整理: TTSEngineManager クラスの新設 (#1234)

* refactor: `TTSEngines` クラスの新設

* add: test

* fix: `TTSEngines` → `TTSEngineManager` へリネーム

* fix: リネーム忘れを修正

* fix: TTSEngineManager の property を関数へ変更

build_util/make_docs.py

@@ -6,7 +6,7 @@
 from voicevox_engine.dev.tts_engine.mock import MockTTSEngine
 from voicevox_engine.preset.PresetManager import PresetManager
 from voicevox_engine.setting.SettingLoader import USER_SETTING_PATH, SettingHandler
-from voicevox_engine.tts_pipeline.tts_engine import CoreAdapter
+from voicevox_engine.tts_pipeline.tts_engine import CoreAdapter, TTSEngineManager
 from voicevox_engine.user_dict.user_dict import UserDictionary
 from voicevox_engine.utility.path_utility import engine_root
 
@@ -36,9 +36,11 @@ def generate_api_docs_html(schema: str) -> str:
 
 if __name__ == "__main__":
     mock_core = MockCoreWrapper()
+    tts_engines = TTSEngineManager()
+    tts_engines.register_engine(MockTTSEngine(), "mock")
     # FastAPI の機能を用いて OpenAPI schema を生成する
     app = generate_app(
-        tts_engines={"mock": MockTTSEngine()},
+        tts_engines=tts_engines,
         cores={"mock": CoreAdapter(mock_core)},
         latest_core_version="mock",
         setting_loader=SettingHandler(USER_SETTING_PATH),

run.py

@@ -19,7 +19,6 @@
 from voicevox_engine.setting.SettingLoader import USER_SETTING_PATH, SettingHandler
 from voicevox_engine.tts_pipeline.tts_engine import make_tts_engines_from_cores
 from voicevox_engine.user_dict.user_dict import UserDictionary
-from voicevox_engine.utility.core_version_utility import get_latest_version
 from voicevox_engine.utility.path_utility import engine_root
 
 
@@ -267,8 +266,8 @@ def main() -> None:
         load_all_models=load_all_models,
     )
     tts_engines = make_tts_engines_from_cores(cores)
-    assert len(tts_engines) != 0, "音声合成エンジンがありません。"
-    latest_core_version = get_latest_version(list(tts_engines.keys()))
+    assert len(tts_engines.versions()) != 0, "音声合成エンジンがありません。"
+    latest_core_version = tts_engines.latest_version()
 
     # Cancellable Engine
     enable_cancellable_synthesis: bool = args.enable_cancellable_synthesis

test/e2e/conftest.py

@@ -12,14 +12,13 @@
 from voicevox_engine.setting.SettingLoader import SettingHandler
 from voicevox_engine.tts_pipeline.tts_engine import make_tts_engines_from_cores
 from voicevox_engine.user_dict.user_dict import UserDictionary
-from voicevox_engine.utility.core_version_utility import get_latest_version
 
 
 @pytest.fixture()
 def app_params(tmp_path: Path) -> dict[str, Any]:
     cores = initialize_cores(use_gpu=False, enable_mock=True)
     tts_engines = make_tts_engines_from_cores(cores)
-    latest_core_version = get_latest_version(list(tts_engines.keys()))
+    latest_core_version = tts_engines.latest_version()
     setting_loader = SettingHandler(Path("./not_exist.yaml"))
 
     # 隔離されたプリセットの生成

test/tts_pipeline/test_tts_engines.py

@@ -0,0 +1,124 @@
+""" `TTSEngineManager` クラスのテスト"""
+
+import pytest
+from fastapi import HTTPException
+
+from voicevox_engine.dev.tts_engine.mock import MockTTSEngine
+from voicevox_engine.tts_pipeline.tts_engine import TTSEngineManager
+
+
+def test_tts_engines_register_engine() -> None:
+    """TTSEngineManager.register_engine() で TTS エンジンを登録できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+
+    # Test
+    tts_engines.register_engine(MockTTSEngine(), "0.0.1")
+
+
+def test_tts_engines_versions() -> None:
+    """TTSEngineManager.versions() でバージョン一覧を取得できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engines.register_engine(MockTTSEngine(), "0.0.1")
+    tts_engines.register_engine(MockTTSEngine(), "0.0.2")
+    # Expects
+    true_versions = ["0.0.1", "0.0.2"]
+    # Outputs
+    versions = tts_engines.versions()
+
+    # Test
+    assert true_versions == versions
+
+
+def test_tts_engines_latest_version() -> None:
+    """TTSEngineManager.latest_version() で最新バージョンを取得できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engines.register_engine(MockTTSEngine(), "0.0.1")
+    tts_engines.register_engine(MockTTSEngine(), "0.0.2")
+    # Expects
+    true_latest_version = "0.0.2"
+    # Outputs
+    latest_version = tts_engines.latest_version()
+
+    # Test
+    assert true_latest_version == latest_version
+
+
+def test_tts_engines_get_engine_specified() -> None:
+    """TTSEngineManager.get_engine() で登録済み TTS エンジンをバージョン指定して取得できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engine1 = MockTTSEngine()
+    tts_engine2 = MockTTSEngine()
+    tts_engines.register_engine(tts_engine1, "0.0.1")
+    tts_engines.register_engine(tts_engine2, "0.0.2")
+    # Expects
+    true_acquired_tts_engine = tts_engine2
+    # Outputs
+    acquired_tts_engine = tts_engines.get_engine("0.0.2")
+
+    # Test
+    assert true_acquired_tts_engine == acquired_tts_engine
+
+
+def test_tts_engines_get_engine_latest() -> None:
+    """TTSEngineManager.get_engine() で最新版 TTS エンジンをバージョン未指定で取得できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engine1 = MockTTSEngine()
+    tts_engine2 = MockTTSEngine()
+    tts_engines.register_engine(tts_engine1, "0.0.1")
+    tts_engines.register_engine(tts_engine2, "0.0.2")
+    # Expects
+    true_acquired_tts_engine = tts_engine2
+    # Outputs
+    acquired_tts_engine = tts_engines.get_engine()
+
+    # Test
+    assert true_acquired_tts_engine == acquired_tts_engine
+
+
+def test_tts_engines_get_engine_missing() -> None:
+    """TTSEngineManager.get_engine() で存在しない TTS エンジンを取得しようとするとエラーになる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engine1 = MockTTSEngine()
+    tts_engine2 = MockTTSEngine()
+    tts_engines.register_engine(tts_engine1, "0.0.1")
+    tts_engines.register_engine(tts_engine2, "0.0.2")
+
+    # Test
+    with pytest.raises(HTTPException) as _:
+        tts_engines.get_engine("0.0.3")
+
+
+def test_tts_engines_has_engine_true() -> None:
+    """TTSEngineManager.has_engine() で TTS エンジンが登録されていることを確認できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engines.register_engine(MockTTSEngine(), "0.0.1")
+    tts_engines.register_engine(MockTTSEngine(), "0.0.2")
+    # Expects
+    true_has = True
+    # Outputs
+    has = tts_engines.has_engine("0.0.1")
+
+    # Test
+    assert true_has == has
+
+
+def test_tts_engines_has_engine_false() -> None:
+    """TTSEngineManager.has_engine() で TTS エンジンが登録されていないことを確認できる。"""
+    # Inputs
+    tts_engines = TTSEngineManager()
+    tts_engines.register_engine(MockTTSEngine(), "0.0.1")
+    tts_engines.register_engine(MockTTSEngine(), "0.0.2")
+    # Expects
+    true_has = False
+    # Outputs
+    has = tts_engines.has_engine("0.0.3")
+
+    # Test
+    assert true_has == has

voicevox_engine/app/application.py

@@ -23,13 +23,13 @@
 from voicevox_engine.preset.PresetManager import PresetManager
 from voicevox_engine.setting.Setting import CorsPolicyMode
 from voicevox_engine.setting.SettingLoader import SettingHandler
-from voicevox_engine.tts_pipeline.tts_engine import TTSEngine
+from voicevox_engine.tts_pipeline.tts_engine import TTSEngineManager
 from voicevox_engine.user_dict.user_dict import UserDictionary
 from voicevox_engine.utility.path_utility import engine_root, get_save_dir
 
 
 def generate_app(
-    tts_engines: dict[str, TTSEngine],
+    tts_engines: TTSEngineManager,
     cores: dict[str, CoreAdapter],
     latest_core_version: str,
     setting_loader: SettingHandler,
@@ -69,13 +69,6 @@ def generate_app(
 
     metas_store = MetasStore(root_dir / "speaker_info")
 
-    def get_engine(core_version: str | None) -> TTSEngine:
-        if core_version is None:
-            return tts_engines[latest_core_version]
-        if core_version in tts_engines:
-            return tts_engines[core_version]
-        raise HTTPException(status_code=422, detail="不明なバージョンです")
-
     def get_core(core_version: str | None) -> CoreAdapter:
         """指定したバージョンのコアを取得する"""
         if core_version is None:
@@ -86,10 +79,10 @@ def get_core(core_version: str | None) -> CoreAdapter:
 
     app.include_router(
         generate_tts_pipeline_router(
-            get_engine, get_core, preset_manager, cancellable_engine
+            tts_engines, get_core, preset_manager, cancellable_engine
         )
     )
-    app.include_router(generate_morphing_router(get_engine, get_core, metas_store))
+    app.include_router(generate_morphing_router(tts_engines, get_core, metas_store))
     app.include_router(generate_preset_router(preset_manager))
     app.include_router(generate_speaker_router(get_core, metas_store, root_dir))
     if engine_manifest_data.supported_features.manage_library:

voicevox_engine/app/routers/morphing.py

@@ -21,7 +21,7 @@
 from voicevox_engine.morphing import (
     synthesis_morphing_parameter as _synthesis_morphing_parameter,
 )
-from voicevox_engine.tts_pipeline.tts_engine import TTSEngine
+from voicevox_engine.tts_pipeline.tts_engine import TTSEngineManager
 from voicevox_engine.utility.path_utility import delete_file
 
 # キャッシュを有効化
@@ -31,7 +31,7 @@
 
 
 def generate_morphing_router(
-    get_engine: Callable[[str | None], TTSEngine],
+    tts_engines: TTSEngineManager,
     get_core: Callable[[str | None], CoreAdapter],
     metas_store: MetasStore,
 ) -> APIRouter:
@@ -94,7 +94,7 @@ def _synthesis_morphing(
         指定された2種類のスタイルで音声を合成、指定した割合でモーフィングした音声を得ます。
         モーフィングの割合は`morph_rate`で指定でき、0.0でベースのスタイル、1.0でターゲットのスタイルに近づきます。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         core = get_core(core_version)
 
         try:

voicevox_engine/app/routers/tts_pipeline.py

@@ -27,12 +27,12 @@
     connect_base64_waves,
 )
 from voicevox_engine.tts_pipeline.kana_converter import create_kana, parse_kana
-from voicevox_engine.tts_pipeline.tts_engine import TTSEngine
+from voicevox_engine.tts_pipeline.tts_engine import TTSEngineManager
 from voicevox_engine.utility.path_utility import delete_file
 
 
 def generate_tts_pipeline_router(
-    get_engine: Callable[[str | None], TTSEngine],
+    tts_engines: TTSEngineManager,
     get_core: Callable[[str | None], CoreAdapter],
     preset_manager: PresetManager,
     cancellable_engine: CancellableEngine | None,
@@ -53,7 +53,7 @@ def audio_query(
         """
         音声合成用のクエリの初期値を得ます。ここで得られたクエリはそのまま音声合成に利用できます。各値の意味は`Schemas`を参照してください。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         core = get_core(core_version)
         accent_phrases = engine.create_accent_phrases(text, style_id)
         return AudioQuery(
@@ -82,7 +82,7 @@ def audio_query_from_preset(
         """
         音声合成用のクエリの初期値を得ます。ここで得られたクエリはそのまま音声合成に利用できます。各値の意味は`Schemas`を参照してください。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         core = get_core(core_version)
         try:
             presets = preset_manager.load_presets()
@@ -139,7 +139,7 @@ def accent_phrases(
         * アクセント位置を`'`で指定する。全てのアクセント句にはアクセント位置を1つ指定する必要がある。
         * アクセント句末に`？`(全角)を入れることにより疑問文の発音ができる。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         if is_kana:
             try:
                 return engine.create_accent_phrases_from_kana(text, style_id)
@@ -160,7 +160,7 @@ def mora_data(
         style_id: Annotated[StyleId, Query(alias="speaker")],
         core_version: str | None = None,
     ) -> list[AccentPhrase]:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         return engine.update_length_and_pitch(accent_phrases, style_id)
 
     @router.post(
@@ -173,7 +173,7 @@ def mora_length(
         style_id: Annotated[StyleId, Query(alias="speaker")],
         core_version: str | None = None,
     ) -> list[AccentPhrase]:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         return engine.update_length(accent_phrases, style_id)
 
     @router.post(
@@ -186,7 +186,7 @@ def mora_pitch(
         style_id: Annotated[StyleId, Query(alias="speaker")],
         core_version: str | None = None,
     ) -> list[AccentPhrase]:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         return engine.update_pitch(accent_phrases, style_id)
 
     @router.post(
@@ -213,7 +213,7 @@ def synthesis(
         ] = True,
         core_version: str | None = None,
     ) -> FileResponse:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         wave = engine.synthesize_wave(
             query, style_id, enable_interrogative_upspeak=enable_interrogative_upspeak
         )
@@ -285,7 +285,7 @@ def multi_synthesis(
         style_id: Annotated[StyleId, Query(alias="speaker")],
         core_version: str | None = None,
     ) -> FileResponse:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         sampling_rate = queries[0].outputSamplingRate
 
         with NamedTemporaryFile(delete=False) as f:
@@ -327,7 +327,7 @@ def sing_frame_audio_query(
         """
         歌唱音声合成用のクエリの初期値を得ます。ここで得られたクエリはそのまま歌唱音声合成に利用できます。各値の意味は`Schemas`を参照してください。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         core = get_core(core_version)
         phonemes, f0, volume = engine.create_sing_phoneme_and_f0_and_volume(
             score, style_id
@@ -353,7 +353,7 @@ def sing_frame_volume(
         style_id: Annotated[StyleId, Query(alias="speaker")],
         core_version: str | None = None,
     ) -> list[float]:
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         return engine.create_sing_volume_from_phoneme_and_f0(
             score, frame_audio_query.phonemes, frame_audio_query.f0, style_id
         )
@@ -378,7 +378,7 @@ def frame_synthesis(
         """
         歌唱音声合成を行います。
         """
-        engine = get_engine(core_version)
+        engine = tts_engines.get_engine(core_version)
         wave = engine.frame_synthsize_wave(query, style_id)
 
         with NamedTemporaryFile(delete=False) as f:

voicevox_engine/cancellable_engine.py

@@ -20,7 +20,6 @@
 from .metas.Metas import StyleId
 from .model import AudioQuery
 from .tts_pipeline.tts_engine import make_tts_engines_from_cores
-from .utility.core_version_utility import get_latest_version
 
 
 class CancellableEngine:
@@ -238,15 +237,14 @@ def start_synthesis_subprocess(
     )
     tts_engines = make_tts_engines_from_cores(cores)
 
-    assert len(tts_engines) != 0, "音声合成エンジンがありません。"
-    latest_core_version = get_latest_version(list(tts_engines.keys()))
+    assert len(tts_engines.versions()) != 0, "音声合成エンジンがありません。"
     while True:
         try:
             query, style_id, core_version = sub_proc_con.recv()
             if core_version is None:
-                _engine = tts_engines[latest_core_version]
-            elif core_version in tts_engines:
-                _engine = tts_engines[core_version]
+                _engine = tts_engines.get_engine()
+            elif tts_engines.has_engine(core_version):
+                _engine = tts_engines.get_engine(core_version)
             else:
                 # バージョンが見つからないエラー
                 sub_proc_con.send("")