Add support for Azure authentication using Azure AD Tokens

[alaeddine-13]

Description

Adds support for Azure authentication using Azure AD Tokens

Pull request checklist

• Support from environment variable
• Support from model YAML config
• tests
  • Do we have tests? Should we add more?
  • Should we parametrize tests?
  • Should we add hypothesis tests?
• documentation
  • add docstrings
  • if a module is new: add API reference file for Sphinx in docs/source/api-reference
  • check API doc of Sphinx - build and show with make sphinx && make open-sphinx
• fix checks
  • start locale checks with make check
  • fix formatting with make format
• Python type info
  • check and improve Python type info
  • double check if return types are specified - including -> None
• check TODO comments in code
• check FIXME comments in code
• check if we want to increase the version or rc number
• check if the copyright needs an update
  • file header
  • README.md
  • LICENSE file

[PhilipMay]

@alaeddine-13 many thanks for thos PR.
Can you please add a copyright line to every file you touched.

Like this:

# Copyright (c) 2024 your_name_here, Deutsche Telekom AG

[PhilipMay]

@alaeddine-13 please use the formatter. can do that with make format and make check.
That should fix some CI problems.

A very detailed guide is here: https://github.com/telekom/mltb2/blob/main/CONTRIBUTING.md#testing-linting-and-formatting

[PhilipMay]

@alaeddine-13 please also add your copyright to the README.md and the LICENSE file.

[PhilipMay]

Please increase the rc number in pyproject.toml.

[PhilipMay]

Hi @alaeddine-13 ! To fix the CI issue with ruff: please do a rebase upstream/main. I fixed the issue there.

[alaeddine-13]

Hi @alaeddine-13 ! To fix the CI issue with ruff: please do a rebase upstream/main. I fixed the issue there.

done

[PhilipMay]

We have this remaining mypy issue:

 Found 3 errors in 1 file (checked 28 source files)
mltb2/openai.py:369: note:          @classmethod
mltb2/openai.py:369: note:          def from_yaml(cls, yaml_file: Any, api_key: Optional[str] = ..., **kwargs: Any) -> Any
mltb2/openai.py:369: note:      Subclass:
mltb2/openai.py:369: note:          @classmethod
mltb2/openai.py:369: note:          def from_yaml(cls, yaml_file: Any, api_key: Optional[str] = ..., azure_ad_token: Optional[str] = ...) -> Any

Should we change back to "kwargs everywhere" or ignore this issue?

[PhilipMay]

@alaeddine-13 mypy now fails with:

mltb2/openai.py:364: error: Argument "azure_endpoint" to "AsyncAzureOpenAI" has incompatible type "Optional[str]"; expected "str"  [arg-type]

The azure_endpoint is now optional in your PR. This is the reason for this issue. But AsyncAzureOpenAI expects it.
It is not optional there.

What is the reason to make azure_endpoint optional? Can that be changed back?

[alaeddine-13]

@PhilipMay can you run again the CI ? I implemented this workaround here in order to make azure_endpoint a positional argument