boto3 with auto-complete and dataclasses not dicts

autocomplete, autosuggest, aws, boto, boto3, dataclasses, pycharm, python, python36
pip install autoboto==0.4.3




pip install autoboto


  • Project Status: Alpha. You should use this only for ad-hoc queries when exploring the AWS. It's for people who roughly know what they want from AWS, but don't want to scroll up and down the long (and good) boto3 documentation pages to find out the right method and parameter names. We are in the 21st century and for user-facing code, the auto-complete should work.
  • All response objects are dataclasses.
  • All response objects have a response_metadata which is an unmodified dictionary normally returned under the ResponseMetadata key of the response dictionary.
  • Passing nested objects isn't as easy as passing nested dictionaries, but it has benefits.
  • The method names are as in boto3, but parameter names have been changed from CamelCase to snake_case.
  • Custom methods that boto3 does not generate from botocore (for example, s3.upload_file are simply delegated to the boto3 client and have no documentation available (for now). The parameter names are as in the original methods.
from import s3

s3_client = s3.Client()

for bucket in s3_client.list_buckets().buckets:
    for obj in s3_client.list_objects_v2(
        print(f" - {obj.key}")

You can also paginate:

for page in s3_client.list_objects_v2(
    for obj in page.contents:
        print(f" - {obj.key}")

Code Generation

You don't need to read this section. It's about how to generate the autoboto code.

When you install autoboto from, the package already contains the generated code for all the services that boto3 supports.

If you want to re-generate the code, you can do so with the included botogen.

Code generation includes importing the generated modules and instantiating the generated client classes. For this to work you will need to have AWS_PROFILE environment variable pointing too an AWS profile which has region set.

python -m botogen --services s3,cloudformation,lambda


  • autoboto - package where the generated code is put just before release. Do not add anything there manually. All files in this directory may be overwritten.
  • botogen - the code responsible for autoboto generation
  • botogen.autoboto_template - contents of this package end up in the generated autoboto package.
    • .gitignore file under botogen/autoboto_template instructs git to ignore all files in the directory. This is so that the generated code in autoboto package is never added to version control. Therefore, during autoboto development, when you are adding new files to the botogen/autoboto_template, you need to add -f flag to force-add them to git.
  • botogen.indentist - generic Python code generator

Directory Structure

build/                                  All build artifacts are put here

    release/                            Release builds happen here

        20180909_135602/                individual release build directory; Added to sys.path
            autoboto/                   generated autoboto package; an augmented copy of botogen/autoboto_complete

    test/                               Test builds happen here

        20180909_135330/                individual test build directory; Added to sys.path
            autoboto_20180909_135330/   generated autoboto package; an augmented copy of botogen/autoboto_complete

    test-packages/                      Target directory for test builds -- where the generated
                                        packages are put after successful completion of a build
                                        and tests passing on the generated code.
  • build_dir -- a temporary directory in which all the build artifacts are generated. In the example above, build/test/20180909_135330 and build/release/20180909_135602 are build directories.
  • target_package -- name of the generated target package; used in import statements in the generated code. autoboto when generating the release; autoboto_{timestamp} in tests.
  • target_dir -- the directory in which to put the target package. By default it's the current directory, but in tests it is build/test-packages.


Do not use any imports from botogen.autoboto_template in tests of the generated code because the generated code does not import from there. Instead, use the dedicated fixtures (for which, ironically, the auto-complete won't work).


tests directory contains both unit tests and integration tests. For unit tests you don't need an AWS account.

pytest tests/unit

To run tests across multiple Python versions, use tox. To run tox, you need to first generate the autoboto package or it will fail.

If you're using pyenv and virtualenvs, don't run tox from within a virtualenv.