Opt in voice mode support with /multimodal endpoint

[pranavjoshi001]

Description

This change adds voice mode support to DirectLineJS with a client opt-in mechanism via the enableVoiceMode option. When enabled, it allows audio streaming through WebSocket connections using the /stream/multimodal endpoint.

Background

• DirectLineJS currently routes all activities through HTTP POST instead of WebSocket due to limitations on the ABS side.
• ABS does not process incoming WebSocket traffic; it only supports server-to-client push.
• Voice traffic is not supported over HTTP POST and must be sent through WebSocket instead of API calls.
• This PR introduces an opt-in voice mode that uses the /stream/multimodal endpoint and routes all traffic through WebSocket when enabled.

Changes in this PR

• Added enableVoiceMode option:

  • true → Enables voice mode
  • false → Disables voice mode
  • undefined → Auto-detects iframe microphone permission

• Enhanced stream URL:

  • Voice mode → /stream/multimodal
  • Standard mode → /stream

• Modified activity routing:

  • Voice mode → Sends all activities (text + voice) via WebSocket
  • Standard mode → Uses HTTP POST

• Added server capabilities handling:

  • Parses agent.capabilities event with modalities object to detect audio support

• Added new public methods:

  • getIsVoiceModeEnabled()
  • getVoiceConfiguration()
  • addEventListener()
  • removeEventListener()

• Added test coverage:

  • Explicit enableVoiceMode: true/false
  • Auto-detect in iframe
  • WebSocket vs HTTP routing verification
  • Reconnect behavior
  • 403 retry handling
  • agent.capabilities event handling

Backward Compatibility

• No breaking changes:

  • Default behavior (enableVoiceMode: undefined in non-iframe context) maintains existing HTTP POST flow

• Opt-in only:

  • Voice mode must be explicitly enabled or auto-detected in iframe with microphone permission

[compulim]

Commented.