Skip to content

Tutorial: Add vectors

First, import the LynseDB client:

1
import lynse

And then, Start the server, you can specify the root path directly 

1
2
client = lynse.VectorDBClient(uri="demo")
my_db = client.create_database("my_vec_db", drop_if_exists=False)

Sequentially add vectors

create or truncate a collection: 

1
collection = my_db.require_collection("test_add_vectors", dim=128, drop_if_exists=True)

The add_item API can be used to add vectors to the database one by one, and it is recommended to always submit data within the insert_session context manager to ensure the highest level of data security: 

1
2
3
4
5
6
7
8
9
import numpy as np

# If the field is not passed, a blank string field will be used.
# The add_item function returns the id of the data by default.
with collection.insert_session() as session:
    id = session.add_item(vector=np.random.random(128), id=1, field={'test': 'test', 'test1': 'test2'})

# If you do not use the insert_session function, you need to manually call the commit function to submit the data
# collection.commit()


2024-09-13 10:44:57 - LynseDB - INFO - Saving data...
2024-09-13 10:44:57 - LynseDB - INFO - Writing chunk to storage...
Data persisting: 100%|██████████| 1/1 [00:00<00:00, 264.76chunk/s]

2024-09-13 10:44:57 - LynseDB - INFO - Writing chunk to storage done.
2024-09-13 10:44:57 - LynseDB - INFO - Pre-building the index...
2024-09-13 10:44:57 - LynseDB - INFO - Building an index using the `Flat-IP` index mode...
2024-09-13 10:44:57 - LynseDB - INFO - Index built.

Now we can see what content the id returns

1
print(id)

1

Print collection to get some information about it 

1
print(collection)


LocalCollectionInstance(
    database="my_vec_db",
    collection="test_add_vectors",
    shape=(1, 128)
)

Add vectors in bulk

Now let's try to add vectors in bulk: 

1
collection = my_db.require_collection("test_min_vec", dim=128, drop_if_exists=True)

Similarly, we use the bulk_add_items method within the insert_session context manager to submit all the data at once: 

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
import numpy as np

vectors = []

with collection.insert_session() as session:
    for i in range(10):
        # The order is vector, id, field
        vectors.append((np.random.random(128), i, {'test': f'test_{i}'}))


    ids = session.bulk_add_items(vectors)

print(ids)


2024-09-13 10:45:14 - LynseDB - INFO - Saving data...
2024-09-13 10:45:14 - LynseDB - INFO - Writing chunk to storage...
Data persisting: 100%|██████████| 1/1 [00:00<00:00, 460.86chunk/s]

2024-09-13 10:45:14 - LynseDB - INFO - Writing chunk to storage done.
2024-09-13 10:45:14 - LynseDB - INFO - Pre-building the index...
2024-09-13 10:45:14 - LynseDB - INFO - Building an index using the `Flat-IP` index mode...
2024-09-13 10:45:14 - LynseDB - INFO - Index built.

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

Isn't it simple? At this point, we can check if the shape of the collection is correct: 

1
print(collection)


LocalCollectionInstance(
    database="my_vec_db",
    collection="test_min_vec",
    shape=(10, 128)
)