AWS SDK for Python v4¶
The AWS SDK for Python v4 provides synchronous and asynchronous clients for supported AWS services. Each service has its own package.
Developer Preview
This SDK is for evaluation and testing only. APIs might change before general availability, so we recommend pinning the versions of the packages you install. For production workloads, use Boto3.
Features¶
- Sync and async clients. Each package includes a synchronous client and an
asynchronous client that works with
asyncio. Both clients use the same models. - Per-service packages. Install only the clients your application uses.
- Generated types. Each client and its data types are generated from the service's Smithy model. Operations take keyword arguments and return typed objects instead of dictionaries.
- Streaming support. The asynchronous clients for streaming services support event streams and bidirectional streaming.
The code generator builds each client from its service's Smithy model.
Install a client¶
The clients require Python 3.12 or later. Create and activate a virtual environment:
See available clients for a list of service packages.
List DynamoDB tables¶
Before running the example, grant an identity dynamodb:ListTables permission
and configure AWS credentials. The default credential chain checks these
built-in sources:
- Environment variables such as
AWS_ACCESS_KEY_ID,AWS_SECRET_ACCESS_KEY, andAWS_SESSION_TOKEN - Shared AWS config and credentials files, including
credential_process
Other supported providers come from aws-credentials-* packages. Install and
configure the package for your environment:
aws-credentials-stsfor profile-based AssumeRoleaws-credentials-httpfor Amazon ECS or Amazon EKS container credentialsaws-credentials-imdsfor Amazon EC2 instance metadata
The SDK detects installed provider packages and adds them to the chain automatically. You do not need to pass a credential resolver or change the client code. The IAM Identity Center (SSO) credential provider and the login credentials provider are not yet supported.
Use the Amazon DynamoDB client to list up to ten
tables in us-east-1:
import asyncio
from aws_sdk_dynamodb import AsyncDynamoDBClient, AsyncDynamoDBConfig
async def main():
config = await AsyncDynamoDBConfig.resolve(region="us-east-1")
async with AsyncDynamoDBClient(config=config) as client:
response = await client.list_tables(limit=10)
for table in response.table_names or []:
print(table)
asyncio.run(main())
The config's resolve() method loads shared settings such as retry
configuration. Here, the region argument overrides a region set in the
environment or shared config. The client resolves credentials through the
default chain when it sends the request.