1
00:00:03,650 --> 00:00:12,860
Okay, so in this lesson we are going to present briefly a couple of other interesting level three LM

2
00:00:12,860 --> 00:00:14,180
applications.

3
00:00:14,180 --> 00:00:21,080
So as always you will find the the best information in the notebooks.

4
00:00:21,080 --> 00:00:29,270
Uh, we are going to to work with two notebooks and also to a presentations with the summaries.

5
00:00:29,270 --> 00:00:32,810
But the notebooks are always going to be the best source for you.

6
00:00:33,600 --> 00:00:43,200
So first we are going to review the Create Llama project and the way to develop full stack starters

7
00:00:43,200 --> 00:00:43,650
with it.

8
00:00:43,650 --> 00:00:51,060
So as you know, the Create Llama is a tool developed by Llama Index to streamline the creation of rack

9
00:00:51,060 --> 00:00:53,130
full stack application templates.

10
00:00:54,540 --> 00:01:00,570
Do you remember that we saw uh, we talk about that when we review, uh, llama index.

11
00:01:00,570 --> 00:01:07,560
So right now and I highlight this right now because I think it is going to improve.

12
00:01:07,560 --> 00:01:11,970
But right now Create llama has a few limitations.

13
00:01:12,910 --> 00:01:18,280
It does not include a database to host private documents.

14
00:01:18,610 --> 00:01:23,560
So right now, private documents are manually included in in an app directory.

15
00:01:23,890 --> 00:01:29,770
And therefore it does not include Crud functionality to manage private documents.

16
00:01:29,770 --> 00:01:32,830
So it's a very simple approach.

17
00:01:33,880 --> 00:01:41,860
It currently allows the creation of three types of full stack applications, or more or less pseudo

18
00:01:41,860 --> 00:01:43,030
full stack applications.

19
00:01:43,030 --> 00:01:49,090
In some cases, the option one is front end and back end with Next.js serverless.

20
00:01:49,540 --> 00:01:51,550
So this is not a true full stack.

21
00:01:51,550 --> 00:02:00,340
The option two is front end with Next.js and back end with fast API, which is something we are familiar

22
00:02:00,340 --> 00:02:00,820
with.

23
00:02:00,820 --> 00:02:08,199
And the option three is front end with Next.js and back end with express, which is based on JavaScript.

24
00:02:08,500 --> 00:02:12,100
So some comments about these options.

25
00:02:12,100 --> 00:02:16,900
The first option front end and back end with Next.js serverless.

26
00:02:17,470 --> 00:02:19,540
This is the simplest option.

27
00:02:19,540 --> 00:02:29,530
It only requires deployment on Vercel and in our opinion, it has the limitation that it will fall short

28
00:02:29,530 --> 00:02:31,990
when we try to scale the application.

29
00:02:33,160 --> 00:02:35,830
How does Lamar Index use it?

30
00:02:35,860 --> 00:02:45,070
A you will see that the rag logic is in this file, and you will immediately see that this option one

31
00:02:45,070 --> 00:02:46,540
is written in TypeScript.

32
00:02:46,540 --> 00:02:48,280
So JavaScript.

33
00:02:50,260 --> 00:02:56,440
The second option is a the frontend with Next.js and the backend with fast API.

34
00:02:56,920 --> 00:03:04,420
The Laima index tutorial suggests that deploying both frontend and backend a a.

35
00:03:06,820 --> 00:03:13,930
The suggestion is to deploy both front end and back end on render com, which we don't think is a good

36
00:03:13,930 --> 00:03:19,750
idea, as it will limit us when scaling the front end of the application.

37
00:03:19,930 --> 00:03:24,880
So it seems more appropriate to deploy the front end over cell and the back end on render.

38
00:03:24,880 --> 00:03:28,420
Com as we did with the Todo application, right.

39
00:03:28,540 --> 00:03:33,850
So this one is interesting, but it has also some limitations.

40
00:03:34,510 --> 00:03:41,740
Here you in this file of the starter template you will find the rug logic of the of llama index.

41
00:03:41,740 --> 00:03:47,290
And interesting thing this option two uses Python okay.

42
00:03:47,650 --> 00:03:53,680
The third option is the front end with Next.js and the back end with express.

43
00:03:53,680 --> 00:03:56,050
And as you know express is based on JavaScript.

44
00:03:56,050 --> 00:04:04,510
So considering that llama index as well as long chain and the ChatGPT API are natively developing Python,

45
00:04:05,140 --> 00:04:15,160
our understanding is that it is most advisable to specialize in back end with fast API instead of express.

46
00:04:15,460 --> 00:04:16,209
Why?

47
00:04:16,240 --> 00:04:23,290
Because there are so many artificial intelligence engineers used to work with Python.

