\n\n This program is free software: you can redistribute it and/or modify\n it under the terms of the GNU Affero General Public License as published\n by the Free Software Foundation, either version 3 of the License, or\n (at your option) any later version.\n\n This program is distributed in the hope that it will be useful,\n but WITHOUT ANY WARRANTY; without even the implied warranty of\n MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\n GNU Affero General Public License for more details.\n\n You should have received a copy of the GNU Affero General Public License\n along with this program. If not, see .\n\nAlso add information on how to contact you by electronic and paper mail.\n\n If your software can interact with users remotely through a computer\nnetwork, you should also make sure that it provides a way for users to\nget its source. For example, if your program is a web application, its\ninterface could display a \"Source\" link that leads users to an archive\nof the code. There are many ways you could offer source, and different\nsolutions will be better for different programs; see section 13 for the\nspecific requirements.\n\n You should also get your employer (if you work as a programmer) or school,\nif any, to sign a \"copyright disclaimer\" for the program, if necessary.\nFor more information on this, and how to apply and follow the GNU AGPL, see\n.\n"
},
{
"path": "README.md",
"content": "\n\n# 音谷 - AI 多角色多情绪配音平台\n\n
\n\n\n\n\n ![](\"https://img.shields.io/badge/飞书-使用教程-4285F4?logo=googleclassroom&logoColor=white\")
\n\n![](\"https://img.shields.io/badge/license-AGPLv3-blue?logo=gnu\")
\n![](\"https://img.shields.io/badge/release-v1.1.3-brightgreen?logo=semver\")
\n
\n\n> 一个开源的多角色、多情绪 AI 配音生成平台，支持小说、剧本、视频等内容的自动配音与导出。 \n\n---\n## 📝 详细使用文档\n[音谷 - AI 多角色多情绪配音平台使用教程](https://sw4s2hg7k5y.feishu.cn/wiki/WjbUw1t7JiWIa7k2pFXcxqSbnde?from=from_copylink)\n## 📖 软件简介\n- **软件名称**：音谷 - AI 多角色多情绪配音平台 \n- **定位**：为小说、剧本、视频等内容提供多角色、多情绪的 AI 语音合成与配音服务 \n- **主要功能**：\n - 小说 / 剧本文本导入\n - 多角色角色库管理\n - 情绪音色选择与绑定\n - 台词自动拆分与配音生成\n - 批量任务管理与导出\n - 支持自定义 LLM 接口选择与调用\n - 基于Index-TTS-2.0的多情绪TTS服务\n - 支持精准的音频编辑功能，可以自定义删除音频片段或者添加静音片段。\n - 支持自定义提示词，适配个性化拆分需求\n## 🛠 技术栈\n- **前端**：Electron + Vue + Element Plus \n- **后端**：FastAPI / Python\n- **AI 接口**：兼容 OpenAI API 协议的大模型 \n- **TTS 服务**：IndexTTs-2 + Cloud Native Build 平台（免费 H20 显卡支持）/ 本地部署整合包\n\n## 二次开发说明\n本软件依据 **AGPL-3.0** 开源许可协议发布。基于本项目进行二次开发时，开发者须遵守以下规范：\n### 1. 署名要求\n必须在衍生软件的用户界面及代码文档中清晰标注：\n> \"本软件基于开源项目《音谷》二次开发\"\n\n并附上原项目仓库链接。\n### 2. 商业使用限制\n未获得书面商业授权前，任何基于本项目的衍生作品不得用于商业用途或提供商业服务。\n\n## 🚀 快速开始\n\n### 1️⃣ 克隆项目\n```bash\ngit clone https://github.com/xcLee001/SonicVale.git\ncd SonicVale\n```\n### 2️⃣ 启动后端\n首先，需要下载ffmpeg.exe到app/core/ffmpeg/ffmpeg.exe\n\n\n可以去官网[ffmpeg](https://www.gyan.dev/ffmpeg/builds/packages/ffmpeg-8.0-full_build.7z)\n。也可以使用[此镜像](https://www.alipan.com/s/ey5QRqW3Jji)\n\n然后复制到app/core/ffmpeg/目录下\n\n安装依赖和启动服务\n```bash\ncd SonicVale\npip install -r requirements.txt\nuvicorn app.main:app --reload --port 8200\n```\n\n\n```\napp/\n├── core/ # 全局配置、tts引擎、llm引擎、ffmpeg封装、字幕生成、websocket、异步队列\n├── db/ # 数据库连接和Base\n├── models/ # ORM模型\n├── dto/ # 数据传输对象（请求/响应验证）\n├── entity/ # 实体类（结合 ORM 与业务层）\n├── repositories/ # 数据库封装\n├── services/ # 核心业务逻辑\n├── routers/ # FastAPI路由接口\n└── main.py # FastAPI启动入口\n\n```\n\n\n\n### 3️⃣ 启动前端\n```bash\ncd sonicvale-front\nnpm install # 安装依赖\nnpm run start # 启动前端包括electron\n```\n## Coffe\n如果您觉得我的项目对您有所帮助，欢迎您的赞助。您的支持将使我有更多的动力继续维护和改进这个项目。\n您可以通过扫描下面的二维码来请我喝杯咖啡：\n\n\n![\"赞赏码\"](\"image/赞赏码.jpg\")
\n\n## 🎥 效果演示\n👉 [点击查看 B 站演示效果视频](https://www.bilibili.com/video/BV1tSpTz6EBH/)\n\n\n## 📷 截图\n\nLLM 配置界面\n\n\nTTS 配置界面\n\n\n音色管理界面\n\n\n项目创建界面\n\n\n章节创建界面\n\n\n章节内容导入\n\n\n台词自动拆分\n\n\n角色绑定，多章节共享角色音色\n\n\n台词编辑，高度自定义\n\n - 在台词编辑区，用户可手动修改：\n - 台词文本\n - 角色归属\n - 情绪类型\n - 情绪轻度\n - 修改后自动保存并更新。\n\n配音生成\n\n\n生成后音频可编辑\n\n\n\n\n## 📬 联系方式\n\n如果在使用过程中遇到 **Bug** 或者有 **功能建议**，请通过 [GitHub Issues](https://github.com/xcLee001/SonicVale/issues) 提交，这样可以帮助我们更好地跟踪与解决问题。 \n\n如果你希望加入用户交流社区，欢迎加入我们的 QQ 群： \n\n- 💬 QQ交流群：1060711739（1群已满）、575715633（2群） （验证信息请填写 “音谷配音”）\n\n## 📜 协议\n\n本项目采用 [GNU Affero General Public License v3.0 (AGPL-3.0)](./LICENSE) 开源协议。 \n\n您可以自由地使用、复制、修改、合并、发布和分发本软件及其副本，但必须遵守以下条款：\n\n- 您必须在分发的软件中包含原始许可声明和版权声明。\n\n- 若您修改并发布本软件，或通过网络提供服务（如 SaaS、Web 应用），您必须同时公开修改后的源代码。\n\n- 您不得附加任何与 AGPL-3.0 条款冲突的限制。\n\n## ⚠️ 免责声明\n\n本项目仅供学习与研究使用。 \n用户不得利用本项目从事任何违法违规行为，包括但不限于： \n- 克隆或模仿未经授权的声音； \n- 侵犯他人声音权、肖像权、著作权、名誉权； \n- 其他可能违反法律法规的行为。 \n\n开发者不对用户使用本项目所产生的任何后果负责，所有风险与责任由用户自行承担。 \n使用本项目即表示您已阅读并同意本免责声明。 \n\n---\n\n## ⚠️ Disclaimer\n\nThis project is intended for research and educational purposes only. \nUsers are strictly prohibited from using this project for any unlawful activities, including but not limited to: \n- Cloning or imitating voices without authorization; \n- Infringing upon the rights of others (voice rights, portrait rights, copyrights, reputation rights, etc.); \n- Any other activities in violation of applicable laws and regulations. \n\nThe developer shall not be held liable for any consequences arising from the use of this project. \nAll risks and responsibilities lie solely with the user. \nBy using this project, you acknowledge that you have read and agreed to this disclaimer.\n"
},
{
"path": "SonicVale/.gitignore",
"content": "# python cache\n__pycache__/\n*.pyc\n*.pyo\n*.pyd\n\n# JetBrains IDE\n.idea/\n\n# venv\n.venv/\nenv/\nvenv/\n\n\n# 打包输出\ndist\nbuild\n*.spec\n*.exe\n\n# logs\n*.log\n\n# others\n.DS_Store"
},
{
"path": "SonicVale/README.md",
"content": "```\napp/\n├── core/ # 全局配置、tts引擎、llm引擎、ffmpeg封装、字幕生成、websocket、异步队列\n├── db/ # 数据库连接和Base\n├── models/ # ORM模型\n├── dto/ # 数据传输对象（请求/响应验证）\n├── entity/ # 实体类（结合 ORM 与业务层）\n├── repositories/ # 数据库封装\n├── services/ # 核心业务逻辑\n├── routers/ # FastAPI路由接口\n└── main.py # FastAPI启动入口\n\n\n```\n"
},
{
"path": "SonicVale/app/__init__.py",
"content": ""
},
{
"path": "SonicVale/app/core/__init__.py",
"content": ""
},
{
"path": "SonicVale/app/core/audio_engin.py",
"content": "import os\nimport subprocess\nimport tempfile\nimport soundfile as sf\nimport numpy as np\n\nfrom app.core.config import getFfmpegPath\n\n\nclass AudioProcessor:\n def __init__(self, audio_path: str, keep_format=True, default_sr=44100, default_ch=2):\n self.audio_path = audio_path\n self.keep_format = keep_format\n self.default_sr = default_sr\n self.default_ch = default_ch\n\n info = sf.info(audio_path)\n self.sr = info.samplerate if keep_format else default_sr\n self.ch = info.channels if keep_format else default_ch\n self.duration = info.duration\n\n self.ffmpeg_path = getFfmpegPath()\n self.temp_path = self._create_tmp_file()\n\n def _create_tmp_file(self):\n os.makedirs(os.path.dirname(self.audio_path) or \".\", exist_ok=True)\n tmp = tempfile.NamedTemporaryFile(delete=False, suffix=\".wav\",\n dir=os.path.dirname(self.audio_path) or \".