`assist_satellite.prompt`: An action like `assist_satellite.start_conversation` that doesn't use an LLM, but just returns the text spoken by the user

assist_satellite.start_conversation is great for its intended purpose, but I’d like an action called something like assist_satellite.prompt that:

  1. Speaks a message on the satellite, with or without a preannouncement
  2. Captures the text the user responds with but doesn’t invoke an LLM.
  3. Returns the spoken text as a response variable.

My use case

I want to be able to automate the greeting of visitors at my front door, like the door greeter droid at Jabba’s Palace. What I would like to do is put a Voice Assistant PE in a secure location on my porch and create an automation that:

  1. Is triggered by the doorbell button being pushed
  2. Uses the aforementioned assist_satellite.prompt to greet the visitor and ask what they want.
  3. Sends me a notification of what the person at the door said.

This is the highly simplified version. I ultimately plan on creating an automation (or more likely a Node RED flow) that uses conversation.process to marshal a conversation between the visitor and an LLM, ultimately letting the LLM decide if I want to talk to the person based on some criteria.

To my knowledge, just capturing STT after initiating the conversation isn’t possible with the assist_satellite.start_conversation today.