Added speech synthesis bypass engine

[compulim]

Fixes #2435.

Changelog Entry

Fixed

• Fix #2435. Fix microphone button stuck on voice-triggered expecting input hint without a speech synthesis engine, by @compulim in PR #2445

Description

On a system without speech synthesis engine:

• Queue a message to be synthesized
• Open microphone in a programmatic way and not clicking on microphone button

Since speech synthesis is not available, the queued message is pending to be synthesized and never complete. Internally, we will wait until all queued messages are synthesized before turning on the microphone. This is to prevent echo (bot hearing itself talking). Thus, the microphone will get stuck and never turned on.

The speech synthesis engine is configured on the component side. When the saga in the core side tries to perform synthesis, it has no knowledge whether the engine is available or not.

This can be reproduced by:

• Configure off speech synthesis
• Click microphone button and trigger the bot to respond with expecting input hint
• Click microphone button again

Since no speech synthesis engine is configured, the bot message is being queued but never complete. And the microphone is queued to open after all synthesis is completed. Thus, the microphone button is disabled and never open.

Specific Changes

If speech synthesis engine is not configured, we will provide a bypass engine.

The bypass engine will act just like the real engine, except all synthesis operation will be completed instantly. This will unblock any future issues related to queued messages on a configured off system.

In the future, we could consider moving all speech operations to component side to prevent this issue from happening.

---

• Testing Added

[corinagum]

LGTM