1
00:00:00,700 --> 00:00:04,890
In this video, we will discuss
namespaces, which are an important

2
00:00:04,900 --> 00:00:07,270
concept when working with PineCon.

3
00:00:08,600 --> 00:00:11,310
PineCon allows you to partition the

4
00:00:11,320 --> 00:00:13,890
vectors in an index into namespaces.

5
00:00:16,600 --> 00:00:19,470
Queries and other operations are then

6
00:00:19,480 --> 00:00:24,850
scoped to a specific namespace, allowing
different requests to search different

7
00:00:24,860 --> 00:00:26,650
subsets of your index.

8
00:00:26,660 --> 00:00:30,310
Imagine you are dealing with news

9
00:00:30,320 --> 00:00:35,930
articles, you might want to create a
namespace for indexing articles by

10
00:00:35,940 --> 00:00:40,510
content and another for indexing articles
by title.

11
00:00:41,960 --> 00:00:45,190
Here is the key information about namespaces.

12
00:00:46,460 --> 00:00:49,430
Every index consists of one or more

13
00:00:49,440 --> 00:00:55,070
namespaces, each vector exists in exactly
one namespace.

14
00:00:55,800 --> 00:01:02,910
Namespaces are uniquely identified by a
namespace name and the default namespace

15
00:01:02,920 --> 00:01:10,310
is represented by the empty string and is
used if no specific namespace is specified.

16
00:01:11,360 --> 00:01:16,910
Remember that all the operations you
learned about in the previous video will

17
00:01:16,920 --> 00:01:22,330
be constrained to a namespace if we
include the namespace argument in each method.

18
00:01:23,660 --> 00:01:25,830
Let s see some examples.

19
00:01:27,280 --> 00:01:29,910
I am starting with an empty index.

20
00:01:31,080 --> 00:01:37,430
I have just removed and recreated the
index, it is empty.

21
00:01:38,140 --> 00:01:42,990
I ll work with the index called
LangChain, so I am selecting it.

22
00:01:43,800 --> 00:01:48,150
index equals PC .index of LangChain.

23
00:01:54,590 --> 00:01:58,760
Now I am creating a namespace, namespaces

24
00:01:58,770 --> 00:02:03,860
are automatically created the first time
they are used to upstart records.

25
00:02:05,650 --> 00:02:12,000
From the previous examples, I am copy
-pasting the code that inserts 5 random

26
00:02:12,010 --> 00:02:13,760
vectors into the index.

27
00:02:15,170 --> 00:02:18,780
If I run this code, the vectors will be

28
00:02:18,790 --> 00:02:24,600
inserted in the default namespace which
is represented by the empty string, I am

29
00:02:24,610 --> 00:02:25,100
running it.

30
00:02:26,390 --> 00:02:29,820
Now, I will add another 3 random vectors

31
00:02:29,830 --> 00:02:32,880
in a namespace called firstNamespace.

32
00:02:33,310 --> 00:02:35,160
I am copy -pasting this code.

33
00:02:36,690 --> 00:02:46,390
I will insert 3 vectors in the namespace,
their IDs will be x, y and z.

34
00:02:47,180 --> 00:02:58,770
And I am adding a second argument to the
upstart method, namespace equals firstNamespace.

35
00:03:00,460 --> 00:03:06,770
If the namespace does not exist, it is
created implicitly, I am running it.

36
00:03:09,740 --> 00:03:12,450
Just like that, you can create more than
one namespace.

37
00:03:15,080 --> 00:03:20,780
Here, I am creating another namespace
called secondNamespace and adding 2

38
00:03:20,790 --> 00:03:22,140
random vectors to it.

39
00:03:32,050 --> 00:03:34,500
I am fetching index statistics,

40
00:03:40,430 --> 00:03:47,440
currently, there are 3 namespaces in the
index, the default namespace empty string

41
00:03:47,450 --> 00:03:55,420
contains 5 vectors, the firstNamespace
contains 3 vectors and the

42
00:03:55,430 --> 00:04:03,520
secondNamespace contains 2 vectors, there
are a total of 10 vectors in the index.

43
00:04:04,860 --> 00:04:11,120
All vector operations apply to a single
namespace except describe index statistics.

44
00:04:12,040 --> 00:04:21,560
I am fetching the vector with the ID x,
index .h, IDs equals x.

45
00:04:22,170 --> 00:04:27,900
If I run this code, it will return an
empty vector because it queries the

46
00:04:27,910 --> 00:04:31,820
default namespace and there is no vector
with this ID.

47
00:04:32,730 --> 00:04:40,220
The vector with the ID x is in the
firstNamespace and I have to specify it explicitly.

48
00:04:41,350 --> 00:04:52,350
I am adding namespace equals
firstNamespace and I have got the vector,

49
00:04:55,930 --> 00:05:00,920
that applies to any other operation, if
you try to delete a vector you will have

50
00:05:00,930 --> 00:05:10,000
to specify the namespace to which it
belongs, index .delete of IDs equals x

51
00:05:12,410 --> 00:05:26,990
and the namespace equals firstNamespace,
I am running it, it was deleted so the

52
00:05:27,000 --> 00:05:29,690
fetch operation returns an empty vector.

53
00:05:31,200 --> 00:05:33,910
To delete all records from a namespace,

54
00:05:34,320 --> 00:05:39,890
specify the deleteAll arguments to index
.delete and provide a namespace argument,

55
00:05:40,580 --> 00:05:49,870
index .delete of deleteAll equals true
and the namespace equals firstNamespace,

56
00:05:54,430 --> 00:06:00,520
note that deleting all records from a
namespace also deletes the namespace

57
00:06:00,530 --> 00:06:13,890
itself, I have deleted it, I am calling
index .describeIndexStacks again, there

58
00:06:13,900 --> 00:06:21,970
is no namespace called firstNamespace
anymore, we have covered the concept of

59
00:06:21,980 --> 00:06:27,670
namespaces, now let s take a break,
in the upcoming section we ll consolidate

60
00:06:27,680 --> 00:06:34,190
what we ve learned so far and dive into
developing an app using LLMs, embeddings,

61
00:06:34,580 --> 00:06:40,570
vectors and more, in the next video we ll
explore splitting and embedding text

62
00:06:40,580 --> 00:06:42,530
using Langchain, stay tuned!