Run a voice workflow partly through a fast speech-to-speech loop and delegate complex lookups or reasoning to a cascaded LLM workflow.

Soon you can unlock how this was done.

Behind this: the tool used · the method · what actually resulted.