Allow to return tool call results from server

[ichitaka]

Feature Description

Currently the next-openai-app example for tool calling has two different types of execution:

• Handle tool call on client side, where the actual tool call output is added to the messages on the client side
• Handle tool call on server side, where the actual tool call output (and the original tool call) gets dropped and only the assistants response are streamed back to the client

I think it would be of great value to allow the server to send the execution and results to the client.

Use Case

If I want to show tool calls to the user. This could be beneficial if the user knows about the tools and wants to check function outputs. ChatGPTs UI shows output of tool calls.

Additional context

No response