AI & Vectors

Amazon Bedrock

Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and Amazon. Each model is accessible through a common API which implements a broad set of features to help build generative AI applications with security, privacy, and responsible AI in mind.

This guide will walk you through an example using Amazon Bedrock SDK with vecs. We will create embeddings using the Amazon Titan Embeddings G1 – Text v1.2 (amazon.titan-embed-text-v1) model, insert these embeddings into a Postgres database using vecs, and then query the collection to find the most similar sentences to a given query sentence.

Create an environment

First, you need to set up your environment. You will need Python 3.7+ with the vecs and boto3 libraries installed.

You can install the necessary Python libraries using pip:

1
pip install vecs boto3

You'll also need:

Create embeddings

Next, we will use Amazon’s Titan Embedding G1 - Text v1.2 model to create embeddings for a set of sentences.

1
import boto3
2
import vecs
3
import json
4
5
client = boto3.client(
6
    'bedrock-runtime',
7
    region_name='us-east-1',
8
	# Credentials from your AWS account
9
    aws_access_key_id='<replace_your_own_credentials>',
10
    aws_secret_access_key='<replace_your_own_credentials>',
11
    aws_session_token='<replace_your_own_credentials>',
12
)
13
14
dataset = [
15
    "The cat sat on the mat.",
16
    "The quick brown fox jumps over the lazy dog.",
17
    "Friends, Romans, countrymen, lend me your ears",
18
    "To be or not to be, that is the question.",
19
]
20
21
embeddings = []
22
23
for sentence in dataset:
24
    # invoke the embeddings model for each sentence
25
    response = client.invoke_model(
26
        body= json.dumps({"inputText": sentence}),
27
        modelId= "amazon.titan-embed-text-v1",
28
        accept = "application/json",
29
        contentType = "application/json"
30
    )
31
    # collect the embedding from the response
32
    response_body = json.loads(response["body"].read())
33
    # add the embedding to the embedding list
34
    embeddings.append((sentence, response_body.get("embedding"), {}))

Store the embeddings with vecs

Now that we have our embeddings, we can insert them into a Postgres database using vecs.

1
import vecs
2
3
DB_CONNECTION = "postgresql://<user>:<password>@<host>:<port>/<db_name>"
4
5
# create vector store client
6
vx = vecs.Client(DB_CONNECTION)
7
8
# create a collection named 'sentences' with 1536 dimensional vectors
9
# to match the default dimension of the Titan Embeddings G1 - Text model
10
sentences = vx.get_or_create_collection(name="sentences", dimension=1536)
11
12
# upsert the embeddings into the 'sentences' collection
13
sentences.upsert(records=embeddings)
14
15
# create an index for the 'sentences' collection
16
sentences.create_index()

Querying for most similar sentences

Now, we query the sentences collection to find the most similar sentences to a sample query sentence. First need to create an embedding for the query sentence. Next, we query the collection we created earlier to find the most similar sentences.

1
query_sentence = "A quick animal jumps over a lazy one."
2
3
# create vector store client
4
vx = vecs.Client(DB_CONNECTION)
5
6
# create an embedding for the query sentence
7
response = client.invoke_model(
8
        body= json.dumps({"inputText": query_sentence}),
9
        modelId= "amazon.titan-embed-text-v1",
10
        accept = "application/json",
11
        contentType = "application/json"
12
    )
13
14
response_body = json.loads(response["body"].read())
15
16
query_embedding = response_body.get("embedding")
17
18
# query the 'sentences' collection for the most similar sentences
19
results = sentences.query(
20
    data=query_embedding,
21
    limit=3,
22
    include_value = True
23
)
24
25
# print the results
26
for result in results:
27
    print(result)

This returns the most similar 3 records and their distance to the query vector.

1
('The quick brown fox jumps over the lazy dog.', 0.27600620558852)
2
('The cat sat on the mat.', 0.609986272479202)
3
('To be or not to be, that is the question.', 0.744849503688346)

Resources