[sdk] python 3.9 generics type annotations break compiling pipeline

[racinmat]

Environment

• KFP version: none
• KFP SDK version: 2.11.0
• All dependencies version:

cachetools==5.5.0
certifi==2024.12.14
charset-normalizer==3.4.1
click==8.1.8
colorama==0.4.6
docstring_parser==0.16
google-api-core==2.24.0
google-auth==2.37.0
google-cloud-core==2.4.1
google-cloud-storage==2.19.0
google-crc32c==1.6.0
google-resumable-media==2.7.2
googleapis-common-protos==1.66.0
idna==3.10
kfp==2.11.0
kfp-pipeline-spec==0.6.0
kfp-server-api==2.3.0
kubernetes==30.1.0
oauthlib==3.2.2
proto-plus==1.25.0
protobuf==4.25.5
pyasn1==0.6.1
pyasn1_modules==0.4.1
python-dateutil==2.9.0.post0
PyYAML==6.0.2
requests==2.32.3
requests-oauthlib==2.0.0
requests-toolbelt==0.10.1
rsa==4.9
six==1.17.0
tabulate==0.9.0
urllib3==1.26.20
websocket-client==1.8.0

Steps to reproduce

# pipeline.py
from kfp import compiler, dsl

@dsl.component(base_image='python:3.12')
def comp(message: str, labels: dict[str, str]) -> str:
    print(message)
    print(labels)
    return message

@dsl.pipeline
def my_pipeline(message: str) -> str:
    """My ML pipeline."""
    return comp(message=message, labels={"version": "1.0.0"}).output

compiler.Compiler().compile(my_pipeline, package_path='pipeline.yaml')

run the following code like this:

python pipeline.py

and the output is

Traceback (most recent call last):
  File "C:\Projects\something\kfp_playground\pipeline.py", line 5, in <module>
    @dsl.component(base_image='python:3.12')
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\component_decorator.py", line 126, in component
    return component_factory.create_component_from_func(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\component_factory.py", line 589, in create_component_from_func
    component_spec = extract_component_interface(func)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\component_factory.py", line 451, in extract_component_interface
    name_to_input_spec, name_to_output_spec = get_name_to_specs(
                                              ^^^^^^^^^^^^^^^^^^
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\component_factory.py", line 302, in get_name_to_specs
    name, list(name_to_input_specs))] = make_input_spec(
                                        ^^^^^^^^^^^^^^^^
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\component_factory.py", line 406, in make_input_spec
    return structures.InputSpec(
           ^^^^^^^^^^^^^^^^^^^^^
  File "<string>", line 8, in __init__
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\structures.py", line 57, in __post_init__
    self._validate_type()
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\structures.py", line 131, in _validate_type
    type_utils.validate_bundled_artifact_type(self.type)
  File "C:\Projects\something\kfp_playground\.venv\Lib\site-packages\kfp\dsl\types\type_utils.py", line 544, in validate_bundled_artifact_type
    raise TypeError(
TypeError: Artifacts must have both a schema_title and a schema_version, separated by `@`. Got: dict[str, str]

Expected result

Successfully compile the pipeline.

This works:

from typing import Dict

from kfp import compiler, dsl

@dsl.component(base_image='python:3.12')
def comp(message: str, labels: Dict[str, str]) -> str:
    print(message)
    print(labels)
    return message

@dsl.pipeline
def my_pipeline(message: str) -> str:
    """My ML pipeline."""
    return comp(message=message, labels={"version": "1.0.0"}).output

compiler.Compiler().compile(my_pipeline, package_path='pipeline.yaml')

but it's ugly and the dict[str,str] is in python for over 4 years already.

Materials and Reference

[

---](https://peps.python.org/pep-0585/)

Impacted by this bug? Give it a 👍.