\")\n return tmp.name\n\n def _run_ffmpeg(self, cmd):\n subprocess.run(\n cmd, check=True,\n creationflags=subprocess.CREATE_NO_WINDOW if os.name == \"nt\" else 0\n )\n\n def _normalize(self, path):\n \"\"\"防止音量削波\"\"\"\n data, sr = sf.read(path, dtype=\"float32\", always_2d=True)\n peak = float(np.max(np.abs(data)))\n if peak > 1.0:\n data = data / peak\n sf.write(path, data, sr, format=\"WAV\", subtype=\"PCM_16\")\n\n # ---------------------- 模块功能 ---------------------- #\n\n def cut(self, start_ms: int, end_ms: int):\n \"\"\"删除音频区间 [start_ms, end_ms]\"\"\"\n start_sec = start_ms / 1000\n end_sec = end_ms / 1000\n\n cmd = [\n self.ffmpeg_path, \"-y\", \"-i\", self.audio_path,\n \"-filter_complex\",\n f\"[0:a]atrim=0:{start_sec},asetpts=PTS-STARTPTS[first];\"\n f\"[0:a]atrim={end_sec},asetpts=PTS-STARTPTS[second];\"\n f\"[first][second]concat=n=2:v=0:a=1[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n self._run_ffmpeg(cmd)\n os.replace(self.temp_path, self.audio_path)\n\n def insert_silence(self, insert_ms: int, duration_sec: float):\n \"\"\"在指定时间点插入静音\"\"\"\n insert_sec = insert_ms / 1000\n cmd = [\n self.ffmpeg_path, \"-y\",\n \"-i\", self.audio_path,\n \"-f\", \"lavfi\", \"-t\", str(duration_sec),\n \"-i\", f\"anullsrc=channel_layout={'stereo' if self.ch == 2 else 'mono'}:sample_rate={self.sr}\",\n \"-filter_complex\",\n f\"[0:a]atrim=0:{insert_sec},asetpts=PTS-STARTPTS[first];\"\n f\"[0:a]atrim={insert_sec},asetpts=PTS-STARTPTS[second];\"\n f\"[first][1:a][second]concat=n=3:v=0:a=1[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n self._run_ffmpeg(cmd)\n os.replace(self.temp_path, self.audio_path)\n\n def append_silence(self, duration_sec: float):\n \"\"\"\n 在音频末尾添加或裁剪静音段：\n - duration_sec > 0: 在末尾添加指定秒数静音\n - duration_sec < 0: 从末尾裁剪指定秒数的内容\n \"\"\"\n if duration_sec == 0:\n return # 无需处理\n\n # ---------- 情况1：添加静音 ----------\n if duration_sec > 0:\n cmd = [\n self.ffmpeg_path, \"-y\",\n \"-i\", self.audio_path,\n \"-f\", \"lavfi\", \"-t\", str(duration_sec),\n \"-i\", f\"anullsrc=channel_layout={'stereo' if self.ch == 2 else 'mono'}:sample_rate={self.sr}\",\n \"-filter_complex\",\n \"[0:a][1:a]concat=n=2:v=0:a=1[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n\n # ---------- 情况2：裁剪末尾 ----------\n else:\n cut_dur = self.duration + duration_sec # 因为 duration_sec 为负\n if cut_dur < 0:\n cut_dur = 0 # 防止全裁掉出错\n cmd = [\n self.ffmpeg_path, \"-y\",\n \"-i\", self.audio_path,\n \"-filter_complex\",\n f\"[0:a]atrim=0:{cut_dur},asetpts=PTS-STARTPTS[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n\n # 执行 ffmpeg 命令\n self._run_ffmpeg(cmd)\n os.replace(self.temp_path, self.audio_path)\n # 更新音频时长（防止后续操作出错）\n info = sf.info(self.audio_path)\n self.duration = info.duration\n\n def change_speed(self, speed: float):\n \"\"\"变速处理 (0.5~2.0倍)\"\"\"\n speed = float(np.clip(speed, 0.5, 2.0))\n cmd = [\n self.ffmpeg_path, \"-y\", \"-i\", self.audio_path,\n \"-af\", f\"atempo={speed}\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n self._run_ffmpeg(cmd)\n os.replace(self.temp_path, self.audio_path)\n\n def change_volume(self, volume: float):\n \"\"\"音量调整\"\"\"\n volume = max(0.0, float(volume))\n cmd = [\n self.ffmpeg_path, \"-y\", \"-i\", self.audio_path,\n \"-af\", f\"volume={volume}\",\n \"-ar\", str(self.sr),\n \"-ac\", str(self.ch),\n \"-c:a\", \"pcm_s16le\",\n self.temp_path\n ]\n self._run_ffmpeg(cmd)\n os.replace(self.temp_path, self.audio_path)\n\n def export(self, out_path: str):\n \"\"\"导出音频到目标路径（带软限幅）\"\"\"\n self._normalize(self.audio_path)\n os.replace(self.audio_path, out_path)\n return out_path\n"
},
{
"path": "SonicVale/app/core/config.py",
"content": "import os\n\nimport os, sys\nfrom pathlib import Path\n# 得到默认配置文件\ndef getConfigPath():\n # 用户 目录下SonicVale目录\n user_dir = os.path.join(os.path.expanduser(\"~\"), \"SonicVale\")\n\n # 如果目录不存在，创建它\n if not os.path.exists(user_dir):\n os.makedirs(user_dir, exist_ok=True)\n\n # 返回 config.json 路径（目录已保证存在）\n return user_dir\n\ndef getFfmpegPath():\n BASE_DIR = getattr(sys, \"_MEIPASS\", Path(os.path.abspath(\".\")))\n FFMPEG_PATH = os.path.join(BASE_DIR, \"core\", \"ffmpeg\", \"ffmpeg.exe\")\n return FFMPEG_PATH"
},
{
"path": "SonicVale/app/core/enums.py",
"content": "from enum import Enum\n\nclass TaskEnum(str, Enum):\n DUBBING = \"台词拆分\"\n\n"
},
{
"path": "SonicVale/app/core/llm_engine.py",
"content": "# app/core/llm_engine.py\nimport json\nimport logging\n# app/core/llm_engine.py\n\nimport re\nimport time\nimport random\nfrom openai import OpenAI\nfrom numba.cuda import stream\n\nfrom app.core.prompts import get_auto_fix_json_prompt\n\n\nclass LLMEngine:\n def __init__(self, api_key: str, base_url: str, model_name: str, custom_params: str):\n \"\"\"\n api_key: LLM API Key\n base_url: OpenAI-compatible API URL（例如企业版/自建 LLM）\n model_name: 模型名称\n custom_params: 自定义参数（JSON字符串）\n \"\"\"\n self.api_key = api_key\n self.base_url = base_url.rstrip(\"/\") # 去掉末尾斜杠\n self.model_name = model_name\n \n # custom_params从string转为dict\n custom_params = json.loads(custom_params)\n if not isinstance(custom_params, dict):\n raise ValueError(\"无效的 custom_params\")\n self.custom_params = custom_params\n \n # 使用新版 OpenAI 客户端\n self.client = OpenAI(\n api_key=api_key,\n base_url=self.base_url\n )\n\n def _extract_result_tag(self, text: str) -> str:\n \"\"\"提取 标签内容\"\"\"\n match = re.search(r\"(.*?)\", text, re.DOTALL)\n if not match:\n raise ValueError(\"Response does not contain ... tag\")\n return match.group(1).strip()\n\n def generate_text_test(self, prompt: str) -> str:\n \"\"\"\n 测试：生成结果并返回（非流式）\n \"\"\"\n response = self.client.chat.completions.create(\n model=self.model_name,\n messages=[{\"role\": \"user\", \"content\": prompt}],\n timeout=3000,\n **self.custom_params\n )\n return response.choices[0].message.content\n def generate_text(self, prompt: str, retries: int = 3, delay: float = 1.0) -> str:\n \"\"\"\n 流式生成：边生成边输出\n \"\"\"\n for attempt in range(retries):\n try:\n # 开启流式\n # stream = self.client.chat.completions.create(\n # model=self.model_name,\n # messages=[{\"role\": \"user\", \"content\": prompt}],\n # stream=True,\n # timeout=3000,\n # **self.custom_params\n # )\n\n # 关闭流式，直接获取完整响应\n response = self.client.chat.completions.create(\n model=self.model_name,\n messages=[{\"role\": \"user\", \"content\": prompt}],\n stream=False, # 关键：设置为 False\n timeout=3000,\n **self.custom_params\n )\n\n # 直接获取完整文本\n full_text = response.choices[0].message.content\n return full_text\n\n except Exception as e:\n if attempt < retries - 1:\n sleep_time = delay * (2 ** attempt) + random.random()\n time.sleep(sleep_time)\n else:\n raise e\n def save_load_json(self, json_str: str):\n \"\"\"解析JSON，支持自动提取标签内容\"\"\"\n # 先尝试提取 标签内容\n try:\n json_str = self._extract_result_tag(json_str)\n except ValueError:\n # 没有 标签，直接使用原文本\n pass\n \n # 尝试加载json\n try:\n return json.loads(json_str)\n except json.JSONDecodeError:\n # JSON解析失败，尝试让LLM修复\n prompt = get_auto_fix_json_prompt(json_str)\n res = self.generate_text(prompt)\n # 递归调用，修复后的结果也可能包含 标签\n return self.save_load_json(res)\n\n def generate_smart_text(self, prompt: str) -> str:\n \"\"\"\n 智能文本生成（流式）\n \"\"\"\n stream = self.client.chat.completions.create(\n model=self.model_name,\n messages=[{\"role\": \"user\", \"content\": prompt}],\n stream=True,\n timeout=3000\n )\n\n # 拼接 delta.content\n full_text = \"\"\n for chunk in stream:\n if chunk.choices and len(chunk.choices) > 0:\n delta = chunk.choices[0].delta\n content = delta.content if hasattr(delta, 'content') else None\n if content:\n # print(content, end=\"\", flush=True)\n full_text += content\n\n logging.debug(\"流式生成完成\")\n return full_text\n"
},
{
"path": "SonicVale/app/core/prompts.py",
