Implement TTS service with Kokoro integration

- Created TTSService class with pipeline caching per language
- Implemented text segmentation to identify role markers (colon and bracket patterns)
- Implemented segment synthesis with voice assignment
- Implemented audio concatenation at 24 kHz sample rate
- Added fallback to default voice (af_bella) for unmapped roles
- Supports language codes: a, b, f, j
- Handles long text by chunking on sentence boundaries
- Added comprehensive unit tests (13 tests, all passing)

Requirements: 10.1, 10.3, 10.4, 6.2, 6.3, 6.4, 6.5, 6.6

backend/test_tts_service.py

@@ -0,0 +1,163 @@
+"""
+Basic tests for TTS service functionality.
+"""
+import pytest
+import numpy as np
+from services.tts_service import TTSService
+
+
+def test_tts_service_initialization():
+    """Test that TTSService initializes correctly."""
+    service = TTSService()
+    assert service is not None
+    assert service.pipelines == {}
+    assert service.DEFAULT_VOICE == "af_bella"
+    assert service.SAMPLE_RATE == 24000
+
+
+def test_get_pipeline_caching():
+    """Test that pipelines are cached per language."""
+    service = TTSService()
+    
+    # Get pipeline for American English
+    pipeline1 = service.get_pipeline('a')
+    assert pipeline1 is not None
+    assert 'a' in service.pipelines
+    
+    # Get same pipeline again - should return cached version
+    pipeline2 = service.get_pipeline('a')
+    assert pipeline1 is pipeline2
+
+
+def test_get_pipeline_invalid_language():
+    """Test that invalid language codes raise ValueError."""
+    service = TTSService()
+    
+    with pytest.raises(ValueError, match="Unsupported language code"):
+        service.get_pipeline('x')
+
+
+def test_segment_text_colon_pattern():
+    """Test text segmentation with colon pattern."""
+    service = TTSService()
+    text = "Alice: Hello there\nBob: Hi Alice"
+    role_voice_map = {"Alice": "af_bella", "Bob": "am_adam"}
+    
+    segments = service.segment_text(text, role_voice_map)
+    
+    assert len(segments) == 2
+    assert segments[0] == ("Alice", "af_bella", "Hello there")
+    assert segments[1] == ("Bob", "am_adam", "Hi Alice")
+
+
+def test_segment_text_bracket_pattern():
+    """Test text segmentation with bracket pattern."""
+    service = TTSService()
+    text = "[Narrator] The story begins\n[Alice] Hello world"
+    role_voice_map = {"Narrator": "af_sarah", "Alice": "af_bella"}
+    
+    segments = service.segment_text(text, role_voice_map)
+    
+    assert len(segments) == 2
+    assert segments[0] == ("Narrator", "af_sarah", "The story begins")
+    assert segments[1] == ("Alice", "af_bella", "Hello world")
+
+
+def test_segment_text_default_voice_fallback():
+    """Test that unmapped roles fall back to default voice."""
+    service = TTSService()
+    text = "Alice: Hello\nBob: Hi"
+    role_voice_map = {"Alice": "af_bella"}  # Bob not mapped
+    
+    segments = service.segment_text(text, role_voice_map)
+    
+    assert len(segments) == 2
+    assert segments[0] == ("Alice", "af_bella", "Hello")
+    assert segments[1] == ("Bob", "af_bella", "Hi")  # Falls back to default
+
+
+def test_segment_text_no_markers():
+    """Test text without role markers uses Narrator."""
+    service = TTSService()
+    text = "This is plain text without markers"
+    role_voice_map = {"Narrator": "af_sarah"}
+    
+    segments = service.segment_text(text, role_voice_map)
+    
+    assert len(segments) == 1
+    assert segments[0] == ("Narrator", "af_sarah", "This is plain text without markers")
+
+
+def test_chunk_text_short():
+    """Test that short text is not chunked."""
+    service = TTSService()
+    text = "This is a short sentence."
+    
+    chunks = service._chunk_text(text, max_length=500)
+    
+    assert len(chunks) == 1
+    assert chunks[0] == text
+
+
+def test_chunk_text_long():
+    """Test that long text is chunked on sentence boundaries."""
+    service = TTSService()
+    text = "First sentence. Second sentence. Third sentence. Fourth sentence."
+    
+    chunks = service._chunk_text(text, max_length=30)
+    
+    assert len(chunks) > 1
+    # Each chunk should contain complete sentences
+    for chunk in chunks:
+        assert chunk.strip()
+
+
+def test_concatenate_audio_empty():
+    """Test concatenating empty list returns empty array."""
+    service = TTSService()
+    
+    result = service.concatenate_audio([])
+    
+    assert isinstance(result, np.ndarray)
+    assert len(result) == 0
+
+
+def test_concatenate_audio_single():
+    """Test concatenating single segment."""
+    service = TTSService()
+    audio = np.array([0.1, 0.2, 0.3], dtype=np.float32)
+    
+    result = service.concatenate_audio([audio])
+    
+    assert isinstance(result, np.ndarray)
+    assert len(result) == 3
+    np.testing.assert_array_equal(result, audio)
+
+
+def test_concatenate_audio_multiple():
+    """Test concatenating multiple segments."""
+    service = TTSService()
+    audio1 = np.array([0.1, 0.2], dtype=np.float32)
+    audio2 = np.array([0.3, 0.4], dtype=np.float32)
+    audio3 = np.array([0.5, 0.6], dtype=np.float32)
+    
+    result = service.concatenate_audio([audio1, audio2, audio3])
+    
+    assert isinstance(result, np.ndarray)
+    assert len(result) == 6
+    expected = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6], dtype=np.float32)
+    np.testing.assert_array_almost_equal(result, expected)
+
+
+def test_concatenate_audio_filters_empty():
+    """Test that empty segments are filtered out."""
+    service = TTSService()
+    audio1 = np.array([0.1, 0.2], dtype=np.float32)
+    audio2 = np.array([], dtype=np.float32)
+    audio3 = np.array([0.3, 0.4], dtype=np.float32)
+    
+    result = service.concatenate_audio([audio1, audio2, audio3])
+    
+    assert len(result) == 4
+    expected = np.array([0.1, 0.2, 0.3, 0.4], dtype=np.float32)
+    np.testing.assert_array_almost_equal(result, expected)