Handling Raw User Input Prompts (Open-ended Dictation)

In some of our intents, we want the user to dictate a custom prompt or text body to be stored in the memory cache (e.g., 'Siri, add memory tag with content: Remember to invalid auth caches when schema bumps').

How can we define an AppIntent parameter that accepts open-ended dictation or arbitrary string blocks from Siri's speech-to-text translator? Are there length limits or language restrictions on string parameters parsed by Siri?

Answered by Frameworks Engineer in 891917022

You'll need to use a second turn to do this, since there is no schema domain for this. If you define an App Shortcut that doesn't fill in one of the required parameters of the intent (i.e. a String one in your case), Siri will ask for it in a second turn and the user can say any arbitrary string that will be sent to you. From there it's up to you whether you have that intent open your app, return dialog or a snippet, or just complete the action in the background.

You'll need to use a second turn to do this, since there is no schema domain for this. If you define an App Shortcut that doesn't fill in one of the required parameters of the intent (i.e. a String one in your case), Siri will ask for it in a second turn and the user can say any arbitrary string that will be sent to you. From there it's up to you whether you have that intent open your app, return dialog or a snippet, or just complete the action in the background.

Handling Raw User Input Prompts (Open-ended Dictation)
 
 
Q