Vector Search Quickstart

To enable your machine learning model, Vector Search uses data to be compared by similarity within a database, even if it is not explicitly defined by a connection. A vector is an array of floating point type that represents a specific object or entity.

The foundation of Vector Search lies within the embeddings, which are compact representations of text as vectors of floating-point numbers. These embeddings are generated by feeding the text through an API, which uses a neural network to transform the input into a fixed-length vector. Embeddings capture the semantic meaning of the text, providing a more nuanced understanding than traditional term-based approaches. The vector representation allows for input that is substantially similar to produce output vectors that are geometrically close; inputs that are not similar are geometrically further apart.

To enable Vector Search, a new vector data type is available in your Astra DB, Apache Cassandra®, or DataStax Enterprise (DSE) database with Vector Search.

Prerequisites

There are no prerequisite tasks.

In general, to use Vector Search with DSE, you’ll follow these instructions: The embeddings were randomly generated in this quickstart. Generally, you would run both your source documents/contents through an embeddings generator, as well as the query you were asking to match. This example is simply to show the mechanics of how to use CQL to create vector search data objects.

Create vector search keyspace

Create a new vector search keyspace called cycling:

CREATE KEYSPACE IF NOT EXISTS cycling 
   WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor' : '1' };

Use vector search keyspace

Select the keyspace cycling for creating the vector search example:

USE cycling;

Create vector search table

Create a new table called comments_vs to store comments information:

CREATE TABLE IF NOT EXISTS cycling.comments_vs (
  record_id timeuuid,
  id uuid,
  commenter text,
  comment text,
  comment_vector VECTOR <FLOAT, 5>,
  created_at timestamp,
  PRIMARY KEY (id, created_at)
)
WITH CLUSTERING ORDER BY (created_at DESC);

Alternatively, if you didn’t have a vector column in your original table, you could alter the original table to add a vector column comment_vector to store the embeddings:

-- ALTER TABLE cycling.comments_vs
--     ADD comment_vector VECTOR <FLOAT, 5>; (1)
ALTER TABLE cycling.comments_vs
    ADD comment_vector VECTOR <FLOAT, 5>;

1	Notice that the vector uses the `float` data type and specifies the array dimension of 5 to store the embeddings.

Create vector search index

Create the custom index called comment_vector with Storage Attached Indexing (SAI):

CREATE CUSTOM INDEX IF NOT EXISTS ann_index 
  ON cycling.comments_vs(comment_vector) USING 'StorageAttachedIndex';

For more about SAI, see the Storage Attached Index documentation.

Load Vector Search data into your database

Insert data into the table:

INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector)  
   VALUES (
      now(), 
      e7ae5cf3-d358-4d99-b900-85902fda9bb0, 
      'Alex',
      'Raining too hard should have postponed', 
      '2017-02-14 12:43:20-0800',
      [0.45, 0.09, 0.01, 0.2, 0.11]
);
INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      e7ae5cf3-d358-4d99-b900-85902fda9bb0,
      'Alex', 
      'Second rest stop was out of water', 
      '2017-03-21 13:11:09.999-0800', 
      [0.99, 0.5, 0.99, 0.1, 0.34]
);
INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      e7ae5cf3-d358-4d99-b900-85902fda9bb0, 
      'Alex',
      'LATE RIDERS SHOULD NOT DELAY THE START', 
      '2017-04-01 06:33:02.16-0800',
      [0.9, 0.54, 0.12, 0.1, 0.95]
);

INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      c7fceba0-c141-4207-9494-a29f9809de6f, 
      'Amy',
      'The gift certificate for winning was the best', 
      totimestamp(now()),
      [0.13, 0.8, 0.35, 0.17, 0.03]
);

INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      c7fceba0-c141-4207-9494-a29f9809de6f, 
      'Amy',
      'Glad you ran the race in the rain', 
      '2017-02-17 12:43:20.234+0400',
      [0.3, 0.34, 0.2, 0.78, 0.25]
);

INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      c7fceba0-c141-4207-9494-a29f9809de6f, 
      'Amy',
      'Great snacks at all reststops', 
      '2017-03-22 5:16:59.001+0400',
      [0.1, 0.4, 0.1, 0.52, 0.09]
);
INSERT INTO cycling.comments_vs (record_id, id, commenter, comment, created_at, comment_vector) 
   VALUES (
      now(), 
      c7fceba0-c141-4207-9494-a29f9809de6f, 
      'Amy',
      'Last climb was a killer', 
      '2017-04-01 17:43:08.030+0400',
      [0.3, 0.75, 0.2, 0.2, 0.5]
);

Note the format of the new vector data type.

Query vector data with CQL

Simple query

To query data using Vector Search, use a SELECT query:

CQL
Result

SELECT * FROM cycling.comments_vs 
    ORDER BY comment_vector ANN OF [0.15, 0.1, 0.1, 0.35, 0.55] 
    LIMIT 3;

 id                                   | created_at                      | comment                                | comment_vector               | commenter | record_id
--------------------------------------+---------------------------------+----------------------------------------+------------------------------+-----------+--------------------------------------
 e7ae5cf3-d358-4d99-b900-85902fda9bb0 | 2017-04-01 14:33:02.160000+0000 | LATE RIDERS SHOULD NOT DELAY THE START | [0.9, 0.54, 0.12, 0.1, 0.95] |      Alex | 616e77e0-22a2-11ee-b99d-1f350647414a
 c7fceba0-c141-4207-9494-a29f9809de6f | 2017-02-17 08:43:20.234000+0000 |      Glad you ran the race in the rain | [0.3, 0.34, 0.2, 0.78, 0.25] |       Amy | 6170c1d0-22a2-11ee-b99d-1f350647414a
 c7fceba0-c141-4207-9494-a29f9809de6f | 2017-04-01 13:43:08.030000+0000 |                Last climb was a killer |   [0.3, 0.75, 0.2, 0.2, 0.5] |       Amy | 62105d30-22a2-11ee-b99d-1f350647414a

The limit has to be 1,000 or fewer.

Scrolling to the right on the results shows the comments from the table that most closely matched the embeddings used for the query.

Similarity query

To obtain the similarity calculation of the best scoring node closest to the query data as part of the results, use a modified SELECT query:

CQL

SELECT  comment, similarity_cosine(comment_vector, [0.2, 0.15, 0.3, 0.2, 0.05]) 
    FROM cycling.comments_vs
    ORDER BY comment_vector ANN OF [0.1, 0.15, 0.3, 0.12, 0.05] 
    LIMIT 3;

Result

 comment                                       | system.similarity_cosine(comment_vector, [0.2, 0.15, 0.3, 0.2, 0.05])
-----------------------------------------------+-----------------------------------------------------------------------
             Second rest stop was out of water |                                                              0.949701
 The gift certificate for winning was the best |                                                               0.86062
 The gift certificate for winning was the best |                                                               0.86062

(3 rows)

The supported functions for this type of query are:

similarity_dot_product
similarity_cosine
similarity_euclidean

with the parameters of (<vector_column>, <embedding_value>). Both parameters represent vectors.

Vector Search utilizes Approximate Nearest Neighbor (ANN) that in most cases yields results almost as good as the exact match. The scaling is superior to Exact Nearest Neighbor (KNN).

Least-similar searches are not supported.

Vector Search works optimally on tables with no overwrites or deletions of the comment_vector column. For an comment_vector column with changes, expect slower search results.

The embeddings were randomly generated in this example. Generally, you would run both your source documents/contents through an embeddings generator, as well as the query you were asking to match. This example is simply to show the mechanics of how to use CQL to create vector search data objects.

With the code examples, you have a working example of our Vector Search. Load your own data and use the search function. The embeddings were randomly generated in this quickstart. Generally, you would run both your source documents/contents through an embeddings generator, as well as the query you were asking to match. This example is simply to show the mechanics of how to use CQL to create vector search data objects.