"content": "# 根据小说内容生成\n\nimport textwrap\n\n\ndef get_context2lines_prompt(possible_characters, novel_content,possible_emotions,possible_strengths) -> str:\n\n prompt = f\"\"\"\n你的任务是将给定小说内容划分为角色台词和旁白，并输出包含标签的结构化JSON结果。\n\n划分规则：\n\n台词识别:\n识别所有角色说话的内容，包括带引号、破折号、叹号等常见台词标记的文本。\n如果角色在给定角色列表中，使用该角色名；\n如果角色未在列表中出现，根据上下文合理归纳角色名。\n重要规则：相邻台词之间如果角色相同，可以适当合并，但是一段内容最多不超过150字。如果单段内容超过150字，请将内容拆分为多条。\n\n\n旁白识别:\n对叙述性、心理描写、环境描写、动作描写等非台词内容统一标记为“旁白：”。\n重要规则：相邻台词之间如果都为旁白内容，可以适当合并，但是一段内容最多不超过150字。如果单段内容超过150字，请将内容拆分为多条。\n\n情绪以及情绪强弱识别:\n根据上下文场景，识别出每条台词所对应的情绪以及情绪强度。情绪和情绪强度的内容必须来自情绪列表possible_emotions和情绪强度列表possible_strengths。\n旁白的情绪和情绪强度统一为一样的，统一为‘平静’情绪，强度为‘中等’。\n\n特殊情况处理:\n多角色对话连续出现时，每条台词对应正确角色。\n混合旁白和台词的段落可拆分为旁白和台词两条记录。\n避免重复、遗漏台词或旁白。\n\n输出格式:\n输出严格遵循包含标签的JSON数组形式\n\n示例：\n\n[\n{\"role_name\": \"张三\", \"text_content\": \"你到底在干什么！\", \"emotion_name\": \"生气\", \"strength_name\": \"强烈\"},\n{\"role_name\": \"旁白\", \"text_content\": \"此时，张三愤怒站着\", \"emotion_name\": \"平静\", \"strength_name\": \"中等\"},\n{\"role_name\": \"李四\", \"text_content\": \"这可不管我的事儿\", \"emotion_name\": \"害怕\", \"strength_name\": \"微弱\"}\n]\n\n\n注意事项:\n保持文本顺序与逻辑一致。\n不要改写原文台词或旁白内容。\n所有划分结果必须完整输出在 标签内。\n\n输入内容：\n可能包含的角色列表：\n\n{possible_characters}\n\n\n可能包含的情绪列表：\n\n{possible_emotions}\n\n\n可能包含的情绪强弱列表：\n\n{possible_strengths}\n\n\n小说原文：\n\n{novel_content}\n\n\n\n\"\"\"\n return textwrap.dedent(prompt)\n\ndef get_prompt_str():\n prompt = \"\"\"\n 你的任务是将给定小说内容划分为角色和内容，并输出为结构化JSON结果。\n 台词识别规则：\n 1. 必须完整保留原文内容，不得遗漏、删改或省略任何字句。\n 2. 提取角色对话内容喝旁白。识别所有内容，包括带引号（“”）、破折号（——）、感叹号（！）、冒号（：）等常见台词标记的文本，其余均为旁白内容。\n 3. 若角色在已知角色列表中，则直接使用该角色名；若不在列表中，则根据上下文合理判断角色身份。\n 4. 相邻台词如属同一角色，可合并为一条，但单条台词长度不得超过150字。\n 5. 若单条台词超过150字，需按语义完整性拆分为多条，每条不超过150字，并确保原文内容不缺失。\n \n 旁白识别规则：\n 1. 所有非台词的叙述性内容（包括心理活动、环境描写、动作描写、场景过渡等）均标记为“旁白”。\n 2. 必须保留原文的所有文字内容，不得遗漏、删改或省略任何字句。\n 3. 相邻的旁白内容可合并为一条，但单条长度不得超过150字。\n 4. 若单条旁白超过150字，需按语义完整性拆分为多条，每条不超过150字，确保原文内容完整呈现。\n \n 情绪与情绪强度识别规则：\n 1. 根据上下文语境、语气及场景变化，为每条台词识别情绪和情绪强度。\n 2. 情绪与强度必须严格从提供的情绪列表（possible_emotions）与强度列表（possible_strengths）中选择。\n 3. “旁白”内容的情绪与强度统一为：情绪“平静”，强度“中等”。\n 4. 情绪识别不得影响或改写原文内容，仅用于标注。\n \n 特殊情况处理：\n 1. 多角色连续对话时，确保每条台词对应正确角色，避免角色错配。\n 2. 当段落中混合出现旁白与台词时，应拆分为独立记录：旁白一条、台词一条。\n 3. 输出结果不得出现遗漏、重复、合并错误或原文缺失的情况。\n 4. 拆分、合并及情绪标注仅为结构化目的，须保证原文内容100%完整保留。\n \n 输出格式:\n 严格输出为 json数组。\n \n 示例：\n 小说原文：\n \n 一名靠前的灰衣少年似乎与石台上的少年颇为熟悉，他听得大伙的窃窃私语，不由得得意一笑，压低声音道：“牧哥可是被选拔出来参加过“灵路”的人，我们整个北灵境中，可就牧哥一人有名额，你们应该也知道参加“灵路”的都是些什么变态吧？当年我们这北灵境可是因为此事沸腾了好一阵的，从那里出来的人，最后基本全部都是被“五大院”给预定了的。”\n \n 输出：\n [\n {\"role_name\": \"旁白\", \"text_content\": \"一名靠前的灰衣少年似乎与石台上的少年颇为熟悉，他听得大伙的窃窃私语，不由得得意一笑，压低声音道\", \"emotion_name\": \"平静\", \"strength_name\": \"中等\"},\n {\"role_name\": \"灰衣少年\", \"text_content\": \"牧哥可是被选拔出来参加过“灵路”的人，我们整个北灵境中，可就牧哥一人有名额，你们应该也知道参加“灵路”的都是些什么变态吧？当年我们这北灵境可是因为此事沸腾了好一阵的，从那里出来的人，最后基本全部都是被“五大院”给预定了的。\", \"emotion_name\": \"高兴\", \"strength_name\": \"中等\"}\n ]\n \n \n 输入内容：\n 可能包含的角色列表：\n \n {possible_characters}\n \n \n 可能包含的情绪列表：\n \n {possible_emotions}\n \n \n 可能包含的情绪强弱列表：\n \n {possible_strengths}\n \n \n 小说原文：\n \n {novel_content}\n \n\n \"\"\"\n return textwrap.dedent(prompt)\n\n\n\n\ndef get_auto_fix_json_prompt(json_str: str) -> str:\n prompt = f\"\"\"\n 你将收到一段可能出错的 JSON 字符串（它可能是 LLM 生成的结果），其中可能存在以下问题：\n 多余或缺失的逗号\n 缺少引号或多余引号\n 键值格式错误\n JSON 外含无关说明文字\n 非法转义符\n 你的任务是：\n 仅输出一个严格合法、可被 json.loads 解析的 JSON。\n 保持原有数据结构和内容不变（除非必须修正格式）。\n 不要在 JSON 外输出任何解释、额外文字或注释。\n 输出必须完整输出在 标签内。\n 输入内容：\n \n {json_str}\n w\n \"\"\"\n return textwrap.dedent(prompt)\n\n\ndef get_add_smart_role_and_voice(original_text: str, role_name, voice_names):\n prompt = f\"\"\"\n 你是“角色音色匹配助手”。你的任务是：根据小说原文中的角色表现，为每个在中出现的角色匹配最符合其语气与性格的音色。\n\n 原文内容：\n \n {original_text}\n \n\n 角色列表信息：\n \n {role_name}\n \n\n 音色列表信息：\n \n {voice_names}\n \n\n 匹配规则（必须严格遵守）：\n 1. 仅根据【原文内容】判断哪些角色实际出现；未在原文中出现的角色一律忽略，不输出。\n 2. 对于每个实际出现的角色，根据原文中体现的性格特征、语气风格、情绪倾向、年龄感等信息，推断该角色适合的音色类型。\n 3. 再根据音色库中每个音色的名称或描述，为角色挑选最匹配的音色。\n 4. 若某角色最匹配的音色与其他角色重复使用是不允许的（音色数量可能不足）。\n 5. 若确实存在无法匹配的角色（例如原文完全无语气风格线索），则该角色不输出。\n 6. 不得臆造原文中不存在的角色特征或音色特征。\n 7. 最终输出必须是一个标准 JSON 数组，且数组中的每个对象必须包含：\n - \"role_name\": 角色名\n - \"voice_name\": 匹配的音色名\n\n 输出格式要求：\n - 严格输出 JSON 数组。\n - 不得输出任何解释说明、自然语言、注释或多余内容。\n\n 示例输出（格式示例）：\n [\n {{ \"role_name\": \"灰衣少年\", \"voice_name\": \"小王\" }},\n {{ \"role_name\": \"白衣少年\", \"voice_name\": \"小正\" }}\n ]\n \"\"\"\n\n return textwrap.dedent(prompt)\n\n\ndef get_subtitle_correction_prompt(original_text: str, subtitle_lines: list) -> str:\n \"\"\"\n 生成字幕矫正的prompt\n original_text: 原始正确文本\n subtitle_lines: ASR识别的字幕行列表，格式为 [{\"index\": 1, \"text\": \"...\"}]\n \"\"\"\n subtitle_json = \"\\n\".join([f' {{\"index\": {item[\"index\"]}, \"text\": \"{item[\"text\"]}\"}}' for item in subtitle_lines])\n \n prompt = f\"\"\"\n你是一个专业的字幕校对助手。你的任务是根据原文内容，修正ASR自动识别产生的字幕错误。\n\n## 任务说明\nASR（自动语音识别）生成的字幕可能存在以下问题：\n1. 同音字错误（如\"他\"与\"她\"、\"的\"与\"得\"）\n2. 近音字错误\n3. 词语分割错误\n4. 标点符号错误或缺失\n\n你需要参考原文，将每条字幕修正为正确的文本。\n\n## 重要规则\n1. 严格保持字幕条目数量不变（输入多少条，输出多少条）\n2. 尽量保持每条字幕的长度相近，不要大幅改变字幕的切分位置\n3. 仅修正错误，不要改写原意或增删内容\n4. 如果某条字幕已经正确，原样保留\n5. 输出格式必须是JSON数组\n\n## 原文内容\n\n{original_text}\n\n\n## 待矫正的字幕\n\n[\n{subtitle_json}\n]\n\n\n## 输出格式\n严格输出JSON数组，每个元素包含index和corrected_text字段：\n\n[\n {{\"index\": 1, \"corrected_text\": \"修正后的文本\"}},\n {{\"index\": 2, \"corrected_text\": \"修正后的文本\"}}\n]\n\n\n请开始矫正：\n\"\"\"\n return textwrap.dedent(prompt)"
},
{
"path": "SonicVale/app/core/response.py",
"content": "# app/core/response.py\nfrom pydantic.generics import GenericModel\nfrom typing import Generic, TypeVar, Optional\n\nT = TypeVar(\"T\")\n\nclass Res(GenericModel, Generic[T]):\n code: int = 200\n message: str = \"success\"\n data: Optional[T] = None\n"
},
{
"path": "SonicVale/app/core/subtitle/ASRData.py",
"content": "import json\nimport logging\nimport re\nfrom typing import List\nfrom pathlib import Path\n\nclass ASRDataSeg:\n def __init__(self, text, start_time, end_time):\n self.text = text\n self.start_time = start_time\n self.end_time = end_time\n\n def to_srt_ts(self) -> str:\n \"\"\"Convert to SRT timestamp format\"\"\"\n return f\"{self._ms_to_srt_time(self.start_time)} --> {self._ms_to_srt_time(self.end_time)}\"\n\n\n def to_lrc_ts(self) -> str:\n \"\"\"Convert to LRC timestamp format\"\"\"\n return f\"[{self._ms_to_lrc_time(self.start_time)}]\"\n \n def to_ass_ts(self) -> tuple[str, str]:\n \"\"\"Convert to ASS timestamp format\"\"\"\n return self._ms_to_ass_ts(self.start_time), self._ms_to_ass_ts(self.end_time)\n\n def _ms_to_lrc_time(self, ms) -> str:\n seconds = ms / 1000\n minutes, seconds = divmod(seconds, 60)\n return f\"{int(minutes):02}:{seconds:.2f}\"\n \n @staticmethod\n def _ms_to_srt_time(ms) -> str:\n \"\"\"Convert milliseconds to SRT time format (HH:MM:SS,mmm)\"\"\"\n total_seconds, milliseconds = divmod(ms, 1000)\n minutes, seconds = divmod(total_seconds, 60)\n hours, minutes = divmod(minutes, 60)\n return f\"{int(hours):02}:{int(minutes):02}:{int(seconds):02},{int(milliseconds):03}\"\n\n @staticmethod\n def _ms_to_ass_ts(ms) -> str:\n \"\"\"Convert milliseconds to ASS timestamp format (H:MM:SS.cc)\"\"\"\n total_seconds, milliseconds = divmod(ms, 1000)\n minutes, seconds = divmod(total_seconds, 60)\n hours, minutes = divmod(minutes, 60)\n # ASS格式使用厘秒(1/100秒)而不是毫秒\n centiseconds = int(milliseconds / 10)\n return f\"{int(hours):01}:{int(minutes):02}:{int(seconds):02}.{centiseconds:02}\"\n\n @property\n def transcript(self) -> str:\n \"\"\"Return segment text\"\"\"\n return self.text\n\n def __str__(self) -> str:\n return f\"ASRDataSeg({self.text}, {self.start_time}, {self.end_time})\"\n\n\nclass ASRData:\n def __init__(self, segments: List[ASRDataSeg]):\n self.segments = segments\n\n def __iter__(self):\n return iter(self.segments)\n \n def __len__(self) -> int:\n return len(self.segments)\n \n def has_data(self) -> bool:\n \"\"\"Check if there are any utterances\"\"\"\n return len(self.segments) > 0\n \n def is_word_timestamp(self) -> bool:\n \"\"\"\n 判断是否是字级时间戳\n 规则：\n 1. 对于英文，每个segment应该只包含一个单词\n 2. 对于中文，每个segment应该只包含一个汉字\n 3. 