refactor: remove websocket streaming support and clean up associated tests

Overview

This Merge Request removes all deprecated WebSocket streaming functionality from the backend and simplifies the ASR pipeline to support only REST-based transcription and diarization workflows.

The cleanup includes removal of streaming services, PCM16-specific methods, unused configuration, and related tests, while ensuring all existing REST APIs continue to work correctly with full test coverage maintained.

---

Changes Made

WebSocket & Streaming Cleanup

• Removed /stream WebSocket endpoint from app/main.py

• Deleted app/services/streaming_service.py

• Removed streaming session cleanup lifecycle logic

• Removed PCM16 streaming transcription flow from:

  • app/asr_service.py
  • app/asr/router.py
  • ASR engine implementations

ASR Engine Refactor

• Removed:

  • transcribe_pcm16
  • detect_language_pcm16

• Cleaned up unused imports and streaming-only utilities

Configuration Cleanup

Removed obsolete configuration related to:

• STREAMING_*
• VAD_*

from app/config.py

Test & Documentation Updates

• Deleted tests/test_streaming_service.py

• Updated affected unit tests:

  • test_main.py
  • test_engines.py
  • test_asr_service.py
  • test_model_router.py
  • test_whisper_lid_unit.py

• Updated README to remove streaming/WebSocket references

Additional Fixes

• Fixed legacy exception syntax issues:

except ValueError, OSError:

updated to:

except (ValueError, OSError):

• Applied:

  • ruff format
  • ruff check --fix

---

Validation

• All tests passing successfully
• Maintained required coverage threshold (~95%)
• REST transcription and diarization flows verified after cleanup

---

Impact

Benefits

• Reduced backend complexity
• Easier maintenance and debugging
• Cleaner ASR architecture
• Removed unused/dead code paths
• Faster and more stable test execution

No Impact On

• REST transcription APIs
• Diarization workflows
• Benchmark jobs
• Existing ASR model functionality

*Closes #29 (closed)