WEBVTT

1
00:00:01.060 --> 00:00:03.887
Hi everyone. Today I'm going to show you how

2
00:00:03.887 --> 00:00:06.361
the document library works. The first thing we

3
00:00:06.361 --> 00:00:09.189
need to do is go to

4
00:00:09.189 --> 00:00:11.663
«read and scan». And in here

5
00:00:11.663 --> 00:00:13.783
we have my library.

6
00:00:13.783 --> 00:00:15.904
We can see all the documents we

7
00:00:15.904 --> 00:00:18.024
had. Right now we can see here

8
00:00:18.024 --> 00:00:20.498
that a document was uploaded a

9
00:00:20.498 --> 00:00:22.618
little while ago, to the queue of existing

10
00:00:22.618 --> 00:00:25.246
books. And it's in the process of

11
00:00:25.246 --> 00:00:27.409
indexing. What does that mean? Well,

12
00:00:27.409 --> 00:00:29.140
when we upload a document

13
00:00:29.140 --> 00:00:31.304
—which, as I was saying

14
00:00:31.304 --> 00:00:33.468
here, okay—, we can upload documents:

15
00:00:33.468 --> 00:00:36.064
PDF, among other formats, and also

16
00:00:36.064 --> 00:00:38.661
images, and what it does is

17
00:00:38.661 --> 00:00:41.257
take the text and split it into

18
00:00:41.257 --> 00:00:42.988
little chunks. And those chunks

19
00:00:42.988 --> 00:00:45.782
go through a process of

20
00:00:45.782 --> 00:00:48.105
vectorization that makes advanced

21
00:00:48.105 --> 00:00:49.035
search easier for

22
00:00:49.035 --> 00:00:51.716
the AI. And

23
00:00:51.716 --> 00:00:54.385
here, for example, we can see that right now

24
00:00:54.385 --> 00:00:57.054
this document, a book, has been

25
00:00:57.054 --> 00:00:59.343
split intelligently into chunks, and

26
00:00:59.343 --> 00:01:01.631
in total there will be 416

27
00:01:01.631 --> 00:01:04.300
chunks. So, this process is

28
00:01:04.300 --> 00:01:06.588
being indexed. When all the documents

29
00:01:06.588 --> 00:01:08.495
say «ready», it means that

30
00:01:08.495 --> 00:01:11.164
the vectorization process is already

31
00:01:11.164 --> 00:01:13.781
complete. After that, it's ready

32
00:01:13.781 --> 00:01:16.573
for those questions, so we're going to

33
00:01:16.573 --> 00:01:19.366
use the assistant. One more thing I wanted

34
00:01:19.366 --> 00:01:21.760
to mention: we can organize the

35
00:01:21.760 --> 00:01:24.718
documents into folders. Here, for example,

36
00:01:24.718 --> 00:01:27.032
«Innova», for example, is in a

37
00:01:27.032 --> 00:01:29.676
folder called «my things». If I

38
00:01:29.676 --> 00:01:32.320
go to the category, I could change it to

39
00:01:32.320 --> 00:01:34.634
another category and give it any name. This

40
00:01:34.634 --> 00:01:36.948
helps a lot. It organizes the

41
00:01:36.948 --> 00:01:39.261
data: if we have many

42
00:01:39.261 --> 00:01:41.575
documents on the same topic, what

43
00:01:41.575 --> 00:01:44.482
we can do is give them the same category. And

44
00:01:44.482 --> 00:01:46.608
then, if I query this category,

45
00:01:46.608 --> 00:01:49.342
the assistant will only take

46
00:01:49.342 --> 00:01:51.467
these documents into account when

47
00:01:51.467 --> 00:01:53.290
answering. If it doesn't find information,

48
00:01:53.290 --> 00:01:55.719
it'll say: «No, I don't have information about this».

49
00:01:57.010 --> 00:01:59.385
Okay, let's use the assistant.

