AWS Builder Center

Builder Center

Your first Amazon Bedrock call with Python

Your first Amazon Bedrock call with Python

Amazon Bedrock lets you use AI models (from Amazon, Anthropic, Meta, Mistral and others) from your code, without managing servers or GPUs. In this article I do the bare minimum: send a prompt from Python and read the response.

Rodrigo Ruiz Orozco

Mecatrónica IPN | Python + IA | Aprendiendo AWS para proyectos de automatización
What you need
  • An AWS account with a user that has the bedrock:InvokeModel permission.
  • Python 3 and the AWS CLI configured (aws configure sso or aws configure).
  • boto3, the AWS SDK for Python:
1
pip install boto3

1. Check that you have access to the model

In the console, open Amazon Bedrock → Model catalog and search for Nova Micro. If the model asks you to enable access or fill out a form, do that first. Work in us-east-1, where almost every model is available.
I’m using Nova Micro because it’s one of the cheapest models on Bedrock, which makes it ideal for practice.

2. The script

I’m using the Converse API, which uses the same request format for every model. If you want to try a different one later, you only change the modelId.
python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
import boto3

client = boto3.client("bedrock-runtime", region_name="us-east-1")

response = client.converse(
modelId="us.amazon.nova-micro-v1:0",
messages=[
{
"role": "user",
"content": [{"text": "Explain what a PID controller is in 3 sentences."}],
}
],
inferenceConfig={"maxTokens": 300, "temperature": 0.5},
)

print(response["output"]["message"]["content"][0]["text"])
print("Tokens used:", response["usage"]["totalTokens"])
Run it:
bash
1
python bedrock_test.py

3. What each part does

PartWhat it’s for
bedrock-runtimeClient for using models (the bedrock client is for managing them)
modelIdThe model you’re calling
messagesThe conversation; each message has a role (user or assistant)
maxTokensCaps response length, and therefore cost
temperatureHow varied the response is (0 = more predictable)
usageHow many tokens you consumed

4. Common errors

  • AccessDeniedException: your user lacks bedrock:InvokeModel, or the model isn’t enabled.
  • ValidationException on the model ID: check that the ID and region match.
  • NoCredentialsError: the CLI isn’t configured. Don’t hardcode your keys in the script.

5. A note on cost

Bedrock charges per input and output token, and it’s not part of the Free Tier. With a small model like Nova Micro and a low maxTokens, a few test calls cost cents, but it’s still worth having a budget alert in place.

Next step

With this you can plug a model into any script. Next for me: adding message history for multi-turn conversations, and trying converse_stream to get the response in real time.
# amazon-bedrock# python# generative-ai# beginner-projects
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article