允许20%的误差率\n \"\"\"\n if not self.segments:\n return False\n \n valid_segments = 0\n total_segments = len(self.segments)\n \n for seg in self.segments:\n text = seg.text.strip()\n # 检查是否只包含一个英文单词或一个汉字\n if (len(text.split()) == 1 and text.isascii()) or len(text.strip()) <= 2:\n valid_segments += 1\n logging.info(\"valid_segments: %s, total_segments: %s\", valid_segments, total_segments)\n return (valid_segments / total_segments) >= 0.8\n\n\n def save(self, save_path: str, ass_style: str = None, layout: str = \"原文在上\") -> None:\n \"\"\"Save the ASRData to a file\"\"\"\n # 根据文件后缀名选择保存格式\n Path(save_path).parent.mkdir(parents=True, exist_ok=True)\n if save_path.endswith('.srt'):\n self.to_srt(save_path=save_path)\n elif save_path.endswith('.txt'):\n with open(save_path, 'w', encoding='utf-8') as f:\n f.write(self.to_txt())\n elif save_path.endswith('.json'):\n with open(save_path, 'w', encoding='utf-8') as f:\n json.dump(self.to_json(), f, ensure_ascii=False)\n elif save_path.endswith('.ass'):\n self.to_ass(save_path=save_path, style_str=ass_style, layout=layout)\n else:\n raise ValueError(f\"Unsupported file extension: {save_path}\")\n\n def to_txt(self) -> str:\n \"\"\"Convert to plain text subtitle format (without timestamps)\"\"\"\n return \"\\n\".join(seg.transcript for seg in self.segments)\n\n def to_srt(self, save_path=None) -> str:\n \"\"\"Convert to SRT subtitle format\"\"\"\n srt_text = \"\\n\".join(\n f\"{n}\\n{seg.to_srt_ts()}\\n{seg.transcript}\\n\"\n for n, seg in enumerate(self.segments, 1))\n if save_path:\n with open(save_path, 'w', encoding='utf-8') as f:\n f.write(srt_text)\n return srt_text\n\n def to_lrc(self, save_path=None) -> str:\n \"\"\"Convert to LRC subtitle format\"\"\"\n lrc_text = \"\\n\".join(\n f\"{seg.to_lrc_ts()}{seg.transcript}\" for seg in self.segments\n )\n if save_path:\n with open(save_path, 'w', encoding='utf-8') as f:\n f.write(lrc_text)\n return lrc_text\n\n def to_json(self) -> dict:\n result_json = {}\n for i, segment in enumerate(self.segments, 1):\n # 检查是否有换行符\n if \"\\n\" in segment.text:\n original_subtitle, translated_subtitle = segment.text.split(\"\\n\")\n else:\n original_subtitle, translated_subtitle = segment.text, \"\"\n\n result_json[str(i)] = {\n \"start_time\": segment.start_time,\n \"end_time\": segment.end_time,\n \"original_subtitle\": original_subtitle,\n \"translated_subtitle\": translated_subtitle\n }\n return result_json\n\n def to_ass(self, style_str: str = None, layout: str = \"原文在上\", save_path: str = None) -> str:\n \"\"\"转换为ASS字幕格式\n \n Args:\n style_str: ASS样式字符串,为空则使用默认样式\n layout: 字幕布局,可选值[\"译文在上\", \"原文在上\", \"仅原文\", \"仅译文\"]\n \n Returns:\n ASS格式字幕内容\n \"\"\"\n # 默认样式\n if not style_str:\n style_str = (\n \"[V4+ Styles]\\n\"\n \"Format: Name,Fontname,Fontsize,PrimaryColour,SecondaryColour,OutlineColour,BackColour,\"\n \"Bold,Italic,Underline,StrikeOut,ScaleX,ScaleY,Spacing,Angle,BorderStyle,Outline,Shadow,\"\n \"Alignment,MarginL,MarginR,MarginV,Encoding\\n\"\n \"Style: Default,微软雅黑,66,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,\"\n \"0,0,1,2,0,2,10,10,10,1\\n\"\n \"Style: Translate,微软雅黑,40,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,\"\n \"0,0,1,2,0,2,10,10,10,1\"\n )\n\n # 构建ASS文件头\n ass_content = (\n \"[Script Info]\\n\"\n \"; Script generated by VideoCaptioner\\n\"\n \"; https://github.com/weifeng2333\\n\"\n \"ScriptType: v4.00+\\n\"\n \"PlayResX: 1280\\n\"\n \"PlayResY: 720\\n\\n\"\n f\"{style_str}\\n\\n\"\n \"[Events]\\n\"\n \"Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text\\n\"\n )\n\n # 根据布局生成对话内容\n for seg in self.segments:\n start_time = seg.to_ass_ts()[0]\n end_time = seg.to_ass_ts()[1]\n dialogue_template = 'Dialogue: 0,{},{},{},,0,0,0,,{}\\n'\n\n # 检查是否有换行符分隔的原文和译文\n if \"\\n\" in seg.text:\n original, translate = seg.text.split(\"\\n\")\n if layout == \"译文在上\" and translate:\n ass_content += dialogue_template.format(start_time, end_time, \"Secondary\", original)\n ass_content += dialogue_template.format(start_time, end_time, \"Default\", translate)\n elif layout == \"原文在上\" and translate:\n ass_content += dialogue_template.format(start_time, end_time, \"Secondary\", translate)\n ass_content += dialogue_template.format(start_time, end_time, \"Default\", original)\n elif layout == \"仅原文\":\n ass_content += dialogue_template.format(start_time, end_time, \"Default\", original)\n elif layout == \"仅译文\" and translate:\n ass_content += dialogue_template.format(start_time, end_time, \"Default\", translate)\n else:\n original = seg.text\n ass_content += dialogue_template.format(start_time, end_time, \"Default\", original)\n # 根据布局生成对话行\n \n if save_path:\n with open(save_path, 'w', encoding='utf-8') as f:\n f.write(ass_content)\n return ass_content\n\n def merge_segments(self, start_index: int, end_index: int, merged_text: str = None):\n \"\"\"合并从 start_index 到 end_index 的段（包含）。\"\"\"\n if start_index < 0 or end_index >= len(self.segments) or start_index > end_index:\n raise IndexError(\"无效的段索引。\")\n merged_start_time = self.segments[start_index].start_time\n merged_end_time = self.segments[end_index].end_time\n if merged_text is None:\n merged_text = ''.join(seg.text for seg in self.segments[start_index:end_index+1])\n merged_seg = ASRDataSeg(merged_text, merged_start_time, merged_end_time)\n # 替换 segments[start_index:end_index+1] 为 merged_seg\n self.segments[start_index:end_index+1] = [merged_seg]\n\n def merge_with_next_segment(self, index: int) -> None:\n \"\"\"合并指定索引的段与下一个段。\"\"\"\n if index < 0 or index >= len(self.segments) - 1:\n raise IndexError(\"索引超出范围或没有下一个段可合并。\")\n current_seg = self.segments[index]\n next_seg = self.segments[index + 1]\n\n # 合并文本\n merged_text = f\"{current_seg.text} {next_seg.text}\"\n merged_start_time = current_seg.start_time\n merged_end_time = next_seg.end_time\n merged_seg = ASRDataSeg(merged_text, merged_start_time, merged_end_time)\n\n # 替换当前段为合并后的段\n self.segments[index] = merged_seg\n # 删除下一个段\n del self.segments[index + 1]\n\n def __str__(self):\n return self.to_txt()\n\ndef from_subtitle_file(file_path: str) -> 'ASRData':\n \"\"\"从文件路径加载ASRData实例\n \n Args:\n file_path: 字幕文件路径，支持.srt、.vtt、.ass、.json格式\n \n Returns:\n ASRData: 解析后的ASRData实例\n \n Raises:\n ValueError: 不支持的文件格式或文件读取错误\n \"\"\"\n file_path = Path(file_path)\n if not file_path.exists():\n raise FileNotFoundError(f\"文件不存在: {file_path}\")\n \n try:\n content = file_path.read_text(encoding='utf-8')\n except UnicodeDecodeError:\n content = file_path.read_text(encoding='gbk')\n \n suffix = file_path.suffix.lower()\n \n if suffix == '.srt':\n return from_srt(content)\n elif suffix == '.vtt':\n if '' in content: # YouTube VTT格式包含字级时间戳\n return from_youtube_vtt(content)\n return from_vtt(content)\n elif suffix == '.ass':\n return from_ass(content)\n elif suffix == '.json':\n return from_json(json.loads(content))\n else:\n raise ValueError(f\"不支持的文件格式: {suffix}\")\n\ndef from_json(json_data: dict) -> 'ASRData':\n \"\"\"从JSON数据创建ASRData实例\"\"\"\n segments = []\n for i in sorted(json_data.keys(), key=int):\n segment_data = json_data[i]\n text = segment_data['original_subtitle']\n if segment_data['translated_subtitle']:\n text += '\\n' + segment_data['translated_subtitle']\n segment = ASRDataSeg(\n text=text,\n start_time=segment_data['start_time'],\n end_time=segment_data['end_time']\n )\n segments.append(segment)\n return ASRData(segments)\n\ndef from_srt(srt_str: str) -> 'ASRData':\n \"\"\"\n 从SRT格式的字符串创建ASRData实例。\n\n :param srt_str: 包含SRT格式字幕的字符串。\n :return: 解析后的ASRData实例。\n \"\"\"\n segments = []\n srt_time_pattern = re.compile(\n r'(\\d{2}):(\\d{2}):(\\d{1,2})[.,](\\d{3})\\s-->\\s(\\d{2}):(\\d{2}):(\\d{1,2})[.,](\\d{3})'\n )\n\n for block in re.split(r'\\n\\s*\\n', srt_str.strip()):\n lines = block.splitlines()\n if len(lines) < 3:\n continue\n\n match = srt_time_pattern.match(lines[1])\n if not match:\n continue\n\n time_parts = list(map(int, match.groups()))\n start_time = sum([\n time_parts[0] * 3600000,\n time_parts[1] * 60000,\n time_parts[2] * 1000,\n time_parts[3]\n ])\n end_time = sum([\n time_parts[4] * 3600000,\n time_parts[5] * 60000,\n time_parts[6] * 1000,\n time_parts[7]\n ])\n\n text = '\\n'.join(lines[2:]).strip()\n segments.append(ASRDataSeg(text, start_time, end_time))\n\n return ASRData(segments)\n\ndef from_vtt(vtt_str: str) -> 'ASRData':\n \"\"\"\n 从YouTube VTT格式的字符串创建ASRData实例。