@brownard I also implemented two tasker actions (and their unit tests) in this branch that stacks on top of this PR. Would you prefer I keep it as a separate PR for later, or push that commit in this PR?

Unit-testing that state condition plugin proved tricky, because of the need to inject a Context

I'm good with this in principle but want to take a bit of care to keep the Tasker stuff as loosely coupled as possible. I'm particularly thinking about the follow on commit that adds a direct dependency in the voice satellite.

I intend to refactor the ESPHome implementation slightly to make the voice satellite a component of an ESPHome device, rather than inheriting from it, which should simplify its dependencies, in the process I can add public methods for wake/stop timer/etc which will hopefully mean the Tasker plugin can call those directly without it needing to be internal to the satellite.

I'll install Tasker and have a play around.

Cool! I wired up the Wake Satellite action on the bixby button, which helps a lot with testing (I don't wake up the other satellites accidentally 😅). Set the bixby button to open the camera: Tasker ships a helper app that fulfills the camera intent and triggers a Tasker event.

For actions, I specifically went with simple callback functions to give us flexibility in refactoring the underlying code: we can easily register them from VoiceSatelliteService once the associated methods are exposed there. I didn't do it in my PR to reduce the diff size, but I could make wakeSatellite and stopTimer public if you prefer.

I'm curious about your advice on unit testing. The two actions were easy to unit test by copying from SatelliteTest.kt, but to test the state plugin we need to run VoiceSatelliteService into an Android Context. requestQuery requires a valid context (it's emitting an Intent, to trigger a pull from Tasker).

I have a plan for the refactoring so I will try and get that done first.

Regarding the testing you can add an instrumentation test in the androidTest source, like I've done for wake word testing here. These run on your connected device/emulator so give you access to a context, etc, but are obviously a bit slower, and are not currently run on Github.

Thanks for this, I have merged it in as is. I will refactor the voice satellite/device in a later pull request but feel free to open further pull requests with additional actions/conditions and I will work around them.