any updates on this?

Apologies for the slow response here, @ahmedmoorsy, and thanks to @RobertCraigie for the tag!

Azure OpenAI parity support for these embedded PDF input_file content parts isn't yet available. It's in progress now, but no ETA harder than "soon, and as soon as possible."

In the interim, although it's much more cumbersome, the functionality can be approximated by uploading the file (with a purpose of assistants, as the new user_data is also coming soon together with input_file parts) and providing a vector store ID with that file to a file_search tool configuration.

+1 to implement this

+1

+1, file upload is a core functionality of LLMs right now

Another +1, * 10,000.

This is a gap that you can't bridge with vector search.

+1

My understanding is Semantic Kernel just added the support to send a PDF file to Azure Open AI with the BinaryContent feature that was part of the 1.49 release on May 8th 2025. Now we're waiting for Azure OpenAI to be able to accept that BinaryContent as part of input_file? I'm not asking for a release date, just confirming once this Issue is resolved I should be able to send PDFs to models that support them?

https://github.com/microsoft/semantic-kernel/blob/6f80214e404424411d1c39c0f896f376c4134306/dotnet/samples/Concepts/ChatCompletion/OpenAI_ChatCompletionWithFile.cs#L20