\n \n :param vtt_str: YouTube VTT格式的字幕字符串\n :return: ASRData实例\n \"\"\"\n segments = []\n # 跳过头部元数据\n content = vtt_str.split('\\n\\n')[2:]\n \n current_text = \"\"\n current_start = 0\n current_end = 0\n \n for block in content:\n lines = block.strip().split('\\n')\n if not lines:\n continue\n \n # 解析时间戳行\n timestamp_line = lines[0]\n if '-->' not in timestamp_line:\n continue\n \n # 提取开始和结束时间\n times = timestamp_line.split(' --> ')[0]\n hours, minutes, seconds = times.split(':')\n seconds, milliseconds = seconds.split('.')\n start_time = (int(hours) * 3600 + int(minutes) * 60 + int(seconds)) * 1000 + int(milliseconds)\n \n times = timestamp_line.split(' --> ')[1].split()[0]\n hours, minutes, seconds = times.split(':')\n seconds, milliseconds = seconds.split('.')\n end_time = (int(hours) * 3600 + int(minutes) * 60 + int(seconds)) * 1000 + int(milliseconds)\n \n # 提取并清文本内容\n if len(lines) > 1:\n text_line = lines[1]\n # 移除时间戳和样式标记\n cleaned_text = re.sub(r'<\\d{2}:\\d{2}:\\d{2}\\.\\d{3}>', '', text_line)\n cleaned_text = re.sub(r'', '', cleaned_text)\n cleaned_text = cleaned_text.strip()\n \n if cleaned_text and cleaned_text != \" \":\n segments.append(ASRDataSeg(cleaned_text, start_time, end_time))\n \n return ASRData(segments)\n\ndef from_youtube_vtt(vtt_str: str) -> 'ASRData':\n \"\"\"\n 从YouTube VTT格式的字符串创建ASRData实例，提取字级时间戳。\n \n :param vtt_str: 包含VTT格式字幕的字符串\n :return: 解析后的ASRData实例\n \"\"\"\n def parse_timestamp(ts: str) -> int:\n \"\"\"将时间戳字符串转换为毫秒\"\"\"\n h, m, s = ts.split(':')\n return int(float(h) * 3600000 + float(m) * 60000 + float(s) * 1000)\n \n def split_timestamped_text(text: str) -> List[ASRDataSeg]:\n \"\"\"分离带时间戳的文本为单词段\"\"\"\n # 匹配 <时间戳>文本 的模式\n pattern = re.compile(r'<(\\d{2}:\\d{2}:\\d{2}\\.\\d{3})>([^<]*)')\n matches = list(pattern.finditer(text))\n word_segments = []\n \n for i in range(len(matches) - 1):\n current_match = matches[i]\n next_match = matches[i + 1]\n \n start_time = parse_timestamp(current_match.group(1))\n end_time = parse_timestamp(next_match.group(1))\n word = current_match.group(2).strip()\n \n if word: # 只有当文本不为空时才创建segment\n word_segments.append(ASRDataSeg(word, start_time, end_time))\n \n return word_segments\n \n segments = []\n # 跳过WEBVTT头部\n blocks = re.split(r'\\n\\n+', vtt_str.strip())\n \n # 时间戳匹配模式\n timestamp_pattern = re.compile(\n r'(\\d{2}):(\\d{2}):(\\d{2}\\.\\d{3})\\s*-->\\s*(\\d{2}):(\\d{2}):(\\d{2}\\.\\d{3})'\n ) \n for block in blocks:\n lines = block.strip().split('\\n')\n if not lines:\n continue\n \n # 匹配时间戳行\n match = timestamp_pattern.match(lines[0])\n if not match:\n continue\n \n # 计算块的开始和结束时间\n block_start_time = (\n int(match.group(1)) * 3600000 +\n int(match.group(2)) * 60000 +\n float(match.group(3)) * 1000\n )\n block_end_time = (\n int(match.group(4)) * 3600000 +\n int(match.group(5)) * 60000 +\n float(match.group(6)) * 1000\n )\n \n # 获取文本内容\n text = '\\n'.join(lines)\n \n timestamp_row = re.search(r'\\n(.*?.*?.*)', block)\n if timestamp_row:\n text = re.sub(r'|', '', timestamp_row.group(1))\n block_start_time_string = f\"{match.group(1)}:{match.group(2)}:{match.group(3)}\"\n block_end_time_string = f\"{match.group(4)}:{match.group(5)}:{match.group(6)}\"\n text = f\"<{block_start_time_string}>{text}<{block_end_time_string}>\"\n \n # 分离每个带时间戳的单词\n word_segments = split_timestamped_text(text)\n segments.extend(word_segments)\n \n return ASRData(segments)\n\ndef from_ass(ass_str: str) -> 'ASRData':\n \"\"\"\n 从ASS格式的字符串创建ASRData实例。\n \n :param ass_str: 包含ASS格式字幕的字符串\n :return: ASRData实例\n \"\"\"\n segments = []\n # ASS时间戳格式: H:MM:SS.cc\n ass_time_pattern = re.compile(r'Dialogue: \\d+,(\\d+:\\d{2}:\\d{2}\\.\\d{2}),(\\d+:\\d{2}:\\d{2}\\.\\d{2}),(.*?),.*?,\\d+,\\d+,\\d+,.*?,(.*?)$')\n \n def parse_ass_time(time_str: str) -> int:\n \"\"\"将ASS时间戳转换为毫秒\"\"\"\n hours, minutes, seconds = time_str.split(':')\n seconds, centiseconds = seconds.split('.')\n return (int(hours) * 3600000 + \n int(minutes) * 60000 + \n int(seconds) * 1000 + \n int(centiseconds) * 10) # 厘秒转毫秒\n \n # 按行处理ASS文件\n for line in ass_str.splitlines():\n if line.startswith('Dialogue:'):\n match = ass_time_pattern.match(line)\n if match:\n start_time = parse_ass_time(match.group(1))\n end_time = parse_ass_time(match.group(2))\n text = match.group(4)\n \n # 清理ASS格式标记\n text = re.sub(r'\\{[^}]*\\}', '', text) # 移除样式标记 {xxx}\n text = text.replace('\\\\N', '\\n') # 处理换行符\n text = text.strip()\n \n if text: # 只有当文本不为空时才创建segment\n segments.append(ASRDataSeg(text, start_time, end_time))\n \n return ASRData(segments)\n\nif __name__ == '__main__':\n ass_style_str = \"\"\"[V4+ Styles]\nFormat: Name,Fontname,Fontsize,PrimaryColour,SecondaryColour,OutlineColour,BackColour,Bold,Italic,Underline,StrikeOut,ScaleX,ScaleY,Spacing,Angle,BorderStyle,Outline,Shadow,Alignment,MarginL,MarginR,MarginV,Encoding\nStyle: Default,微软雅黑,62,&H0017f1be,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,1.0,0,1,0.8,0,2,10,10,10,1\nStyle: Secondary,微软雅黑,40,&H00ffffff,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,0.0,0,1,0.0,0,2,10,10,10,1\"\"\"\n # 测试\n from pathlib import Path\n # vtt_file_path = r\"E:\\GithubProject\\VideoCaptioner\\app\\work_dir\\Setting the record straight\\subtitle\\original_subtitle.en.vtt\"\n # vtt_file_path = r\"E:\\GithubProject\\VideoCaptioner\\work_dir\\Wake up babe a dangerous new open-source AI model is here\\subtitle\\original.en.vtt\"\n # asr_data = from_youtube_vtt(Path(vtt_file_path).read_text(encoding=\"utf-8\"))\n srt_file_path = r\"E:\\GithubProject\\VideoCaptioner\\app\\work_dir\\低视力音乐助人者_mp4\\result_subtitle.srt\"\n asr_data = from_srt(Path(srt_file_path).read_text(encoding=\"utf-8\"))\n\n logging.info(\"%s\", asr_data.to_ass(style_str=ass_style_str, save_path=srt_file_path.replace(\".srt\", \".ass\")))\n # pass\n # asr_data = ASRData(seg)\n # Uncomment to test different formats:\n # print(asr_data.to_srt(save_path=vtt_file_path.replace(\".vtt\", \".srt\")))\n # print(asr_data.to_lrc())\n # print(asr_data.to_txt())\n # print(asr_data.to_json())\n # print(asr_data.to_json())\n\n\n\n"
},
{
"path": "SonicVale/app/core/subtitle/BaseASR.py",
