['artist', 'Led Zeppelin', 'title', 'Stairway To Heaven'] instead of
relying on one single type-filter pair
- Improved condition matching - don't skip the last condition token if
there are still event tokens to add
prevents a PortAudio write on input-only stream error in the Assistant
SDK, see https://github.com/googlesamples/assistant-sdk-python/issues/185
- Make sure that user_request is always defined before returning it
- Use assistant.google.pushtotalk as a fallback if assistant.google is
not configured/available
- More consistent naming for many methods, plus added a more extensive doc.
- Refactored the entry points for the daemon and the pusher into two
classes, easier to encapsulate and wrap into tests.
- Removed the local backend - managing the concurrency of two processes
reading and writing on the same socket at the same time was too much,
and its utility outside of the tests (which can have mock backends as
well) is quite modest.
- Managing stop events on the bus. Still some work to do tho.
- Fixed several bugs.
