Fix echo cancellation android

[alfonsomorab]

Summary

Fixes the infinite echo loop issue when using SDK .speak() calls on Android by implementing STT (Speech-to-Text) processing pause during TTS playbook with comprehensive reliability safeguards.

Note: This PR addresses the Android implementation only. iOS still needs this solution implemented.

Problem

When using SDK .speak() calls with phone speaker + phone microphone:

1. User says "hello"
2. App calls session.audio.speak("hello")
3. TTS plays through phone speaker
4. Phone microphone captures TTS audio
5. Speech recognition processes captured audio as new speech
6. App responds with another .speak() call → Infinite loop

Solution

Instead of trying to disable microphone hardware (which had timing and acoustic coupling issues), this solution pauses STT processing during SDK .speak() calls:

• Mobile side: AudioManagerModule sends STT pause/resume commands via broadcast
• Android Core: AugmentosService receives commands and posts PauseAsrEvent
• Leverages existing infrastructure: Uses the same STT pause system that native TTS already uses successfully
• Zero latency: Mobile-only solution with no network calls
• Multi-layered reliability: Comprehensive error handling and safety mechanisms

Architecture

SDK .speak() → Mobile AudioManagerModule → Broadcast Intent →
AugmentosService STTControlReceiver → PauseAsrEvent →
SpeechRecSwitchSystem → SpeechRecFramework.pauseAsr()

Files Changed

• mobile/android/app/src/main/java/com/mentra/mentra/AudioManagerModule.java: Added STT pause/resume commands with reliability mechanisms
• android_core/app/src/main/java/com/augmentos/augmentos_core/AugmentosService.java: Added STT control BroadcastReceiver with Android 14+ compatibility

Edge Cases Handled

The solution includes comprehensive safeguards for reliability:

Failure Scenarios

• Failed audio requests: STT resumes immediately if playAudio() throws exception
• Network/URL errors: Timeout safety resumes STT after 30 seconds if audio fails to load
• Rapid .speak() calls: State tracking prevents redundant commands; timer cancellation prevents race conditions
• Service crashes: STT pause state is managed independently and will resume via timeout
• Audio loading failures: If audio fails during playback, timeout mechanism ensures STT eventually resumes

Reliability Mechanisms

• Error handling: Try/catch in playAudio() ensures STT resumes on immediate failures
• Success-based resume: STT only resumes when audio completes successfully
• State tracking: Prevents redundant pause/resume commands from conflicting
• Safety timeout: 30-second auto-resume mechanism for any stuck states
• Timer management: Proper cancellation prevents old timers from interfering with new audio
• Android 14+ compatibility: Uses RECEIVER_NOT_EXPORTED for proper BroadcastReceiver registration

Test Plan

• Tested with phone speaker + microphone configuration on Android
• Verified SDK .speak() calls no longer create infinite loops
• Confirmed STT resumes correctly after TTS completes successfully
• Validated STT resumes after failed audio requests
• Tested rapid consecutive .speak() calls don't cause issues
• Verified timeout safety mechanism works for edge cases
• Confirmed existing native TTS functionality remains unaffected

Testing App

This solution can be tested using the MentraOS Echo Test App:
https://github.com/alfonsomorab/MentraOS-Echo-App

The test app demonstrates the echo loop scenario and validates the fix across various edge cases.

Platform Support

• ✅ Android: Fixed with this PR
• ⚠️ iOS: Still needs implementation (similar approach should work)

Related Issues

Resolves #1036 - SDK .speak() infinite echo loop (Android only)