"content": "import json\nimport logging\nimport os\nimport zlib\nimport tempfile\nimport threading\n\nfrom .ASRData import ASRDataSeg, ASRData\n\n\nclass BaseASR:\n SUPPORTED_SOUND_FORMAT = [\"flac\", \"m4a\", \"mp3\", \"wav\"]\n CACHE_FILE = os.path.join(tempfile.gettempdir(), \"bk_asr\", \"asr_cache.json\")\n _lock = threading.Lock()\n\n def __init__(self, audio_path: [str, bytes], use_cache: bool = False):\n self.audio_path = audio_path\n self.file_binary = None\n\n self.crc32_hex = None\n self.use_cache = use_cache\n\n self._set_data()\n\n self.cache = self._load_cache()\n\n def _load_cache(self):\n if not self.use_cache:\n return {}\n os.makedirs(os.path.dirname(self.CACHE_FILE), exist_ok=True)\n with self._lock:\n if os.path.exists(self.CACHE_FILE):\n try:\n with open(self.CACHE_FILE, 'r', encoding='utf-8') as f:\n cache = json.load(f)\n if isinstance(cache, dict):\n return cache\n except (json.JSONDecodeError, IOError):\n return {}\n return {}\n\n def _save_cache(self):\n if not self.use_cache:\n return\n with self._lock:\n try:\n with open(self.CACHE_FILE, 'w', encoding='utf-8') as f:\n json.dump(self.cache, f, ensure_ascii=False, indent=2)\n if os.path.exists(self.CACHE_FILE) and os.path.getsize(self.CACHE_FILE) > 10 * 1024 * 1024:\n os.remove(self.CACHE_FILE)\n except IOError as e:\n logging.error(f\"Failed to save cache: {e}\")\n\n def _set_data(self):\n if isinstance(self.audio_path, bytes):\n self.file_binary = self.audio_path\n else:\n ext = self.audio_path.split(\".\")[-1].lower()\n assert ext in self.SUPPORTED_SOUND_FORMAT, f\"Unsupported sound format: {ext}\"\n assert os.path.exists(self.audio_path), f\"File not found: {self.audio_path}\"\n with open(self.audio_path, \"rb\") as f:\n self.file_binary = f.read()\n crc32_value = zlib.crc32(self.file_binary) & 0xFFFFFFFF\n self.crc32_hex = format(crc32_value, '08x')\n\n def _get_key(self):\n return f\"{self.__class__.__name__}-{self.crc32_hex}\"\n\n def run(self):\n k = self._get_key()\n if k in self.cache and self.use_cache:\n resp_data = self.cache[k]\n else:\n resp_data = self._run()\n # Cache the result\n self.cache[k] = resp_data\n self._save_cache()\n segments = self._make_segments(resp_data)\n return ASRData(segments)\n\n def _make_segments(self, resp_data: dict) -> list[ASRDataSeg]:\n raise NotImplementedError(\"_make_segments method must be implemented in subclass\")\n\n def _run(self) -> dict:\n \"\"\" Run the ASR service and return the response data. \"\"\"\n raise NotImplementedError(\"_run method must be implemented in subclass\")\n\n\n\n"
},
{
"path": "SonicVale/app/core/subtitle/BcutASR.py",
"content": "import json\nimport logging\nimport time\nfrom os import PathLike\nfrom typing import Optional\n\nimport requests\n\nfrom .ASRData import ASRData, ASRDataSeg\nfrom .BaseASR import BaseASR\n\n\n__version__ = \"0.0.3\"\n\nAPI_BASE_URL = \"https://member.bilibili.com/x/bcut/rubick-interface\"\n\n# 申请上传\nAPI_REQ_UPLOAD = API_BASE_URL + \"/resource/create\"\n\n# 提交上传\nAPI_COMMIT_UPLOAD = API_BASE_URL + \"/resource/create/complete\"\n\n# 创建任务\nAPI_CREATE_TASK = API_BASE_URL + \"/task\"\n\n# 查询结果\nAPI_QUERY_RESULT = API_BASE_URL + \"/task/result\"\n\n\nclass BcutASR(BaseASR):\n \"\"\"必剪 语音识别接口\"\"\"\n headers = {\n 'User-Agent': 'Bilibili/1.0.0 (https://www.bilibili.com)',\n 'Content-Type': 'application/json'\n }\n\n def __init__(self, audio_path: [str, bytes], use_cache: bool = False):\n super().__init__(audio_path, use_cache=use_cache)\n self.session = requests.Session()\n self.task_id = None\n self.__etags = []\n\n self.__in_boss_key: Optional[str, None] = None\n self.__resource_id: Optional[str, None] = None\n self.__upload_id: Optional[str, None] = None\n self.__upload_urls: Optional[list[str]] = []\n self.__per_size: Optional[int, None] = None\n self.__clips: Optional[int, None] = None\n\n self.__etags: Optional[list[str]] = []\n self.__download_url: Optional[str, None] = None\n self.task_id: Optional[str, None] = None\n\n\n def upload(self) -> None:\n \"\"\"申请上传\"\"\"\n if not self.file_binary:\n raise ValueError(\"none set data\")\n payload = json.dumps({\n \"type\": 2,\n \"name\": \"audio.mp3\",\n \"size\": len(self.file_binary),\n \"ResourceFileType\": \"mp3\",\n \"model_id\": \"8\",\n })\n\n resp = requests.post(\n API_REQ_UPLOAD,\n data=payload,\n headers=self.headers\n )\n resp.raise_for_status()\n resp = resp.json()\n resp_data = resp[\"data\"]\n\n self.__in_boss_key = resp_data[\"in_boss_key\"]\n self.__resource_id = resp_data[\"resource_id\"]\n self.__upload_id = resp_data[\"upload_id\"]\n self.__upload_urls = resp_data[\"upload_urls\"]\n self.__per_size = resp_data[\"per_size\"]\n self.__clips = len(resp_data[\"upload_urls\"])\n\n logging.info(\n f\"申请上传成功, 总计大小{resp_data['size'] // 1024}KB, {self.__clips}分片, 分片大小{resp_data['per_size'] // 1024}KB: {self.__in_boss_key}\"\n )\n self.__upload_part()\n self.__commit_upload()\n\n def __upload_part(self) -> None:\n \"\"\"上传音频数据\"\"\"\n for clip in range(self.__clips):\n start_range = clip * self.__per_size\n end_range = (clip + 1) * self.__per_size\n logging.info(f\"开始上传分片{clip}: {start_range}-{end_range}\")\n resp = requests.put(\n self.__upload_urls[clip],\n data=self.file_binary[start_range:end_range],\n headers=self.headers\n )\n resp.raise_for_status()\n etag = resp.headers.get(\"Etag\")\n self.__etags.append(etag)\n logging.info(f\"分片{clip}上传成功: {etag}\")\n\n def __commit_upload(self) -> None:\n \"\"\"提交上传数据\"\"\"\n data = json.dumps({\n \"InBossKey\": self.__in_boss_key,\n \"ResourceId\": self.__resource_id,\n \"Etags\": \",\".join(self.__etags),\n \"UploadId\": self.__upload_id,\n \"model_id\": \"8\",\n })\n resp = requests.post(\n API_COMMIT_UPLOAD,\n data=data,\n headers=self.headers\n )\n resp.raise_for_status()\n resp = resp.json()\n self.__download_url = resp[\"data\"][\"download_url\"]\n logging.info(f\"提交成功\")\n\n def create_task(self) -> str:\n \"\"\"开始创建转换任务\"\"\"\n resp = requests.post(\n API_CREATE_TASK, json={\"resource\": self.__download_url, \"model_id\": \"8\"}, headers=self.headers\n )\n resp.raise_for_status()\n resp = resp.json()\n self.task_id = resp[\"data\"][\"task_id\"]\n logging.info(f\"任务已创建: {self.task_id}\")\n return self.task_id\n\n def result(self, task_id: Optional[str] = None):\n \"\"\"查询转换结果\"\"\"\n resp = requests.get(API_QUERY_RESULT, params={\"model_id\": 7, \"task_id\": task_id or self.task_id}, headers=self.headers)\n resp.raise_for_status()\n resp = resp.json()\n return resp[\"data\"]\n\n def _run(self):\n self.upload()\n self.create_task()\n # 轮询检查任务状态\n for _ in range(500):\n task_resp = self.result()\n if task_resp[\"state\"] == 4:\n break\n time.sleep(1)\n logging.info(f\"转换成功\")\n return json.loads(task_resp[\"result\"])\n\n def _make_segments(self, resp_data: dict) -> list[ASRDataSeg]:\n return [ASRDataSeg(u['transcript'], u['start_time'], u['end_time']) for u in resp_data['utterances']]\n\n\nif __name__ == '__main__':\n logging.basicConfig(level=logging.INFO, format=\"%(asctime)s - %(levelname)s - %(message)s\")\n # Example usage\n audio_file = r\"test.mp3\"\n asr = BcutASR(audio_file)\n asr_data = asr.run()\n logging.info(\"%s\", asr_data)\n"
},
{
"path": "SonicVale/app/core/subtitle/JianYingASR.py",
"content": "import datetime\nimport hashlib\nimport hmac\nimport json\nimport os\nimport time\nimport uuid\nfrom typing import Dict, Tuple, Union\n\nimport requests\n\nfrom .ASRData import ASRDataSeg\nfrom .BaseASR import BaseASR\n\n\n# from ASRData import ASRDataSeg\n# from BaseASR import BaseASR\n\nclass JianYingASR(BaseASR):\n def __init__(self, audio_path: Union[str, bytes], use_cache: bool = False, need_word_time_stamp: bool = False,\n start_time: float = 0, end_time: float = 6000):\n super().