1 00:00:03,300 --> 00:00:12,570 Okay, so in this blog we are going to work in the same way we work in some blocks before we are going 2 00:00:12,570 --> 00:00:15,090 to divide the screen into areas. 3 00:00:15,090 --> 00:00:22,470 In one area, we are going to see the slides that you have attached, and in the other we are going 4 00:00:22,470 --> 00:00:26,640 to, uh, take a look at the notebooks we have prepared for you. 5 00:00:26,640 --> 00:00:29,850 And as you as you will see, they are connected. 6 00:00:29,850 --> 00:00:30,630 Right. 7 00:00:30,630 --> 00:00:35,040 So we have divided this block in two lessons. 8 00:00:35,040 --> 00:00:37,710 The first lesson is more introductory. 9 00:00:37,710 --> 00:00:43,980 The second lesson it goes more in detail, uh, around lambda index. 10 00:00:43,980 --> 00:00:44,490 So. 11 00:00:45,630 --> 00:00:49,650 In this one we have a little introduction. 12 00:00:49,650 --> 00:00:57,600 So remember that in his first version, Lamar Index, uh, is a framework less generalistic than long 13 00:00:57,600 --> 00:00:58,080 chain. 14 00:00:58,680 --> 00:01:03,660 It aims to do fewer things, but do them better. 15 00:01:04,400 --> 00:01:09,950 It specialises in generating possibilities for professional drag applications. 16 00:01:09,950 --> 00:01:14,150 So the initial name of llama index. 17 00:01:14,790 --> 00:01:17,670 Was GPT index. 18 00:01:18,030 --> 00:01:25,710 So remember that in the context of LM applications, when we use the word index we are referring to 19 00:01:25,710 --> 00:01:26,820 indexing. 20 00:01:27,180 --> 00:01:36,720 And when you name your company GPT index what you are referring is okay I'm going to use GPT meaning 21 00:01:36,720 --> 00:01:38,550 the new LM models. 22 00:01:38,730 --> 00:01:45,390 As you know the the the top LM models are they come from the GPT family. 23 00:01:45,390 --> 00:01:49,590 So chat, GPT and many other LM models. 24 00:01:49,590 --> 00:01:53,370 They come from a base LM model called GPT. 25 00:01:53,490 --> 00:02:02,460 So when you name your company GPT index, what you are referring to is okay, I'm going to use GPT. 26 00:02:02,490 --> 00:02:07,320 So LM models, the new LM models in order to index. 27 00:02:07,320 --> 00:02:11,310 So in order to search into data. 28 00:02:12,640 --> 00:02:17,500 So you will see that in the next lesson. 29 00:02:17,500 --> 00:02:24,370 When we go in detail around la mendes, this initial name makes a lot of sense. 30 00:02:25,250 --> 00:02:27,440 So a layman. 31 00:02:27,440 --> 00:02:31,760 This is still very young and a small company. 32 00:02:32,630 --> 00:02:37,100 It still needs to find its strategic vision and business model. 33 00:02:37,100 --> 00:02:40,730 It has a great dependency on the evolution of ChatGPT. 34 00:02:40,850 --> 00:02:43,850 All these things are also applicable to landscape. 35 00:02:43,880 --> 00:02:49,970 So these are both very similar companies in very similar stages. 36 00:02:50,570 --> 00:03:02,450 They are they they are both a they have a raised a a similar amount of investment between 10 and $20 37 00:03:02,450 --> 00:03:02,930 million. 38 00:03:02,930 --> 00:03:08,570 They are both in Silicon Valley I and they come both from the same company. 39 00:03:08,570 --> 00:03:14,930 So the two founders of Llama and Landscape, they used to work together in the same company. 40 00:03:14,930 --> 00:03:18,470 So it's a very interesting, uh, history. 41 00:03:18,590 --> 00:03:27,410 So a although the first version of Le Mendes was not very user friendly, they are now striving to make 42 00:03:27,410 --> 00:03:29,630 a second improved version in that regard. 43 00:03:29,630 --> 00:03:37,400 So it's very interesting to see how Llama index is trying to get more simple. 44 00:03:37,400 --> 00:03:45,440 And at the same time, launching seems to be a being more complex. 