.Net: Bug: When using GroupChatOrchestration, InteractiveReplyCallback and OpenAIResponseAgent it will fail if there is a function call in the chat history

[DanielMcAssey]

Describe the bug
When using GroupChatOrchestration with 2 or more OpenAIResponseAgent's and an InteractiveReplyCallback, if one of the agents calls a Function, then asks for UserInput, the input is received, the next agent to be invoked will (silently) fail as there is a FunctionCall and FunctionCallResponse messages, and then the chat indefinitely hangs.

It fails because these lines throw when there is unexpected input:

• Throws when item is FunctionCall and FunctionCallResponse (As well as others):

  semantic-kernel/dotnet/src/Agents/OpenAI/Extensions/KernelContentExtensions.cs

  Line 13 in e73446e

  internal static ResponseContentPart ToResponseContentPart(this KernelContent content)

• Throws when message role is "tool":

  semantic-kernel/dotnet/src/Agents/OpenAI/Extensions/ChatContentMessageExtensions.cs

  Line 44 in e73446e

  public static ResponseItem ToResponseItem(this ChatMessageContent message)

To Reproduce
Steps to reproduce the behavior:

1. Create a GroupOrchestrationChat with 2 OpenAIResponseAgent's
2. Add a function call to one of them
3. Get one of the agents to call the function, then stop and ask for User Input
4. Respond with user input, try continue the conversation with the other Agent
5. Will indefinitely hang

Expected behavior
It should not hang, and the conversation should continue as expected.

Screenshots
If applicable, add screenshots to help explain your problem.

Platform

• Language: C#
• Source: NuGet package 1.65.0
• AI model: GPT-4.1
• IDE: VS Code
• OS: Windows