__init__(audio_path, use_cache)\n self.audio_path = audio_path\n self.end_time = end_time\n self.start_time = start_time\n\n # AWS credentials\n self.session_token = None\n self.secret_key = None\n self.access_key = None\n\n # Upload details\n self.store_uri = None\n self.auth = None\n self.upload_id = None\n self.session_key = None\n self.upload_hosts = None\n\n self.need_word_time_stamp = need_word_time_stamp\n self.tdid = \"3943278516897751\" if datetime.datetime.now().year != 2024 else f\"{uuid.getnode():012d}\"\n\n def submit(self) -> str:\n \"\"\"Submit the task\"\"\"\n url = \"https://lv-pc-api-sinfonlinec.ulikecam.com/lv/v1/audio_subtitle/submit\"\n payload = {\n \"adjust_endtime\": 200,\n \"audio\": self.store_uri,\n \"caption_type\": 2,\n \"client_request_id\": \"45faf98c-160f-4fae-a649-6d89b0fe35be\",\n \"max_lines\": 1,\n \"songs_info\": [{\"end_time\": self.end_time, \"id\": \"\", \"start_time\": self.start_time}],\n \"words_per_line\": 16\n }\n\n sign, device_time = self._generate_sign_parameters(url='/lv/v1/audio_subtitle/submit', pf='4', appvr='6.6.0',\n tdid=self.tdid)\n headers = self._build_headers(device_time, sign)\n response = requests.post(url, json=payload, headers=headers)\n\n resp_data = response.json()\n\n if resp_data.get('ret') != '0':\n error_msg = f\"API Error: {resp_data.get('errmsg', 'Unknown error')} (ret: {resp_data.get('ret')})\"\n raise ValueError(error_msg)\n\n query_id = resp_data['data']['id']\n return query_id\n\n def upload(self):\n \"\"\"Upload the file\"\"\"\n self._upload_sign()\n self._upload_auth()\n self._upload_file()\n self._upload_check()\n uri = self._upload_commit()\n return uri\n\n def query(self, query_id: str):\n \"\"\"Query the task\"\"\"\n url = \"https://lv-pc-api-sinfonlinec.ulikecam.com/lv/v1/audio_subtitle/query\"\n payload = {\n \"id\": query_id,\n \"pack_options\": {\"need_attribute\": True}\n }\n sign, device_time = self._generate_sign_parameters(url='/lv/v1/audio_subtitle/query', pf='4', appvr='6.6.0',\n tdid=self.tdid)\n headers = self._build_headers(device_time, sign)\n response = requests.post(url, json=payload, headers=headers)\n resp_data = response.json()\n\n if resp_data.get('ret') != '0':\n error_msg = f\"API Error: {resp_data.get('errmsg', 'Unknown error')} (ret: {resp_data.get('ret')})\"\n\n raise ValueError(error_msg)\n\n return resp_data\n\n\n def _run(self, callback=None):\n # logging.info(\"正在上传文件...\")\n if callback:\n callback(20, \"正在上传...\")\n self.upload()\n if callback:\n callback(50, \"提交任务...\")\n query_id = self.submit()\n if callback:\n callback(60, \"获取结果...\")\n resp_data = self.query(query_id)\n if callback:\n callback(100, \"转录完成\")\n return resp_data\n\n def _make_segments(self, resp_data: dict) -> list[ASRDataSeg]:\n if self.need_word_time_stamp:\n return [ASRDataSeg(w['text'].strip(), w['start_time'], w['end_time']) for u in\n resp_data['data']['utterances'] for w in u['words']]\n else:\n return [ASRDataSeg(u['text'], u['start_time'], u['end_time']) for u in resp_data['data']['utterances']]\n\n def _get_key(self):\n return f\"{self.__class__.__name__}-{self.crc32_hex}-{self.need_word_time_stamp}\"\n\n def _generate_sign_parameters(self, url: str, pf: str = '4', appvr: str = '6.6.0', tdid='') -> \\\n Tuple[str, str]:\n \"\"\"Generate signature and timestamp via an HTTP request\"\"\"\n current_time = str(int(time.time()))\n data = {\n 'url': url,\n 'current_time': current_time,\n 'pf': pf,\n 'appvr': appvr,\n 'tdid': self.tdid\n }\n # Replace with your actual endpoint URL\n get_sign_url = 'https://asrtools-update.bkfeng.top/sign'\n try:\n response = requests.post(get_sign_url, json=data)\n response.raise_for_status()\n response_data = response.json()\n sign = response_data.get('sign')\n if not sign:\n raise ValueError(\"No 'sign' in response\")\n except requests.exceptions.RequestException as e:\n raise SystemExit(f\"HTTP Request failed: {e}\")\n except ValueError as ve:\n raise SystemExit(f\"Invalid response: {ve}\")\n return sign.lower(), current_time\n\n def _build_headers(self, device_time: str, sign: str) -> Dict[str, str]:\n \"\"\"Build headers for requests\"\"\"\n return {\n 'User-Agent': \"Cronet/TTNetVersion:d4572e53 2024-06-12 QuicVersion:4bf243e0 2023-04-17\",\n 'appvr': \"6.6.0\",\n 'device-time': str(device_time),\n 'pf': \"4\",\n 'sign': sign,\n 'sign-ver': \"1\",\n 'tdid': self.tdid,\n }\n\n def _uplosd_headers(self):\n headers = {\n 'User-Agent': \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/81.0.4044.138 Safari/537.36 Thea/1.0.1\",\n 'Authorization': self.auth,\n 'Content-CRC32': self.crc32_hex,\n }\n return headers\n\n def _upload_sign(self):\n \"\"\"Get upload sign\"\"\"\n url = \"https://lv-pc-api-sinfonlinec.ulikecam.com/lv/v1/upload_sign\"\n payload = json.dumps({\"biz\": \"pc-recognition\"})\n sign, device_time = self._generate_sign_parameters(url='/lv/v1/upload_sign', pf='4', appvr='6.6.0',\n tdid=self.tdid)\n headers = self._build_headers(device_time, sign)\n response = requests.post(url, data=payload, headers=headers)\n response.raise_for_status()\n login_data = response.json()\n self.access_key = login_data['data']['access_key_id']\n self.secret_key = login_data['data']['secret_access_key']\n self.session_token = login_data['data']['session_token']\n return self.access_key, self.secret_key, self.session_token\n\n def _upload_auth(self):\n \"\"\"Get upload authorization\"\"\"\n if isinstance(self.audio_path, bytes):\n file_size = len(self.audio_path)\n else:\n file_size = os.path.getsize(self.audio_path)\n request_parameters = f'Action=ApplyUploadInner&FileSize={file_size}&FileType=object&IsInner=1&SpaceName=lv-mac-recognition&Version=2020-11-19&s=5y0udbjapi'\n\n t = datetime.datetime.utcnow()\n amz_date = t.strftime('%Y%m%dT%H%M%SZ')\n datestamp = t.strftime('%Y%m%d')\n headers = {\n \"x-amz-date\": amz_date,\n \"x-amz-security-token\": self.session_token\n }\n signature = aws_signature(self.secret_key, request_parameters, headers, region=\"cn\", service=\"vod\")\n authorization = f\"AWS4-HMAC-SHA256 Credential={self.access_key}/{datestamp}/cn/vod/aws4_request, SignedHeaders=x-amz-date;x-amz-security-token, Signature={signature}\"\n headers[\"authorization\"] = authorization\n response = requests.get(f\"https://vod.bytedanceapi.com/?{request_parameters}\", headers=headers)\n store_infos = response.json()\n\n self.store_uri = store_infos['Result']['UploadAddress']['StoreInfos'][0]['StoreUri']\n self.auth = store_infos['Result']['UploadAddress']['StoreInfos'][0]['Auth']\n self.upload_id = store_infos['Result']['UploadAddress']['StoreInfos'][0]['UploadID']\n self.session_key = store_infos['Result']['UploadAddress']['SessionKey']\n self.upload_hosts = store_infos['Result']['UploadAddress']['UploadHosts'][0]\n self.store_uri = store_infos['Result']['UploadAddress']['StoreInfos'][0]['StoreUri']\n return store_infos\n\n def _upload_file(self):\n \"\"\"Upload the file\"\"\"\n url = f\"https://{self.upload_hosts}/{self.store_uri}?partNumber=1&uploadID={self.upload_id}\"\n headers = self._uplosd_headers()\n response = requests.put(url, data=self.file_binary, headers=headers)\n resp_data = response.json()\n assert resp_data['success'] == 0, f\"File upload failed: {response.text}\"\n return resp_data\n\n def _upload_check(self):\n \"\"\"Check upload result\"\"\"\n url = f\"https://{self.upload_hosts}/{self.store_uri}?uploadID={self.upload_id}\"\n payload = f\"1:{self.crc32_hex}\"\n headers = self._uplosd_headers()\n response = requests.post(url, data=payload, headers=headers)\n resp_data = response.json()\n return resp_data\n\n def _upload_commit(self):\n \"\"\"Commit the uploaded file\"\"\"\n url = f\"https://{self.upload_hosts}/{self.store_uri}?uploadID={self.upload_id}&partNumber=1&x-amz-security-token={self.session_token}\"\n headers = self._uplosd_headers()\n response = requests.put(url, data=self.file_binary, headers=headers)\n return self.store_uri\n\n\ndef sign(key: bytes, msg: str) -> bytes:\n \"\"\"使用HMAC-SHA256生成签名\"\"\"\n return hmac.new(key, msg.encode('utf-8'), hashlib.sha256).digest()\n\n\ndef get_signature_key(secret_key: str, date_stamp: str, region_name: str, service_name: str) -> bytes:\n \"\"\"生成用于AWS签名的密钥\"\"\"\n k_date = sign(('AWS4' + secret_key).encode('utf-8'), date_stamp)\n k_region = sign(k_date, region_name)\n k_service = sign(k_region, service_name)\n k_signing = sign(k_service, 'aws4_request')\n return k_signing\n\n\ndef aws_signature(secret_key: str, request_parameters: str, headers: Dict[str, str],\n method: str = \"GET\", payload: str = '', region: str = \"cn\", service: str = \"vod\") -> str:\n \"\"\"生成AWS签名\"\"\"\n canonical_uri = '/'\n canonical_querystring = request_parameters\n canonical_headers = '\\n'.join([f\"{key}:{value}\" for key, value in headers.items()]) + '\\n'\n signed_headers = ';'.join(headers.keys())\n payload_hash = hashlib.sha256(payload.encode('utf-8')).hexdigest()\n canonical_request = f\"{method}\\n{canonical_uri}\\n{canonical_querystring}\\n{canonical_headers}\\n{signed_headers}\\n{payload_hash}\"\n\n amzdate = headers[\"x-amz-date\"]\n datestamp = amzdate.split('T')[0]\n\n algorithm = 'AWS4-HMAC-SHA256'\n credential_scope = f\"{datestamp}/{region}/{service}/aws4_request\"\n string_to_sign = f\"{algorithm}\\n{amzdate}\\n{credential_scope}\\n{hashlib.sha256(canonical_request.encode('utf-8')).hexdigest()}\"\n\n signing_key = get_signature_key(secret_key, datestamp, region, service)\n signature = hmac.new(signing_key, string_to_sign.encode('utf-8'), hashlib.sha256).hexdigest()\n return signature\n"
