06:35.270 The AI agent is a node cluster. 06:35.270 --> 06:38.380 And then the thing we had to to connect to super base. 06:38.380 --> 06:40.260 That was another node cluster. 06:40.460 --> 06:46.500 Uh, and then we did the loading and put it in super base in a Postgres database that we set up in Super 06:46.500 --> 06:46.860 Base. 06:46.940 --> 06:50.980 So that was the whole of the flow of our data pipes. 06:51.100 --> 06:55.780 And we have a dedicated workflow to handle our data pipes. 06:55.780 --> 06:57.900 And it's triggered right now by pressing a button. 06:57.900 --> 07:00.700 And you could turn that to be something that's triggered on a schedule. 07:00.700 --> 07:03.900 You could do it with something that's triggered when the spreadsheet changes. 07:03.900 --> 07:08.380 You can experiment with different triggers to see what could kick off your data pipes. 07:08.420 --> 07:08.900 All right. 07:08.900 --> 07:15.260 And that is a wrap on data ingest and connecting to super base and working with vectors. 07:16.100 --> 07:19.420 It was uh, hopefully it was uh, it was it was pretty satisfying. 07:19.420 --> 07:25.820 But I think that definitely we hit the bullseye tomorrow when we pull this together and we complete 07:25.820 --> 07:33.380 our project for week two, all about accelerating, building with voice agents and with Rag. 07:33.380 --> 07:41.460 And that brings us to to the final slide, to say that you have reached the 60% point 60% tomorrow. 07:41.500 --> 07:42.740 Conclusion of week two. 07:42.780 --> 07:43.620 You don't want to miss it.