GeezerVison

API Reference | Developer Guide

This repository contains the next-generation AWS SDK for Python v4, rebuilt from the ground up with an async-first, modular architecture. These new clients allow you to interact with select AWS services using Python's native async/await functionality. Unlike Boto3, these clients are distributed per-service, leaving you the option to install only what fits your needs.

• Native asynchronous APIs: Service clients use Python async and await for non-blocking operations and concurrent I/O.
• Bidirectional streaming: Supported clients can send and receive event streams concurrently over HTTP/2.
• Modular packages: Each service client is available as a separate package (e.g., aws-sdk-dynamodb), reducing the dependencies you install and deploy.
• Generated types: Type annotations provide editor completion and static-analysis support.

Choose Boto3 if you need production stability, full AWS service coverage, synchronous workloads, or features like paginators, waiters, and presigned URLs.

Choose the AWS SDK for Python v4 if you are evaluating native async capabilities, building high-throughput async prototypes, working with streaming services in test environments, or want to provide feedback to shape the GA release.

The SDK requires Python 3.12 or later and publishes each service client as a separate package. For example, the following command installs the DynamoDB and Transcribe Streaming clients as individual packages:

python -m pip install aws-sdk-dynamodb aws-sdk-transcribe-streaming

The same clients are also available as optional dependencies of the aws-sdk-python meta-package:

python -m pip install "aws-sdk-python[dynamodb,transcribe_streaming]"

The meta-package installs only the service clients that you select as extras; without extras, it installs no clients. It coordinates client versions through its MAJOR.MINOR version, so it installs compatible client versions together.

Use the meta-package when your application depends on several clients. Install individual client packages when your application uses only one or two services and you want the smallest dependency set.

First, create a virtual environment to keep this project's packages separate from other Python projects, and activate it:

python -m venv .venv
source .venv/bin/activate

Then install the Amazon STS client used in this example:

python -m pip install aws-sdk-sts

Next, set up credentials (in e.g. ~/.aws/credentials):

[default]
aws_access_key_id = YOUR_ACCESS_KEY_ID
aws_secret_access_key = YOUR_SECRET_ACCESS_KEY
aws_session_token = YOUR_SESSION_TOKEN

For the credential sources that the experimental SDK supports, see Credential providers in the developer guide.

Then, set up a default Region (in e.g. ~/.aws/config):

[default]
region = us-east-1

Finally, run the following example, which calls Amazon STS GetCallerIdentity to verify your setup:

import asyncio

from aws_sdk_sts.aio.client import AsyncSTSClient
from aws_sdk_sts.models import GetCallerIdentityInput


async def main():
    async with AsyncSTSClient() as client:
        response = await client.get_caller_identity(GetCallerIdentityInput())
        print(f"Account: {response.account}")
        print(f"Arn: {response.arn}")


if __name__ == "__main__":
    asyncio.run(main())

Some operations use bidirectional (duplex) event streams. The default aiohttp transport doesn't support duplex streaming. To use these operations, install the client's awscrt extra and set AWSCRTHTTPClient as the transport:

python -m pip install "aws-sdk-bedrock-runtime[awscrt]"

from aws_sdk_bedrock_runtime.aio.client import AsyncBedrockRuntimeClient
from aws_sdk_bedrock_runtime.aio.config import AsyncBedrockRuntimeConfig
from smithy_http.aio.crt import AWSCRTHTTPClient


async def create_client() -> AsyncBedrockRuntimeClient:
    config = await AsyncBedrockRuntimeConfig.resolve(
        region="us-east-1",
        transport=AWSCRTHTTPClient(),
    )
    return AsyncBedrockRuntimeClient(config=config)

• SDK Homepage
• Developer Guide
• API Reference

The SDK uses GitHub Issues to track feature requests and issues with the SDK. You can provide feedback or report a bug by submitting an issue. This is the preferred mechanism to give feedback so that other users can engage in the conversation, +1 issues, etc.

See CONTRIBUTING for more information.

This project is licensed under the Apache-2.0 License.