search

Azure / azure-sdk-for-net

This repository is for active development of the Azure SDK for .NET. For consumers of the SDK we recommend visiting our public developer docs at https://learn.microsoft.com/dotnet/azure/ or our versioned developer docs at https://azure.github.io/azure-sdk-for-net.
MIT License
5.25k stars 4.58k forks source link

[QUERY] StreamingChatCompletionUpdate.AuthorName property #44930

Open SergeyMenshykh opened 1 month ago

SergeyMenshykh commented 1 month ago

Library name and version

Azure.AI.OpenAI 2.0.0-beta.2

Query/Question

During the migration of our codebase to the latest Azure.AI.OpenAI 2.0.0-beta.2 package, we've noticed that the new equivalent StreamingChatCompletionUpdate of the former StreamingChatCompletionsUpdate from Azure.AI.OpenAI 1.0.0-beta.17 does not have the AuthorName property. The question is: is this a bug that needs to be fixed? Or was the property dropped intentionally? Or maybe it's available in some other way?

Environment

No response

github-actions[bot] commented 1 month ago

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @jpalvarezl @ralph-msft @trrwilson.

trrwilson commented 1 month ago

Thanks for following up with this and the other issues, @SergeyMenshykh!

Did you previously see this AuthorName field populated? If so, this is certainly a regression. The status of name in chat completions has long been a bit underspecified; it used to appear in the Azure OpenAI service specification for non-streaming Choices (which is where the prior streaming implementation inherited it from), but it's since been removed on response models (together with most request message types) and has no presence in OpenAI's base specification for responses, either.

If you have usage that's actively exercising this undocumented field, I'd like to follow up with the Azure OpenAI service teams involved and potentially OpenAI's team, also, as this would represent a clear gap that should be addressed holistically.

If this was rather just a parity observation and there's no current exercise, I'd consider the previous inclusion of name (to AuthorName) to have been spurious and the 1.x->2.x removal as a correction of an unintended field.