aws-sdk-python-usage

A reference for AWS SDK for Python (boto3/botocore) development patterns, to be used whenever writing Python code that uses AWS services via boto3 or botocore. It covers creating service clients or resources, configuring sessions and credentials, error handling with ClientError, paginators and waiters, S3 file transfers and presigned URLs, DynamoDB table operations, and client configuration, and applies whenever Python code imports boto3 or botocore or when the user asks about AWS operations in Python. It also instructs that no emojis be used in code, comments, or output while active.

The guidance starts from the client-versus-resource distinction: clients map 1:1 to service APIs and return plain dicts and exist for every service, while resources provide an object-oriented interface and exist only for select services such as S3, DynamoDB, EC2, IAM, SQS, SNS, CloudFormation, CloudWatch, and Glacier, auto-marshaling types in a way that is especially useful for DynamoDB. It advises reusing a single client instance rather than creating clients inside loops, noting clients are thread-safe once instantiated, and that parameter names match the AWS API casing, typically PascalCase.

For error handling it recommends catching only when there is an actionable response, preferring typed exceptions on the client (via client.exceptions) over generic ClientError string matching, and reserving ClientError as a top-level catch-all that lives in botocore rather than boto3. It promotes a script structure where the __main__ guard is a single call to main(), keeping argument parsing, error presentation, and exit codes in main() and never calling sys.exit from business logic. It directs the use of paginators with JMESPath .search() instead of manual NextToken loops, waiters for resource states, and botocore.config.Config for retries, timeouts, and connection pools, and shows logging via set_stream_logger. A common-issue note clarifies that ClientError must be imported from botocore.exceptions, not boto3.exceptions. Service-specific reference files for S3 and DynamoDB, plus references for configuration, credentials, error handling, pagination, and waiters, must be loaded when relevant.