Another spin at #292 which is a bit better than my attempt in #295. As explained in #292:

Of course, it's possible that the speech that Dicio is hearing is not directed at it, so it doesn't make sense to retry indefinitely: if the user's second attempt is also something that Dicio can't understand, this implementation omits the "Could you repeat?" part of the response and stops re-enabling the microphone until the user successfully performs another interaction.

In order to make this possible, two architectural changes were needed:

• Allowing the fallback skill to check if it was triggered on the previous execution, too. This can be done by adding a new field to SkillContext:
  • in Smarter fallback responses #292 it was InteractionLog
  • in Smarter fallback responses 2 #295 it was a boolean previousInteractionWasFallback
  • in this PR it's the previous SkillOutput.
• Communicating to SkillEvaluator that the microphone should be enabled again
  • in Smarter fallback responses #292 this was done by adding a getKeepListening() function to SkillOutput
  • in Smarter fallback responses 2 #295 there were no architectural changes, but the text fallback skill returned an unmatchable skill in getNextSkills(), which still made SkillEvaluator think the skill was requesting more output
  • in this PR SkillOutput.getNextSkills() is renamed to getInteractionPlan() with various options on how the interaction should continue (see the sealed class InteractionPlan), which also includes whether the microphone should be reenabled immediately or not

Fixes #280
Testing APK: https://github.com/Stypox/testing-apks/releases/download/12/app-debug.apk