50
00:01:59.385 --> 00:02:01.761
I come in here, to the assistant. By default it's

51
00:02:01.761 --> 00:02:03.345
asking the

52
00:02:03.345 --> 00:02:06.117
LLM and the library

53
00:02:06.117 --> 00:02:08.493
by default, but let's set it here.

54
00:02:08.493 --> 00:02:11.265
I only want the book by Miguel de

55
00:02:11.265 --> 00:02:13.641
Cervantes. I'll search in document: Miguel de

56
00:02:13.641 --> 00:02:15.621
Cervantes, Don Quixote

57
00:02:15.621 --> 00:02:17.601
de la Mancha, PDF.

58
00:02:23.252 --> 00:02:25.463
Okay, I'm going to ask it a question. I'm

59
00:02:25.463 --> 00:02:28.305
going to ask something about Sancho Panza, to see what

60
00:02:28.305 --> 00:02:30.516
information it has. «I'm Bri, a voice

61
00:02:30.516 --> 00:02:33.043
assistant from Elbrinner Labs. I'm glad to

62
00:02:33.043 --> 00:02:35.677
help you with whatever you need.» If

63
00:02:35.677 --> 00:02:38.596
I mute it, it's so it doesn't bother us.

64
00:02:38.596 --> 00:02:40.786
Okay, here, at first, we can see

65
00:02:40.786 --> 00:02:43.704
that it starts to answer here. It said yes,

66
00:02:43.704 --> 00:02:46.259
that it has information about Sancho Panza, it starts

67
00:02:46.259 --> 00:02:48.813
to say it has that information from

68
00:02:48.813 --> 00:02:51.002
page 3. If we go here,

69
00:02:51.002 --> 00:02:53.921
to «sources», we could see where it got

70
00:02:53.921 --> 00:02:56.840
all the data, which page it used

71
00:02:56.840 --> 00:02:59.171
to answer the question

72
00:02:59.171 --> 00:03:01.447
the user asked, so

73
00:03:01.447 --> 00:03:03.880
that... well. The

74
00:03:03.880 --> 00:03:06.190
user has the freedom to ask

75
00:03:06.190 --> 00:03:08.499
things, topics that are in the

76
00:03:08.499 --> 00:03:10.424
book, and the assistant will be

77
00:03:10.424 --> 00:03:12.349
smart and won't make

78
00:03:12.349 --> 00:03:14.658
things up. And I think this has

79
00:03:14.658 --> 00:03:17.353
quite a few use cases, not only for

80
00:03:17.353 --> 00:03:19.663
people with vision problems, but

81
00:03:19.663 --> 00:03:22.357
even for myself, when I have to

82
00:03:22.357 --> 00:03:24.667
look up technical documents and so on, it

83
00:03:24.667 --> 00:03:26.982
makes searching for documentation

84
00:03:26.982 --> 00:03:29.466
much easier, and it's a way to ensure

85
00:03:29.466 --> 00:03:32.306
that the AI doesn't make things up. For example,

86
00:03:32.306 --> 00:03:34.790
if I ask here... I'm going to

87
00:03:34.790 --> 00:03:37.630
ask for my name, which is not

88
00:03:37.630 --> 00:03:40.114
in this document. «I don't have information

89
00:03:40.114 --> 00:03:42.599
about that in the document: Miguel de

90
00:03:42.599 --> 00:03:44.374
Cervantes, Don Quixote

91
00:03:44.374 --> 00:03:47.219
de la Mancha, PDF.» Okay, so

92
00:03:47.219 --> 00:03:49.321
as you can see, the assistant

93
00:03:49.321 --> 00:03:51.773
made it quite clear that it has no information

94
00:03:51.773 --> 00:03:54.575
about... well, about my name in this

95
00:03:54.575 --> 00:03:57.027
document. It's a way to show

96
00:03:57.027 --> 00:03:59.129
that it's taking the

97
00:03:59.129 --> 00:04:01.712
information into account. And that's

98
00:04:01.712 --> 00:04:04.358
it for today's video. Soon,

99
00:04:04.358 --> 00:04:05.240
more. Thanks!