Python client for Vektonn

Vektonn is a high-performance battle-tested kNN vector search engine for your data science applications. It helps you manage vectors' lifecycle and radically reduces time to market.

See documentation for more info.

Installation

Install the latest version:

$ pip install vektonn

Install specific version:

$ pip install vektonn==1.2.3

Upgrade to the latest version:

$ pip install --upgrade vektonn

Usage

Having Vektonn running on localhost:8081 and configured for QuickStart examples one can access it with synchronous Python client:

from vektonn import Vektonn

vektonn_client = Vektonn('http://localhost:8081')

or asynchronous one:

from vektonn import VektonnAsync

vektonn_client_async = VektonnAsync('http://localhost:8081')

To upload data to Vektonn use upload() method:

from vektonn.dtos import Attribute, AttributeValue, InputDataPoint, Vector

vektonn_client.upload(
    data_source_name='QuickStart.Source',
    data_source_version='1.0',
    input_data_points=[
        InputDataPoint(
            attributes=[
                Attribute(key='id', value=AttributeValue(int64=1)),
                Attribute(key='payload', value=AttributeValue(string='sample data point')),
            ],
            vector=Vector(is_sparse=False, coordinates=[3.14, 2.71]))
    ])

To query Vektonn for k nearest data points to the given query_vector use search() method:

from vektonn.dtos import Vector, SearchQuery

k = 10
query_vector = Vector(is_sparse=False, coordinates=[1.2, 3.4])

search_results = vektonn_client.search(
    index_name='QuickStart.Index',
    index_version='1.0',
    search_query=SearchQuery(k=k, query_vectors=[query_vector]))

print(f'For query vector {query_vector.coordinates} {k} nearest data points are:')
for fdp in search_results[0].nearest_data_points:
    attrs = {x.key: x.value for x in fdp.attributes}
    distance, vector, dp_id, payload = fdp.distance, fdp.vector, attrs['id'].int64, attrs['payload'].string
    print(f' - "{payload}" with id = {dp_id}, vector = {vector.coordinates}, distance = {distance}')