45 00:03:45,890 --> 00:03:53,240 This is very interesting in my in my impression as a, you know, a spectator looking at these two frameworks. 46 00:03:53,240 --> 00:03:55,670 For me, this is interesting. 47 00:03:56,990 --> 00:04:07,340 So in this initial lesson, we have focused our attention in the quick start guide that llama index 48 00:04:07,340 --> 00:04:07,880 provides. 49 00:04:07,880 --> 00:04:15,410 We will see where in the next lesson, but as you will see in the notebook that we attach, we have 50 00:04:15,410 --> 00:04:18,350 started playing around with llama index. 51 00:04:18,350 --> 00:04:27,500 The first thing we do is we load the dot m file with our secret credentials to access our OpenAI key, 52 00:04:27,500 --> 00:04:34,970 because as long chain llama index is going to use mostly ChatGPT as the LM foundation models for the 53 00:04:34,970 --> 00:04:38,060 LM applications built on llama index. 54 00:04:38,060 --> 00:04:40,280 Okay, so we do this. 55 00:04:40,280 --> 00:04:51,680 We load the file and then see how fast and how easy it is to load a private document, create a vector 56 00:04:51,680 --> 00:04:55,070 database and ask questions to this private document. 57 00:04:55,070 --> 00:04:59,150 In llama, index is just three lines is really amazing. 58 00:04:59,150 --> 00:05:04,280 So see that the way we use is very similar to to line chain. 59 00:05:04,280 --> 00:05:08,780 But instead of loading from line chain we are going to load from llama index. 60 00:05:08,780 --> 00:05:10,340 See the hyphen here. 61 00:05:11,570 --> 00:05:14,270 So this is the way we load a private document. 62 00:05:14,300 --> 00:05:17,840 See that in this example we are loading a document. 63 00:05:17,840 --> 00:05:20,030 In this case we are loading an article. 64 00:05:20,030 --> 00:05:27,800 This is an article talking about startups and the concept of good and bad in startups etc. with an evil 65 00:05:27,800 --> 00:05:28,940 with an evil. 66 00:05:29,300 --> 00:05:36,080 But see that right now we are loading this document, this article which is a text. 67 00:05:36,080 --> 00:05:42,650 TXT file, but we are loading it from a folder called 000 hyphen data. 68 00:05:42,650 --> 00:05:47,120 This is not going to be the case with you because you are using a data folder. 69 00:05:47,120 --> 00:05:56,030 Okay, so the article that we are using is a I think is is called good good dot txt I think it is in 70 00:05:56,030 --> 00:06:01,370 your data folder in the sample data folder we have provided in GitHub. 71 00:06:01,370 --> 00:06:04,880 But you can use whatever txt document you want. 72 00:06:04,880 --> 00:06:07,370 So this is the way you load the document. 73 00:06:07,370 --> 00:06:09,980 This is the way you create a vector database. 74 00:06:09,980 --> 00:06:13,310 This vector database is loading the private document. 75 00:06:13,310 --> 00:06:19,820 And this is the way we start, um asking questions with this uh query function. 76 00:06:19,820 --> 00:06:20,210 Okay. 77 00:06:20,210 --> 00:06:23,810 So super easy to start with Lamar index. 78 00:06:25,720 --> 00:06:31,270 Here we have a couple of additional notes about, you know, what they are. 79 00:06:32,850 --> 00:06:36,150 Teaching in the quick start C under the heart in my opinion. 80 00:06:36,330 --> 00:06:39,210 Under the hood in my opinion, this is not relevant right now. 81 00:06:39,540 --> 00:06:42,990 How to save the vector database okay, interesting. 82 00:06:43,260 --> 00:06:52,560 And in my opinion these are, uh, interesting uh, points, uh, about what is coming in the next lesson. 83 00:06:52,560 --> 00:07:00,810 So we start talking about the customization options that you can find in the mind. 84 00:07:00,960 --> 00:07:10,770 This is the most powerful, uh, thing about Lambda Index, the customization and the optimization options 85 00:07:10,770 --> 00:07:11,520 it provides. 