Let Decision Models Guide Your Voice Assistant

A voice assistant's effectiveness can be improved by using a decision model to guide the conversation. This architecture separates tasks, with a speech model like GPT-Live handling the listening and talking, while a decision model tracks what has been confirmed and what is still missing. The decision model analyzes the conversation transcript against predefined requirements and criteria to determine the current status. It then selects the most logical next action or question for the assistant to ask from a set of defined options. The process is illustrated by building an assistant for a cleaner inspecting a rental apartment, guiding them through a checklist of tasks.