Disabling asynchronous tool calling with gpt-realtime

API
, ,
1

In the gpt-realtime announcement blog post, they mention improvements to asynchronous tool calling:

Long-running function calls will no longer disrupt the flow of a session

Is it possible to make tool calling synchronous? My use case depends on the model blocking and effectively ignoring further inputs until the tool call returns.

2

Welcome to the dev community, @tmtcam-042.

You could simply soft-mute the mic once you receive a response.function_call_arguments.delta event and “hold” the conversation until a tool call result is submitted back.