guidance on how to implement tool confirmation responses in v5?

[nikitatome]

In this v4 example, before running streamText we process the parts of the lastMessage. What is the recommended way to do this in v5? When I'm trying to do this with this approach we're observing multiple assistant messages in the UI response (with beta.14) It's also unclear to me what order the messages are processed in. When streamText is waiting for a response, does the UI message stream's execute function rerun before continuing streamText?

https://ai-sdk.dev/cookbook/next/human-in-the-loop#handle-confirmation-response

We could use prepareStep, however prepareStep's messages are modelMessages so I'm not sure we should be modifying them?

[lgrammel]

we are working on an updated docs page cc @nicoalbanese

[mediocreaverage1]

Hi Igrammel,
I know you're busy and thank you for your contributions. Is there any progress on the docs update? Or at least give us a quick answer here which will benefit everyone waiting for this feature?

If the user enter yes or no, how to update the llm model with the user input?