86 00:07:11,520 --> 00:07:13,890 So we will see more about that. 87 00:07:13,950 --> 00:07:21,540 Uh, you will see that even when they talk about different use cases like agents, multi-modality, 88 00:07:21,540 --> 00:07:28,710 blah, blah, the right now, the number one use case and the I would say 99% of the cases, all the 89 00:07:28,710 --> 00:07:32,640 cases in the real world are, uh, a rack techniques. 90 00:07:32,640 --> 00:07:34,620 The first two are techniques. 91 00:07:34,740 --> 00:07:43,020 Multimodality is is is is getting a lot of attention around the lambda index community as it is around 92 00:07:43,020 --> 00:07:48,210 the long chain community is still very early, like in the case with agents. 93 00:07:48,210 --> 00:07:58,080 But the interesting thing for you to to understand is that the multimodal, uh, applications are initially 94 00:07:58,080 --> 00:07:59,460 LM applications. 95 00:07:59,460 --> 00:08:07,740 So it is an LM application that uh, apart from using text, like all the LM applications, it is going 96 00:08:07,740 --> 00:08:12,390 to use images, videos, audio, etc., other kinds of data. 97 00:08:12,390 --> 00:08:12,870 Right? 98 00:08:12,870 --> 00:08:14,250 So this is the multimodality. 99 00:08:14,250 --> 00:08:21,060 But the interesting thing is that the the base, the foundation of this multimodal application is an 100 00:08:21,060 --> 00:08:22,020 LM application. 101 00:08:22,020 --> 00:08:29,940 So you are in a very good position because you are learning the foundations of the LM applications and 102 00:08:29,940 --> 00:08:32,580 the multimodal applications as well. 103 00:08:32,850 --> 00:08:41,549 So, so, uh, this is the excuse me, this is the most interesting thing about the mind is, as we 104 00:08:41,549 --> 00:08:49,770 will see in the next lesson, customization and optimizing other interesting things around Lambda Index. 105 00:08:50,280 --> 00:09:00,150 They have a templates like launching very, uh, very early, uh, new uh functionality, as we will 106 00:09:00,150 --> 00:09:00,630 see. 107 00:09:00,630 --> 00:09:11,610 And in my opinion, by far the most interesting thing about Lambda Index is the SEQ Insights project. 108 00:09:11,730 --> 00:09:17,040 So these are project that has been built by the Lambda index team. 109 00:09:17,040 --> 00:09:19,110 And they have open source. 110 00:09:19,110 --> 00:09:26,760 This project, meaning that this project is available for everybody to to look under the hood. 111 00:09:27,000 --> 00:09:35,580 Uh, so we are going to use this project as a way to understand very, very, very in detail what an 112 00:09:35,580 --> 00:09:37,530 LM application can do. 113 00:09:37,680 --> 00:09:48,480 We have even, uh, extended the level of research, investigation and analysis of this project, uh, 114 00:09:48,480 --> 00:09:56,100 because in our opinion, Lambda Index gives you the project, but the explanations and tutorial that 115 00:09:56,100 --> 00:10:05,760 are, uh, going with the project are a little bit, uh, in my opinion, too light to, to, uh, to 116 00:10:05,760 --> 00:10:06,510 short. 117 00:10:07,050 --> 00:10:11,820 So we have a invested a lot of time studying this project. 118 00:10:11,820 --> 00:10:16,920 And we are going to go into detail, uh, with you. 119 00:10:16,920 --> 00:10:25,980 So in this initial lesson, we just wanted to give you an initial introduction about lambda index and 120 00:10:25,980 --> 00:10:34,560 a way your interest, uh, about the next lesson, which is the lesson we are going to to focus on the 121 00:10:34,560 --> 00:10:40,200 details, uh, around the, the, the lambda index framework. 122 00:10:40,200 --> 00:10:40,650 Okay. 123 00:10:40,650 --> 00:10:43,380 So let's go to the next lesson and see that.