How to Conduct a Similarity Search in KDB.AI

This page provides details of how to execute similarity searches. For more advanced search filters, see Customize Filters.

Note

Similarity searches in KDB.AI are based on Approximate Nearest Neighbor (ANN) algorithms.

Setup

Before you start, make sure you have:

Select the table to search

To perform a search, specify the name of the table in which the relevant vector embeddings are stored. Using Python Client, you can create a table object from the session:

Python

Copy 
documents = session.table("documents")

Now that you have a vector embedding, you can perform a search for the nearest neighbors. Python Client uses the table object, whereas REST Client uses the table name as above. In this example, the embeddings are assumed to be eight dimensional and the number of nearest neighbors is set to 3.

Use the following command to search for the nearest neighbors:

Python

REST

q

Python

Copy 
index_name = 'vector_index'
documents.search(vectors={index_name: [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}, n=3)

Shell

Copy 
curl -s -H "Content-Type: application/json" localhost:8082/api/v2/databases/default/tables/documents/search \
-d '{"n": 3, "vectors" {"vector_index":[[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}}'

q

Copy 
// gw is a handler to the gateway
vectors: enlist[`vector_index]!enlist enlist 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0;
gw(`search;`database`table`n`vectors!(`default;`documents;3;vectors))

Batch searches

For larger workloads you can send multiple query vectors at once as seen in the following command.

Python

REST

q

Python

Copy 
index_name = 'vector_index'
documents.search(vectors={index_name: [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0],[1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0]]}, n=3)

Shell

Copy 
curl -s -H "Content-Type: application/json" localhost:8082/api/v2/databases/default/tables/documents/search \ 
-d '{"n":3,
"vectors": {"vector_index": [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0],[1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0]]}}' 

q

Copy 
// gw is a handler to the gateway
vectors: enlist[`vector_index]!enlist (1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0; 1.0 7.0 1.0 1.0 7.0 1.0 1.0 7.0 1.0 1.0 7.0 1.0);
gw(`search;`database`table`n`vectors!(`default;`documents;3;vectors))

Customize searches using Range

Use the keyword range to find all vectors within a distance you define. This applies to the qFlat index only.

Python

REST

q

Python

Copy 
index_name = 'qFlat_index'
documents.search(vectors={qFlat_index: [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0],[1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0]]}, range=5.5)

Shell

Copy 
curl -s -H "Content-Type: application/json" localhost:8082/api/v2/databases/default/tables/documents/search \ 
-d '{"vectors": {"vector_index": [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0],[1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0,1.0,7.0,1.0]]}, "range":5.5}' 

q

Copy 
// gw is a handler to the gateway
vectors: enlist[`vector_index]!enlist (1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0; 1.0 7.0 1.0 1.0 7.0 1.0 1.0 7.0 1.0 1.0 7.0 1.0);
gw(`search;`database`table`vectors`range!(`default;`documents;vectors;5.5e))

Processing results

You can return a subset of the columns in the table, reducing the amount to data sent back to the client:

Python

REST

q

Python

Copy 
index_name = "vector_index"
documents.search(vectors={index_name: [[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}, n=3, aggs=[["author", "content"]])

Shell

Copy 
curl -s -H "Content-Type: application/json" localhost:8082/api/v2/databases/default/tables/documents/search \ 
-d '{"n":3,
"vectors":{"vector_index":[[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}, aggs:[["author","content"]]}'

q

Copy 
// gw is a handler to the gateway
vectors: enlist[`vector_index]!enlist enlist 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0;
gw(`search;`database`table`vectors`n`aggs!(`default;`documents;vectors;3;{x!x}(`author;`content)))

In addition to returning a subset of the columns, you can return aggregated results, grouped by categorical variables, and sorted by column name:

Python

REST

q

Python

Copy 
index_name = "vector_index"
documents.search(vectors={index_name:[[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}, n=3, aggs={"sumLength":["sum","length"]}, group_by=["author"], sort_columns=["sumLength"])

Shell

Copy 
curl -s -H "Content-Type: application/json" localhost:8082/api/v2/databases/default/tables/documents/search \ 
-d '{"n":3,
"vectors":{"vector_index":[[1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.0,1.0]]}, aggs:{"sumLength":["sum","length"]},
groupBy:["author"],sort_columns:["sumLength"]}'

q

Copy 
// gw is a handler to the gateway
vectors: enlist[`vector_index]!enlist enlist 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0;
gw(`search;`database`table`vectors`n`aggs`groupBy`sortColumns!(`default;`documents;vectors;3;enlist[`sumLength]!enlist (`sum;`length);`author;`sumLength))

You can find all supported aggregations listed here.

Next steps

Now that you're familiar with how to perform a similarity search, move on to the following: