From 539c71f5cb11bac2bb2dcdcfb758c8ad182498a7 Mon Sep 17 00:00:00 2001 From: xiaoxia Date: Sun, 30 Aug 2026 16:12:57 +0800 Subject: [PATCH] =?UTF-8?q?feat(tts):=20=E6=AD=A3=E5=BC=8F=E5=90=88?= =?UTF-8?q?=E6=88=90=E6=94=AF=E6=8C=81=E5=85=8B=E9=9A=86=E9=9F=B3=E8=89=B2?= =?UTF-8?q?=E3=80=81=E4=BF=9D=E5=AD=98=E5=88=B0=E9=85=8D=E9=9F=B3=E5=BA=93?= =?UTF-8?q?=E6=94=B9=E5=86=99=20assets=20=E7=B4=A0=E6=9D=90=E4=BD=93?= =?UTF-8?q?=E7=B3=BB=E3=80=81=E5=85=8B=E9=9A=86=E6=8E=A5=E5=8F=A3=E6=94=AF?= =?UTF-8?q?=E6=8C=81=20asset=5Fid?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 1. TTS synthesize() 复用 preview 的克隆音色解析逻辑：voice_id 传克隆 profile UUID 时 校验归属(403)/克隆完成状态(400 音色克隆尚未完成)，job.voice_id 存解析后的 CosyVoice voice_id，voice_clone_profile_id 记录 profile id。 2. save_tts_job_to_library 改写为新素材体系：TTS 音频持久化独立副本到 uploads/voice/tts/，创建 file_type=audio/status=ready 的 asset 挂用户 voice 素材库（无则自动创建），duration/metadata(source=tts_job) 完整记录； 返回结构保持 id/name/audio_url/duration 兼容前端；移除旧 voice_library 表写入。 3. voice clone 创建接口支持 asset_id 字段：校验素材归属与音频类型，用 asset storage_key 生成 7 天有效签名 URL 供 worker 下载，不依赖前端签名； source_audio_url 方式保留兼容。 补集成测试：synthesize 克隆音色 4 用例、save-to-library asset 断言、 clone from asset 5 用例（成功/404/403/非音频400/两参同传400）。 --- apps/api/app/api/routes/tts.py | 222 ++++++++++++++++------ apps/api/app/api/routes/voice_clones.py | 63 +++++- apps/api/app/schemas/voice_clone.py | 3 +- tests/integration/test_tts_api.py | 213 ++++++++++++++++++++- tests/integration/test_voice_clone_api.py | 156 ++++++++++++++- 5 files changed, 581 insertions(+), 76 deletions(-) diff --git a/apps/api/app/api/routes/tts.py b/apps/api/app/api/routes/tts.py index eb9581e07..17ddcde40 100755 --- a/apps/api/app/api/routes/tts.py +++ b/apps/api/app/api/routes/tts.py @@ -7,13 +7,15 @@ from typing import Optional from app.auth import AuthenticatedUser, get_current_user from app.core.celery_app import celery_app +from app.core.storage import get_storage_service from app.dependencies import ( + get_asset_library_repository, + get_asset_repository, get_audio_url_signer, get_cosyvoice_service, get_db_session, - get_user_repository, + get_project_repository, get_voice_clone_profile_repository, - get_voice_library_repository, ) from app.schemas.tts import ( ListTTSJobResponse, @@ -32,7 +34,6 @@ from sqlalchemy.orm import Session from packages.adapters.sqlalchemy_impl.tts_job_repository import ( SQLAlchemyTTSJobRepository, ) -from packages.adapters.sqlalchemy_impl.voice_library_repository import SQLAlchemyVoiceLibraryRepository from packages.application.cosyvoice_service import CosyVoiceError, CosyVoiceService from packages.application.tts_job.streaming_service import TTSStreamingService from packages.application.tts_job.use_cases import ( @@ -44,13 +45,12 @@ from packages.application.tts_job.use_cases import ( TTSJobNotFoundError, ) from packages.application.tts_job.workflow import TTSWorkflowService -from packages.application.voice_library.commands import CreateVoiceLibraryCommand -from packages.application.voice_library.use_cases import ( - CreateVoiceLibraryUseCase, - QuotaExceededError, -) +from packages.domain import Asset, AssetLibrary, AssetLibraryKind, AssetStatus, ClassificationStatus from packages.domain.voice_presets import list_voices -from packages.ports.user_repository import UserRepository +from packages.ports.asset_library_repository import AssetLibraryRepository +from packages.ports.asset_repository import AssetRepository +from packages.ports.project_repository import ProjectRepository +from packages.shared.storage import SharedStorageService logger = logging.getLogger(__name__) @@ -134,27 +134,47 @@ def synthesize( """ user_id = authenticated_user.user.id - # 校验 voice_clone_profile_id 归属（防止越权使用他人克隆音色） - if request.voice_clone_profile_id: - profile = voice_clone_repo.get(request.voice_clone_profile_id) - if profile is None: + # 解析 voice_id：前端可能传克隆音色 profile UUID（而非 CosyVoice voice_id）， + # 与 /tts/preview 保持一致：命中 profile → 校验归属 → 取 CosyVoice voice_id + actual_voice_id = request.voice_id + voice_clone_profile_id = request.voice_clone_profile_id + resolved_profile = None + if actual_voice_id: + resolved_profile = voice_clone_repo.get(actual_voice_id) + if resolved_profile is not None: + voice_clone_profile_id = actual_voice_id + + # 显式传了 voice_clone_profile_id（且与 voice_id 不同）时再查一次归属 + if voice_clone_profile_id and (resolved_profile is None or resolved_profile.id != voice_clone_profile_id): + resolved_profile = voice_clone_repo.get(voice_clone_profile_id) + if resolved_profile is None: raise HTTPException( status_code=status.HTTP_404_NOT_FOUND, detail="Voice clone profile not found", ) - if profile.user_id != user_id: + + if resolved_profile is not None: + if resolved_profile.user_id != user_id: raise HTTPException( status_code=status.HTTP_403_FORBIDDEN, - detail="Access denied to voice clone profile", + detail="无权访问该音色", ) + if not resolved_profile.voice_id: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="音色克隆尚未完成，请稍后再试", + ) + # 命中克隆音色：job 存解析后的 CosyVoice voice_id + if voice_clone_profile_id == actual_voice_id: + actual_voice_id = resolved_profile.voice_id use_case = CreateTTSJobUseCase(repository) job = use_case.execute( user_id=user_id, input_text=request.text, - voice_id=request.voice_id, + voice_id=actual_voice_id, voice_model=request.voice_model, - voice_clone_profile_id=request.voice_clone_profile_id, + voice_clone_profile_id=voice_clone_profile_id, metadata=request.metadata_, ) @@ -284,6 +304,41 @@ def delete_tts_job( return +def _find_or_create_voice_library( + *, + user_id: str, + project_repository: ProjectRepository, + asset_library_repository: AssetLibraryRepository, +) -> AssetLibrary: + """在用户可访问的项目中找到（或自动创建）voice 素材库。 + + 与前端配音素材页逻辑一致：素材库挂在项目下，配音素材读取 + getAssetsByKind("voice") → 用户所有可访问项目中的 voice 库。 + 优先使用已有 voice 库；没有则在第一个可访问项目中自动创建。 + """ + projects = project_repository.find_accessible_projects(user_id) + if not projects: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="没有可用的项目，请先创建项目后再保存配音素材", + ) + + for project in projects: + for lib in asset_library_repository.find_by_project(project.id): + kind = lib.kind.value if hasattr(lib.kind, "value") else lib.kind + if kind == AssetLibraryKind.VOICE.value: + return lib + + # 所有项目都没有 voice 库 → 在第一个可访问项目中自动创建默认配音素材库 + project = projects[0] + library = AssetLibrary.create( + project_id=project.id, + name="配音素材库", + kind=AssetLibraryKind.VOICE, + ) + return asset_library_repository.create(library) + + @router.post( "/jobs/{job_id}/save-to-library", response_model=SaveToLibraryResponse, @@ -294,13 +349,17 @@ def save_tts_job_to_library( request: SaveToLibraryRequest = SaveToLibraryRequest(), authenticated_user: AuthenticatedUser = Depends(get_current_user), tts_repository: SQLAlchemyTTSJobRepository = Depends(_get_repository), - voice_library_repository: SQLAlchemyVoiceLibraryRepository = Depends(get_voice_library_repository), - user_repository: UserRepository = Depends(get_user_repository), + asset_repository: AssetRepository = Depends(get_asset_repository), + asset_library_repository: AssetLibraryRepository = Depends(get_asset_library_repository), + project_repository: ProjectRepository = Depends(get_project_repository), + storage_service: SharedStorageService = Depends(get_storage_service), sign_url=Depends(get_audio_url_signer), ) -> SaveToLibraryResponse: - """将已完成的 TTS 合成结果保存到配音库。 + """将已完成的 TTS 合成结果保存到配音素材库（assets 表新素材体系）。 - 自动携带音色名、时长、语速等元信息。 + 流程：把 TTS 输出音频转存到用户素材 OSS 路径 → 创建 file_type=audio、 + status=ready 的 asset（挂用户 voice 素材库）→ 返回前端可用结构。 + 配额策略与素材上传一致（上传/ingest 链路无额外配额拦截）。 """ user_id = authenticated_user.user.id @@ -318,64 +377,107 @@ def save_tts_job_to_library( detail="TTS job is not completed yet", ) - # 构建配音素材名称 + if not job.output_audio_url and not job.output_audio_key: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="TTS job 缺少输出音频，无法保存", + ) + + # 素材名称 name = request.name or f"TTS-{job.id[:8]}" - # 构建元信息 - metadata_ = { + # 找到（或自动创建）用户 voice 素材库 + library = _find_or_create_voice_library( + user_id=user_id, + project_repository=project_repository, + asset_library_repository=asset_library_repository, + ) + + # 转存音频到素材 OSS 路径（tts-outputs/ 下的产物归 TTS 任务所有， + # 素材独立持有副本，删除 TTS 任务不影响配音库素材） + import tempfile + from pathlib import Path + + audio_format = (job.format or "mp3").strip() or "mp3" + content_type_map = { + "mp3": "audio/mpeg", + "wav": "audio/wav", + "pcm": "audio/pcm", + "opus": "audio/opus", + } + content_type = content_type_map.get(audio_format, "audio/mpeg") + storage_key = f"uploads/voice/tts/{job.id}.{audio_format}" + + tmp_path: Path | None = None + try: + with tempfile.NamedTemporaryFile(suffix=f".{audio_format}", delete=False) as tmp: + tmp_path = Path(tmp.name) + # 优先用 OSS storage_key（走 oss2 SDK，私有 bucket 也可下载）； + # 兜底用 output_audio_url（旧任务可能没有 key） + download_source = job.output_audio_key or job.output_audio_url + downloaded = storage_service.download_asset(download_source, tmp_path) + if not downloaded or not tmp_path.exists() or tmp_path.stat().st_size == 0: + raise HTTPException( + status_code=status.HTTP_502_BAD_GATEWAY, + detail="TTS 音频下载失败，无法保存到配音库", + ) + file_size = tmp_path.stat().st_size + storage_service.upload_file(tmp_path, storage_key, content_type=content_type) + except HTTPException: + raise + except Exception as e: + logger.error("TTS 音频转存素材失败: job_id=%s, error=%s", job.id, e, exc_info=True) + raise HTTPException( + status_code=status.HTTP_502_BAD_GATEWAY, + detail="TTS 音频转存失败，无法保存到配音库", + ) from e + finally: + if tmp_path and tmp_path.exists(): + try: + tmp_path.unlink() + except OSError: + pass + + # 构建素材元信息 + metadata_: dict[str, object] = { "source": "tts_job", "tts_job_id": job.id, "format": job.format, "sample_rate": job.sample_rate, + "voice_id": job.voice_id, + "voice_name": job.voice_model or "", } if job.metadata: - # 保留原始 job 的有用元信息 for key in ("speed", "language"): if key in job.metadata: metadata_[key] = job.metadata[key] - # 获取用户套餐（用于配额检查） - user = user_repository.find_by_id(user_id) - plan_name = getattr(user, "subscription_plan", "free") if user else "free" - - # 构建命令并执行 - command = CreateVoiceLibraryCommand( - user_id=user_id, + asset = Asset.create( + project_id=library.project_id, + library_id=library.id, name=name, - text=job.input_text, - voice_provider="cosyvoice", - voice_id=job.voice_id, - voice_name=job.voice_model or "", - audio_url=job.output_audio_url, - duration=job.duration, - file_size=job.file_size, - status="completed", - project_id=job.project_id or "", - tags=[], - metadata_=metadata_, + storage_key=storage_key, + mime_type=content_type, + metadata=metadata_, + file_size=file_size, + duration=job.duration or None, + status=AssetStatus.READY, + classification_status=ClassificationStatus.PENDING, # 音频不参与内容分类，保持 pending 与 ingest 链路一致 + uploaded_by_user_id=user_id, ) - - use_case = CreateVoiceLibraryUseCase(voice_library_repository) - try: - item = use_case.execute(command, plan_name=plan_name or "free") - except QuotaExceededError as exc: - raise HTTPException( - status_code=status.HTTP_429_TOO_MANY_REQUESTS, - detail=f"配音库配额已满（{exc.used}/{exc.limit}），请升级套餐", - ) from exc + asset = asset_repository.create(asset) return SaveToLibraryResponse( - id=item.id, - name=item.name, - audio_url=sign_url(item.audio_url) if item.audio_url else "", - duration=item.duration, - voice_id=item.voice_id, - voice_name=item.voice_name, - status=item.status, + id=asset.id, + name=asset.name, + audio_url=sign_url(storage_key), + duration=asset.duration or 0.0, + voice_id=job.voice_id, + voice_name=job.voice_model or "", + status="completed", ) - @router.post("/preview", response_model=TTSPreviewResponse) def preview_tts( request: TTSPreviewRequest, diff --git a/apps/api/app/api/routes/voice_clones.py b/apps/api/app/api/routes/voice_clones.py index cc513b615..c97819a54 100755 --- a/apps/api/app/api/routes/voice_clones.py +++ b/apps/api/app/api/routes/voice_clones.py @@ -7,7 +7,13 @@ from typing import Optional from app.auth import AuthenticatedUser, get_current_user from app.core.celery_app import celery_app -from app.dependencies import get_cosyvoice_service, get_voice_clone_profile_repository +from app.core.storage import get_storage_service +from app.dependencies import ( + get_asset_repository, + get_cosyvoice_service, + get_project_repository, + get_voice_clone_profile_repository, +) from app.schemas.voice_clone import ( CreateVoiceCloneRequest, ListVoiceCloneResponse, @@ -32,6 +38,9 @@ from packages.application.voice_clone.use_cases import ( from packages.application.voice_clone.workflow import ( VoiceCloneWorkflowService, ) +from packages.ports.asset_repository import AssetRepository +from packages.ports.project_repository import ProjectRepository +from packages.shared.storage import SharedStorageService logger = logging.getLogger(__name__) @@ -83,23 +92,69 @@ def create_voice_clone( request: CreateVoiceCloneRequest, authenticated_user: AuthenticatedUser = Depends(get_current_user), workflow: VoiceCloneWorkflowService = Depends(_get_workflow_service), + asset_repository: AssetRepository = Depends(get_asset_repository), + project_repository: ProjectRepository = Depends(get_project_repository), + storage_service: SharedStorageService = Depends(get_storage_service), ) -> VoiceCloneProfileResponse: """创建音色克隆任务。 创建 VoiceCloneProfile → 提交 CosyVoice 克隆任务 → 触发 Celery 异步轮询。 - 如果有 source_audio_url，状态会变为 processing；否则保持 pending。 + 参考音频两种来源（二选一）： + - source_audio_url：前端直传后的音频 URL（兼容旧流程） + - asset_id：配音素材库中的音频素材，服务端用其 OSS storage_key 生成 + 预签名下载 URL（不依赖前端签名，避免签名过期导致克隆失败） + 如果有参考音频，状态会变为 processing；否则保持 pending。 """ user_id = authenticated_user.user.id + + source_audio_url = request.source_audio_url + clone_metadata = dict(request.metadata_ or {}) + + if request.asset_id: + if source_audio_url: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="asset_id 与 source_audio_url 只能传一个", + ) + asset = asset_repository.find_by_id(request.asset_id) + if asset is None: + raise HTTPException( + status_code=status.HTTP_404_NOT_FOUND, + detail="素材不存在", + ) + # 归属校验：素材挂在项目素材库下，用户必须能访问该项目 + project = project_repository.find_by_id(asset.project_id) + if project is None or not project.can_access(user_id): + raise HTTPException( + status_code=status.HTTP_403_FORBIDDEN, + detail="无权使用该素材", + ) + # 类型校验：仅支持音频素材 + file_type = asset.file_type if hasattr(asset, "file_type") else "" + if file_type != "audio": + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="仅支持音频素材进行音色克隆", + ) + if not asset.storage_key: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail="该素材缺少音频文件，无法用于克隆", + ) + # 用 OSS storage_key 生成服务端预签名 URL（7 天有效，覆盖克隆重试周期） + source_audio_url = storage_service.get_download_url(asset.storage_key, expires_seconds=7 * 24 * 3600) + clone_metadata["source_asset_id"] = asset.id + profile = workflow.start_clone( user_id=user_id, name=request.name, description=request.description, - source_audio_url=request.source_audio_url, + source_audio_url=source_audio_url, voice_model=request.voice_model, language=request.language, gender=request.gender, max_retries=request.max_retries, - metadata=request.metadata_, + metadata=clone_metadata, ) # 如果 profile 处于 processing 且有 task_id，触发 Celery 异步轮询 diff --git a/apps/api/app/schemas/voice_clone.py b/apps/api/app/schemas/voice_clone.py index b13517366..261fbfaf2 100755 --- a/apps/api/app/schemas/voice_clone.py +++ b/apps/api/app/schemas/voice_clone.py @@ -13,7 +13,8 @@ class CreateVoiceCloneRequest(BaseModel): name: str = Field(..., min_length=1, max_length=100, description="音色名称") description: str = Field("", description="音色描述") - source_audio_url: str = Field("", description="参考音频 URL") + source_audio_url: str = Field("", description="参考音频 URL（与 asset_id 二选一）") + asset_id: str = Field("", description="参考音频素材 ID（配音素材库中的音频 asset，与 source_audio_url 二选一）") voice_model: str = Field("", description="语音模型名称") language: str = Field("zh-CN", description="语言") gender: str = Field("unknown", description="性别") diff --git a/tests/integration/test_tts_api.py b/tests/integration/test_tts_api.py index 63e065b0c..3a1a85216 100644 --- a/tests/integration/test_tts_api.py +++ b/tests/integration/test_tts_api.py @@ -31,8 +31,12 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "..", "apps", " from app.api.routes.tts import router from app.auth import AuthenticatedUser, get_current_user +from app.core.storage import get_storage_service from app.dependencies import ( + get_asset_library_repository, + get_asset_repository, get_cosyvoice_service, + get_project_repository, get_user_repository, get_voice_clone_profile_repository, get_voice_library_repository, @@ -322,6 +326,92 @@ def _make_voice_clone_profile( return profile +# --------------------------------------------------------------------------- +# 3b. 素材体系内存 Repository（save-to-library 新链路） +# --------------------------------------------------------------------------- + + +class InMemoryAssetRepository: + """内存 Asset 仓储（save-to-library 只用到 create/find_by_id）。""" + + def __init__(self): + self._items: dict = {} + + def create(self, asset): + self._items[asset.id] = asset + return asset + + def find_by_id(self, asset_id: str): + return self._items.get(asset_id) + + +class InMemoryAssetLibraryRepository: + """内存 AssetLibrary 仓储。""" + + def __init__(self): + self._items: dict = {} + + def create(self, library): + self._items[library.id] = library + return library + + def get(self, library_id: str): + return self._items.get(library_id) + + def find_by_id(self, library_id: str): + return self._items.get(library_id) + + def find_by_project(self, project_id: str): + return [lib for lib in self._items.values() if lib.project_id == project_id] + + +class InMemoryProjectRepository2: + """内存 Project 仓储（save-to-library 需要）。""" + + def __init__(self): + self._items: dict = {} + + def find_by_id(self, project_id: str): + return self._items.get(project_id) + + def find_accessible_projects(self, user_id: str): + from packages.domain.entities import Project + + projects = [p for p in self._items.values() if p.can_access(user_id)] + if projects: + return projects + # 没有任何项目时自动给一个默认项目（与前端 getOrCreateDefaultProject 行为对齐） + project = Project.create(owner_user_id=user_id, name="默认项目") + self._items[project.id] = project + return [project] + + +class MockStorageService: + """Mock 存储：下载写出小文件，上传不做真实 OSS 操作。""" + + def __init__(self): + self.uploaded_keys: list[str] = [] + self.download_calls: list[str] = [] + + def download_asset(self, storage_key_or_url: str, local_path) -> bool: + self.download_calls.append(storage_key_or_url) + from pathlib import Path + + path = Path(local_path) + path.write_bytes(b"FAKE-AUDIO-BYTES" * 100) + return True + + def upload_file(self, file_or_path, storage_key: str, content_type: str = "application/octet-stream") -> str: + self.uploaded_keys.append(storage_key) + return f"https://cdn.example.com/{storage_key}" + + def get_download_url(self, storage_key: str, expires_seconds: int = 3600) -> str: + return f"https://cdn.example.com/{storage_key}?signed=1" + + def get_url(self, storage_key: str) -> str: + return f"https://cdn.example.com/{storage_key}" + + # --------------------------------------------------------------------------- # 4. Fixtures # --------------------------------------------------------------------------- @@ -342,6 +432,26 @@ def voice_library_repo(): return InMemoryVoiceLibraryRepository() +@pytest.fixture +def asset_repo(): + return InMemoryAssetRepository() + + +@pytest.fixture +def asset_library_repo(): + return InMemoryAssetLibraryRepository() + + +@pytest.fixture +def project_repo2(): + return InMemoryProjectRepository2() + + +@pytest.fixture +def storage_service(): + return MockStorageService() + + @pytest.fixture def user_repo(): repo = InMemoryUserRepository() @@ -355,7 +465,17 @@ def cosyvoice_service(): @pytest.fixture -def client(tts_repo, voice_clone_repo, voice_library_repo, user_repo, cosyvoice_service): +def client( + tts_repo, + voice_clone_repo, + voice_library_repo, + user_repo, + cosyvoice_service, + asset_repo, + asset_library_repo, + project_repo2, + storage_service, +): """创建带有依赖覆盖的 TestClient。""" test_app = FastAPI() test_app.include_router(router, prefix="/tts") @@ -371,6 +491,10 @@ def client(tts_repo, voice_clone_repo, voice_library_repo, user_repo, cosyvoice_ test_app.dependency_overrides[get_voice_clone_profile_repository] = lambda: voice_clone_repo test_app.dependency_overrides[get_voice_library_repository] = lambda: voice_library_repo test_app.dependency_overrides[get_user_repository] = lambda: user_repo + test_app.dependency_overrides[get_asset_repository] = lambda: asset_repo + test_app.dependency_overrides[get_asset_library_repository] = lambda: asset_library_repo + test_app.dependency_overrides[get_project_repository] = lambda: project_repo2 + test_app.dependency_overrides[get_storage_service] = lambda: storage_service # 使用 FastAPI dependency_overrides 覆盖 TTS repository from app.api.routes import tts as tts_module @@ -493,6 +617,65 @@ class TestCreateTTSJob: ) assert resp.status_code == 403 + def test_synthesize_with_clone_profile_in_voice_id(self, client, tts_repo, voice_clone_repo): + """voice_id 直接传克隆 profile UUID（新前端流程）： + + 命中 profile → 校验归属 → job.voice_id 存解析后的 CosyVoice voice_id， + voice_clone_profile_id 记录该 profile。 + """ + profile = _make_voice_clone_profile() # ready，voice_id="clone-voice-001" + voice_clone_repo.create(profile) + + resp = client.post( + "/tts/synthesize", + json={"text": "克隆音色合成", "voice_id": profile.id}, + ) + assert resp.status_code == 201, resp.text + job = tts_repo.get(resp.json()["job_id"]) + assert job.voice_id == "clone-voice-001" + assert job.voice_clone_profile_id == profile.id + + def test_synthesize_with_other_user_clone_voice_id_returns_403(self, client, voice_clone_repo): + """voice_id 传他人克隆 profile UUID → 403。""" + profile = _make_voice_clone_profile(user_id="other-user") + voice_clone_repo.create(profile) + + resp = client.post( + "/tts/synthesize", + json={"text": "越权克隆音色", "voice_id": profile.id}, + ) + assert resp.status_code == 403, resp.text + + def test_synthesize_with_unfinished_clone_voice_id_returns_400(self, client, voice_clone_repo): + """voice_id 传克隆未完成（voice_id 为空）的 profile → 400。""" + from packages.domain.voice_clone_profile import VoiceCloneProfile + + profile = VoiceCloneProfile.create( + user_id="user-test-001", + name="未完成克隆", + voice_model="cosyvoice-v2", + ) + profile.mark_processing() # processing 状态，尚未 mark_ready，voice_id 为空 + voice_clone_repo.create(profile) + + resp = client.post( + "/tts/synthesize", + json={"text": "未完成克隆", "voice_id": profile.id}, + ) + assert resp.status_code == 400, resp.text + assert "克隆尚未完成" in resp.json()["detail"] + + def test_synthesize_with_preset_voice_id_unaffected(self, client, tts_repo): + """预置音色 voice_id 不匹配任何 profile 时走原流程，不受影响。""" + resp = client.post( + "/tts/synthesize", + json={"text": "预置音色", "voice_id": "longxiaochun_v2"}, + ) + assert resp.status_code == 201, resp.text + job = tts_repo.get(resp.json()["job_id"]) + assert job.voice_id == "longxiaochun_v2" + assert job.voice_clone_profile_id == "" + # --------------------------------------------------------------------------- # 6. GET /jobs — 列出 TTS 任务 @@ -805,18 +988,30 @@ class TestSaveToLibrary: # 自动生成的名称应该以 TTS- 开头 assert data["name"].startswith("TTS-") - def test_save_creates_library_item(self, client, tts_repo, voice_library_repo): - """保存后配音库中新增一条记录。""" - before_count = voice_library_repo.count_by_user("user-test-001") - - job = _make_tts_job("入库测试", status=TTSJobStatus.COMPLETED) + def test_save_creates_library_item(self, client, tts_repo, asset_repo, asset_library_repo): + """保存后在素材体系（assets 表）中新增一条 ready 音频素材，并自动创建 voice 库。""" + job = _make_tts_job("入库测试", status=TTSJobStatus.COMPLETED, duration=6.0) tts_repo.create(job) resp = client.post(f"/tts/jobs/{job.id}/save-to-library", json={"name": "入库"}) - assert resp.status_code == 201 + assert resp.status_code == 201, resp.text + data = resp.json() - after_count = voice_library_repo.count_by_user("user-test-001") - assert after_count == before_count + 1 + # asset 已创建 + asset = asset_repo.find_by_id(data["id"]) + assert asset is not None + assert asset.status.value == "ready" + assert asset.file_type == "audio" + assert asset.duration == 6.0 + assert (asset.metadata or {}).get("source") == "tts_job" + assert (asset.metadata or {}).get("tts_job_id") == job.id + assert asset.uploaded_by_user_id == "user-test-001" + assert asset.storage_key.startswith("uploads/voice/tts/") + + # voice 素材库自动创建，asset 挂到该库 + assert asset.library_id in asset_library_repo._items + voice_lib = asset_library_repo.get(asset.library_id) + assert voice_lib is not None # --------------------------------------------------------------------------- diff --git a/tests/integration/test_voice_clone_api.py b/tests/integration/test_voice_clone_api.py index a63d402c1..def454b87 100755 --- a/tests/integration/test_voice_clone_api.py +++ b/tests/integration/test_voice_clone_api.py @@ -29,15 +29,20 @@ from fastapi.testclient import TestClient sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "..", "apps", "api")) +from unittest.mock import MagicMock + from app.api.routes.voice_clones import router from app.auth import AuthenticatedUser, get_current_user +from app.core.storage import get_storage_service from app.dependencies import ( + get_asset_repository, get_audio_url_signer, get_cosyvoice_service, + get_project_repository, get_voice_clone_profile_repository, ) -from packages.domain.entities import User +from packages.domain.entities import Asset, AssetStatus, Project, User from packages.domain.voice_clone_profile import ( VoiceCloneProfile, VoiceCloneStatus, @@ -232,8 +237,78 @@ def cosyvoice_service(): return MockCosyVoiceService(async_mode=True) # 异步模式，匹配真实 CosyVoice API 行为 +def _make_voice_asset( + *, + asset_id: str = "asset-voice-1", + project_id: str = "proj-1", + user_id: str = "user-test-001", + file_type: str = "audio", + storage_key: str = "uploads/voice/sample.m4a", +) -> Asset: + """构造一个音频素材（可改 file_type 模拟非音频）。""" + mime = {"audio": "audio/mp4", "video": "video/mp4", "image": "image/jpeg"}[file_type] + return Asset( + id=asset_id, + project_id=project_id, + library_id="lib-1", + name="配音素材.m4a", + storage_key=storage_key, + mime_type=mime, + file_size=12345, + duration=10.0, + status=AssetStatus.READY, + uploaded_by_user_id=user_id, + ) + + +class InMemoryAssetRepository: + """最小内存 Asset 仓储（find_by_id 即可）。""" + + def __init__(self, assets: list[Asset] | None = None): + self._items = {a.id: a for a in (assets or [])} + + def find_by_id(self, asset_id: str): + return self._items.get(asset_id) + + +class InMemoryProjectRepository: + """最小内存 Project 仓储（find_by_id 即可）。""" + + def __init__(self, projects: list[Project] | None = None): + self._items = {p.id: p for p in (projects or [])} + + def find_by_id(self, project_id: str): + return self._items.get(project_id) + + +class MockStorageService: + """Mock 存储服务：get_download_url 返回固定签名 URL。""" + + def __init__(self): + self.signed_keys: list[str] = [] + + def get_download_url(self, storage_key: str, expires_seconds: int = 3600) -> str: + self.signed_keys.append(storage_key) + return f"https://oss.example.com/{storage_key}?signed=1&exp={expires_seconds}" + + @pytest.fixture -def client(clone_repo, cosyvoice_service): +def asset_repo(): + return InMemoryAssetRepository() + + +@pytest.fixture +def project_repo(): + return InMemoryProjectRepository() + + +@pytest.fixture +def storage_service(): + return MockStorageService() + + +@pytest.fixture +def client(clone_repo, cosyvoice_service, asset_repo, project_repo, storage_service): """创建带有依赖覆盖的 TestClient。""" test_app = FastAPI() test_app.include_router(router, prefix="/voice-clones") @@ -245,6 +320,9 @@ def client(clone_repo, cosyvoice_service): test_app.dependency_overrides[get_voice_clone_profile_repository] = lambda: clone_repo test_app.dependency_overrides[get_cosyvoice_service] = lambda: cosyvoice_service test_app.dependency_overrides[get_audio_url_signer] = lambda: (lambda url: url) + test_app.dependency_overrides[get_asset_repository] = lambda: asset_repo + test_app.dependency_overrides[get_project_repository] = lambda: project_repo + test_app.dependency_overrides[get_storage_service] = lambda: storage_service yield TestClient(test_app) @@ -714,3 +792,77 @@ class TestVoiceCloneLifecycle: if __name__ == "__main__": pytest.main([__file__, "-v"]) + + +# --------------------------------------------------------------------------- +# 6. POST /?asset_id=xxx — 从配音素材创建克隆 +# --------------------------------------------------------------------------- + + +class TestCreateVoiceCloneFromAsset: + """asset_id 参数测试：用配音素材库音频发起克隆。""" + + def test_create_from_asset_success(self, client, clone_repo, asset_repo, project_repo, storage_service): + """正常路径：asset 归属本人且为音频 → 201，source_audio_url 为签名 URL。""" + project = Project.create(owner_user_id="user-test-001", name="默认项目") + project_repo._items[project.id] = project + asset = _make_voice_asset(asset_id="asset-ok", project_id=project.id) + asset_repo._items[asset.id] = asset + + resp = client.post( + "/voice-clones", + json={"name": "素材克隆", "asset_id": asset.id}, + ) + assert resp.status_code == 201, resp.text + data = resp.json() + assert data["name"] == "素材克隆" + # 用 OSS storage_key 生成了服务端签名 URL + assert any("uploads/voice/sample.m4a" in k for k in storage_service.signed_keys) + assert data["source_audio_url"].startswith("https://oss.example.com/uploads/voice/sample.m4a") + + def test_create_from_asset_not_found_returns_404(self, client, asset_repo, project_repo): + """asset 不存在 → 404。""" + resp = client.post("/voice-clones", json={"name": "克隆", "asset_id": "no-such-asset"}) + assert resp.status_code == 404, resp.text + + def test_create_from_asset_other_user_returns_403(self, client, asset_repo, project_repo): + """asset 属于他人项目 → 403。""" + project = Project.create(owner_user_id="other-user", name="他人项目") + project_repo._items[project.id] = project + asset = _make_voice_asset(asset_id="asset-other", project_id=project.id, user_id="other-user") + asset_repo._items[asset.id] = asset + + resp = client.post("/voice-clones", json={"name": "克隆", "asset_id": asset.id}) + assert resp.status_code == 403, resp.text + + def test_create_from_non_audio_asset_returns_400(self, client, asset_repo, project_repo): + """asset 是视频 → 400。""" + project = Project.create(owner_user_id="user-test-001", name="默认项目") + project_repo._items[project.id] = project + asset = _make_voice_asset( + asset_id="asset-video", + project_id=project.id, + file_type="video", + storage_key="uploads/v/test.mp4", + ) + asset_repo._items[asset.id] = asset + + resp = client.post("/voice-clones", json={"name": "克隆", "asset_id": asset.id}) + assert resp.status_code == 400, resp.text + + def test_create_with_both_asset_id_and_url_returns_400(self, client, asset_repo, project_repo): + """asset_id 与 source_audio_url 同时传 → 400。""" + project = Project.create(owner_user_id="user-test-001", name="默认项目") + project_repo._items[project.id] = project + asset = _make_voice_asset(asset_id="asset-both", project_id=project.id) + asset_repo._items[asset.id] = asset + + resp = client.post( + "/voice-clones", + json={ + "name": "克隆", + "asset_id": asset.id, + "source_audio_url": "https://example.com/a.wav", + }, + ) + assert resp.status_code == 400, resp.text