Bump Speech SDK 1.15.1 to support sovereign clouds

[compulim]

Fixes #3666. Fixes #3421. Fixes #3411.

Changelog Entry

Added

• Resolves #3666. Added support of sovereign clouds when using Direct Line Speech, by @compulim in PR #3694
  • Please refer to DIRECT_LINE_SPEECH.md for instructions

Fixed

• Fixes #3411. With Direct Line Speech, clicking on microphone button during speech recognition should no longer stop working, by @compulim in PR #3694
  • Although it no locker lock up microphone, clicking on the microphone button has no effect because Direct Line Speech does not support aborting speech recognition
• Fixes #3421. With Direct Line Speech, after not able to recognize any speech, it should no longer stop working, by @compulim in PR #3694

Changed

• Bumped all dependencies to the latest versions, by @compulim in PR #3594 and PR #3694
  • Production dependencies
    • [email protected]
    • [email protected]

Description

Bump Speech SDK to 1.15.0 to add sovereign cloud support and fixes several issues.

Design

Speech SDK 1.15.0 uses ES Promise interface, instead of a Promise-like interface ported from C#. We updated our test harness to support ES Promise.

We also added sovereign clouds support for Direct Line Speech. The fetchCredentials option passed to DirectLineSpeechSDK.createAdapters() call should return directLineSpeechHostname, instead of region. For example:

return { authorizationToken: '...', directLineSpeechHostname: 'virginia.convai.speech.azure.us' };

This change is outlined in the updated DIRECT_LINE_SPEECH.md.

There are few tracking issues open due to bugs in 1.15.0: #3692 and #3693.

One of the issue requires us to hardcode the audio output format. We will check if the output format is wrong: if SDK pass us the output format of MP3, it must be wrong, because MP3 is not a streamable format and cannot be used in Direct Line Speech. Then we hardcode it with RIFF 16kHz instead.

One potential risk: in the future, if the service decided to send us 44.1kHz (or any other values) instead. This version of Web Chat will still use the hardcoded value 16kHz.

I am not foreseeing the service will update to 44.1kHz sooner than the SDK resolving the issue and Web Chat bumping the SDK. So, IMO, the risk is small.

Specific Changes

• Bumped [email protected]
• Bumped [email protected], changes can be found here
  • Bumped [email protected] for peerDependencies
  • Bumped all dependencies to latest versions
  • Fixed race conditions for using different implementation of Promise interface
  • Updated AudioConfig to newer interface by 1.15.0
  • Updated Custom Voice API endpoint from cris.ai to customvoice.speech.microsoft.com
  • Moved build pipeline from Travis to GitHub Actions
• Added tests for testing speech between hostname and region
• Updated microphone button CSS: will now lit up in red during dictation using Direct Line Speech
  • Previously, it didn't lit up because it is disabled (Direct Line Speech does not support abort, thus, the button is disabled)
• Removed workaround code for Speech SDK 1.12.0-1.13.1
  • Safari was not supported in previous version of SDKs because AudioContext was vendor-prefixed in Safari

• I have added tests and executed them locally
• I have updated CHANGELOG.md
• I have updated documentation

Review Checklist

This section is for contributors to review your work.

• Accessibility reviewed (tab order, content readability, alt text, color contrast)
• Browser and platform compatibilities reviewed
• Tested on Edge (Chromium), Chrome, and iOS Safari only. More intensive tests needed during release testing
• CSS styles reviewed (minimal rules, no z-index)
• Documents reviewed (docs, samples, live demo)
• Internationalization reviewed (strings, unit formatting)
• package.json and package-lock.json reviewed
• Security reviewed (no data URIs, check for nonce leak)
• Tests reviewed (coverage, legitimacy)

[corinagum]

LGTM pending a few comments.