48
00:04:23,620 --> 00:04:28,960
That is going to be very difficult to, uh, move them to JavaScript.

49
00:04:28,960 --> 00:04:32,380
So let's see what happens in the future.

50
00:04:32,860 --> 00:04:42,430
But I don't think a Python is going anywhere in artificial intelligence because the main community of

51
00:04:42,430 --> 00:04:44,650
users are Python users.

52
00:04:44,650 --> 00:04:52,120
So I don't think this option three has a lot of appeal for, uh, artificial intelligence engineers.

53
00:04:52,120 --> 00:04:59,110
So for this reason, we will not cover this third option in the course, although it may be a good alternative

54
00:04:59,110 --> 00:05:05,140
for developers with a background in JavaScript of or working in teams specialized in JavaScript.

55
00:05:05,140 --> 00:05:09,730
So let's see what happens a in our opinion.

56
00:05:10,730 --> 00:05:17,780
Create the llama right now is not the perfect solution, but it can be.

57
00:05:18,140 --> 00:05:28,670
So we expect that especially this second option is going to improve and become a much better option

58
00:05:28,670 --> 00:05:29,510
for us.

59
00:05:31,130 --> 00:05:39,740
Here you have a detailed notes in the notebook regarding the second, uh, the second level three LM

60
00:05:39,740 --> 00:05:42,080
application we wanted to highlight here.

61
00:05:43,530 --> 00:05:44,460
The project.

62
00:05:44,460 --> 00:05:54,150
Chat Llama index this is a starter template for creating a rack application for chatting with your private

63
00:05:54,150 --> 00:05:56,700
documents, which is very interesting.

64
00:05:57,630 --> 00:06:05,640
It is similar to the Create Llama starter template, but with this is Next.js.

65
00:06:07,690 --> 00:06:11,140
So it is very similar to the first option of create lemma.

66
00:06:11,170 --> 00:06:11,620
Okay.

67
00:06:11,620 --> 00:06:14,470
The one that was using Next.js serverless.

68
00:06:14,470 --> 00:06:21,040
But in this case the application allows file uploads through the user interface instead of having to

69
00:06:21,040 --> 00:06:26,140
do that manually in the application file, as we do with create lemma.

70
00:06:26,170 --> 00:06:26,950
Okay.

71
00:06:27,790 --> 00:06:30,280
Comments on this option.

72
00:06:30,760 --> 00:06:36,700
First one, it uses the TypeScript version of llama index, not the Python version.

73
00:06:36,700 --> 00:06:37,390
Baths.

74
00:06:38,290 --> 00:06:46,240
And second, it falls a bit short for professional applications as it does not use backend API or back

75
00:06:46,240 --> 00:06:47,260
end database.

76
00:06:47,260 --> 00:06:56,410
So we think a that this option and this option.

77
00:06:56,410 --> 00:07:02,290
The second option of create Llama, we have the feeling that they are going to evolve in a positive,

78
00:07:02,290 --> 00:07:04,060
uh, direction.

79
00:07:05,110 --> 00:07:16,390
The current status of both of them is not the ideal one, but we think a llama index and I would bet

80
00:07:16,390 --> 00:07:25,570
also long chain are going to provide more professional starters or starters for better prepared for

81
00:07:25,570 --> 00:07:26,800
professional applications.

82
00:07:26,800 --> 00:07:32,530
So my recommendation for you is explore these ones experiment a little bit.

83
00:07:32,530 --> 00:07:38,110
And the most important, pay attention to how they evolve.

84
00:07:38,110 --> 00:07:43,840
Because I think, uh, options like these ones can be super interesting for us.

85
00:07:43,840 --> 00:07:45,520
Artificial intelligence engineer.

86
00:07:45,520 --> 00:07:49,120
So with this we are finishing this blog.

87
00:07:49,120 --> 00:07:51,700
I think this is a very important blog.

88
00:07:51,700 --> 00:08:00,730
Uh, and it is very important for you to be comfortable and familiar with the different, uh, lessons

89
00:08:00,730 --> 00:08:02,590
and subsections of this blog.

90
00:08:02,590 --> 00:08:09,790
So please go back if you are not, it is very important for you this this this one, this blog is very,

91
00:08:09,790 --> 00:08:10,780
very important for you.

92
00:08:10,780 --> 00:08:18,640
So please spend the necessary time to consolidate, uh, your, your, uh, lessons, uh, here.

93
00:08:18,640 --> 00:08:19,240
Okay.

94
00:08:19,240 --> 00:08:27,970
And once you are ready, we will go to the next, uh, blog where we are going to start talking about

95
00:08:27,970 --> 00:08:33,730
a very important advanced concepts regarding LM applications.