Can we add that default_language has been removed, and give a link to how to customers can change the service default instead?

sure

Is there an actual guide/documentation that we can link to about how to configure this at the project level? Or in the portal?

I just check the service side, the default_language is hardcoded as "en",can not be configured at the project level, remove related line in changelog

Is there a reason we're no longer providing this default?

I remember it is not needed, but I'm not sure now. Add this back

Let's make a note to confirm this before GA.

this is useful, add it back

We're missing the overloads for this model.
I'm also not sure if all the documentation will be inherited - so we should check how it renders.

add the overloads and update the documentation

I think we're still missing the base overload that just supports a raw mapping? (You can find an example of this on all the generated models) - something like:

    @overload
    def __init__(self, mapping: Mapping[str, Any]) -> None:
        ...

delete this patch file, it's false added

You don't need separate imports of Mapping from typing vs collections.abc - they are aliases for the same object:
https://docs.python.org/3/library/typing.html#typing.Mapping

So you only need to import once from collections.abc

fixed

I wouldn't support a union sequence like this - I don't think it's too much to ask that users provide a homogenous list. Keeps things a bit more simple.

you're right, fix it

I think we're still missing the base overload that just supports a raw mapping? (You can find an example of this on all the generated models) - something like:

    @overload
    def __init__(self, mapping: Mapping[str, Any]) -> None:
        ...

Let's make a note to confirm this before GA.

Rather than do this, shouldn't we do something like:

    def __init__(
        self,
        *,
        question: str,
        text_documents: Sequence[Union[str, TextDocument]],
        language: Optional[str] = None,
        **kwargs: Any,
    ) -> None:
        string_index_type = kwargs.pop("string_index_type", "UnicodeCodePoint")
        super().__init__(question=question, text_documents=normalized, language=language, string_index_type=string_index_type)

from the changelog, the string_index_type is hardcoded to Utf16CodeUnit

we don't need to use kwargs.pop("string_index_type", "UnicodeCodePoint"), it's designed to use hardcode UnicodeCodePoint