According to the documentation under response.create calls theres a field “conversation” in the response object that can only be set to “auto“ or “none”.
When left on “auto” it appears to add messages to a conversation with a separate conversation_id from the conversation_id used for the audio/vad_detection_turned_on.
This makes sending side channeled messages intended to be relevant/related to the current realtime audio conversation’s context useless. There ends up being two separate conversations with different contexts one for text based messages one for audio based.
Is this a known bug or am I missing something?
Also it would be awesome if we could specify the conversation_id manually instead of having to trust the “auto“ option.