},
{
"path": "SonicVale/app/core/subtitle/KuaiShouASR.py",
"content": "import requests\n\nfrom .ASRData import ASRDataSeg\nfrom .BaseASR import BaseASR\n\n\nclass KuaiShouASR(BaseASR):\n def __init__(self, audio_path: [str, bytes], use_cache: bool = False):\n super().__init__(audio_path, use_cache)\n\n def _run(self) -> dict:\n return self._submit()\n\n def _make_segments(self, resp_data: dict) -> list[ASRDataSeg]:\n return [ASRDataSeg(u['text'], u['start_time'], u['end_time']) for u in resp_data['data']['text']]\n\n def _submit(self) -> dict:\n payload = {\n \"typeId\": \"1\"\n }\n files = [('file', ('test.mp3', self.file_binary, 'audio/mpeg'))]\n result = requests.post(\"https://ai.kuaishou.com/api/effects/subtitle_generate\", data=payload, files=files)\n return result.json()\n"
},
{
"path": "SonicVale/app/core/subtitle/WhisperASR.py",
"content": "import os\n\nfrom openai import OpenAI\n\nfrom .ASRData import ASRDataSeg\nfrom .BaseASR import BaseASR\n\n\n\nclass WhisperASR(BaseASR):\n def __init__(self, audio_path: [str, bytes], model: str = MODEL, use_cache: bool = False):\n super().__init__(audio_path, use_cache)\n self.base_url = os.getenv('OPENAI_BASE_URL')\n self.api_key = os.getenv('OPENAI_API_KEY')\n if not self.base_url or not self.api_key:\n raise ValueError(\"环境变量 OPENAI_BASE_URL 和 OPENAI_API_KEY 必须设置\")\n self.model = model\n self.client = OpenAI(base_url=self.base_url, api_key=self.api_key)\n\n def _run(self) -> dict:\n return self._submit()\n\n def _make_segments(self, resp_data: dict) -> list[ASRDataSeg]:\n return [ASRDataSeg(u['text'], u['start'], u['end']) for u in resp_data['segments']]\n\n def _get_key(self) -> str:\n return f\"{self.__class__.__name__}-{self.model}-{self.crc32_hex}-{self.model}\"\n\n def _submit(self) -> dict:\n completion = self.client.audio.transcriptions.create(\n model=self.model,\n temperature=0,\n response_format=\"verbose_json\",\n file=(\"test.mp3\", self.file_binary, \"audio/mp3\"),\n prompt=\"\",\n language=\"zh\"\n )\n return completion.to_dict()\n\n\n"
},
{
"path": "SonicVale/app/core/subtitle/__init__.py",
"content": ""
},
{
"path": "SonicVale/app/core/subtitle/subtitle_engine.py",
"content": "from app.core.subtitle.BcutASR import BcutASR\nfrom app.core.subtitle.JianYingASR import JianYingASR\nfrom app.core.prompts import get_subtitle_correction_prompt\nfrom app.core.llm_engine import LLMEngine\n\n\ndef generate_subtitle(audio_file,save_path):\n # asr = JianYingASR(audio_file)\n asr = BcutASR(audio_file)\n result = asr.run()\n result.to_srt(save_path)\n return result\n\n\n\n# 字幕矫正\nimport re\nimport difflib\nimport shutil\nimport logging\nfrom pypinyin import lazy_pinyin\n\n# -------------------- 基础工具 --------------------\n\ndef is_same_char(c1: str, c2: str) -> bool:\n \"\"\"字面相同或拼音相同（处理同音字）\"\"\"\n if c1 == c2:\n return True\n return lazy_pinyin(c1) == lazy_pinyin(c2)\n\ndef correct_text_with_pinyin(original: str, recognized: str) -> str:\n \"\"\"全局文本纠正：把 recognized 纠正到 original\"\"\"\n sm = difflib.SequenceMatcher(None, recognized, original, autojunk=False)\n out = []\n for tag, i1, i2, j1, j2 in sm.get_opcodes():\n if tag == \"equal\":\n out.append(recognized[i1:i2])\n else:\n r = recognized[i1:i2]\n o = original[j1:j2]\n seg = []\n L = max(len(r), len(o))\n for k in range(L):\n c1 = r[k] if k < len(r) else \"\"\n c2 = o[k] if k < len(o) else \"\"\n if c1 and c2 and is_same_char(c1, c2):\n seg.append(c2)\n elif c2:\n seg.append(c2)\n out.append(\"\".join(seg))\n return \"\".join(out)\n\n# -------------------- SRT 读写 --------------------\n\nSRT_BLOCK = re.compile(\n r\"(\\d+)\\s+([\\d:,]+ --> [\\d:,]+)\\s+([\\s\\S]*?)(?=\\n\\n|\\Z)\", re.MULTILINE\n)\n\ndef read_srt(path: str):\n with open(path, \"r\", encoding=\"utf-8\") as f:\n content = f.read()\n blocks = SRT_BLOCK.findall(content)\n entries = []\n for idx, ts, txt in blocks:\n text_raw = txt.strip(\"\\r\\n\")\n entries.append((int(idx), ts, text_raw))\n return entries\n\ndef write_srt(path: str, entries):\n with open(path, \"w\", encoding=\"utf-8\") as f:\n for idx, ts, text in entries:\n f.write(f\"{idx}\\n{ts}\\n{text}\\n\\n\")\n\n# -------------------- 对齐切分 --------------------\n\ndef flatten_for_align(text: str) -> str:\n return text.replace(\"\\r\", \"\").replace(\"\\n\", \"\")\n\ndef segment_corrected_by_recognized_boundaries(recognized_full: str,\n corrected_full: str,\n line_lengths: list[int]):\n boundaries = [0]\n acc = 0\n for L in line_lengths:\n acc += L\n boundaries.append(acc)\n\n sm = difflib.SequenceMatcher(None, recognized_full, corrected_full, autojunk=False)\n ops = sm.get_opcodes()\n\n out_lines, buf = [], []\n r_pos, next_bi = 0, 1\n next_boundary = boundaries[next_bi] if next_bi < len(boundaries) else len(recognized_full)\n\n def flush_line():\n nonlocal buf, out_lines, next_bi, next_boundary\n out_lines.append(\"\".join(buf))\n buf = []\n next_bi += 1\n next_boundary = boundaries[next_bi] if next_bi < len(boundaries) else boundaries[-1]\n\n for tag, i1, i2, j1, j2 in ops:\n if tag in (\"equal\", \"replace\"):\n while r_pos < i2:\n take = min(i2 - r_pos, next_boundary - r_pos)\n recog_len_total = i2 - i1\n corr_len_total = j2 - j1\n start_ratio = (r_pos - i1) / max(1, recog_len_total)\n end_ratio = (r_pos + take - i1) / max(1, recog_len_total)\n cj_start = j1 + round(start_ratio * corr_len_total)\n cj_end = j1 + round(end_ratio * corr_len_total)\n if cj_start < cj_end:\n buf.append(corrected_full[cj_start:cj_end])\n r_pos += take\n if r_pos == next_boundary:\n flush_line()\n\n elif tag == \"delete\":\n while r_pos < i2:\n take = min(i2 - r_pos, next_boundary - r_pos)\n r_pos += take\n if r_pos == next_boundary:\n flush_line()\n\n elif tag == \"insert\":\n buf.append(corrected_full[j1:j2])\n\n if buf:\n while len(out_lines) < len(line_lengths) - 1:\n out_lines.append(\"\")\n out_lines.append(\"\".join(buf))\n\n if len(out_lines) < len(line_lengths):\n out_lines += [\"\"] * (len(line_lengths) - len(out_lines))\n elif len(out_lines) > len(line_lengths):\n extra = \"\".join(out_lines[len(line_lengths)-1:])\n out_lines = out_lines[:len(line_lengths)-1] + [extra]\n\n cleaned = []\n for line in out_lines:\n # line = re.sub(r\"\\s+\", \"\", line)\n # line = re.sub(r'^(…{1,2}|\\.{3,}|[，。！？；：、”])+', '', line)\n # line = re.sub(r'(…{1,2}|\\.{3,}|[，。！？；：、“])+$', '', line)\n # 同时匹配中英文符号\n line = re.sub(r\"\\s+\", \"\", line)\n line = re.sub(r'^(…{1,2}|\\.{3,}|[，,。.!！？?；;：:、”“\"“])+', '', line)\n line = re.sub(r'(…{1,2}|\\.{3,}|[，,。.!！？?；;：:、”“\"“])+$', '', line)\n\n cleaned.append(line)\n\n return cleaned\n\n# -------------------- 外部调用 --------------------\n\ndef correct_srt_file(original_text: str, srt_path: str,\n overwrite: bool = True, backup: bool = False,\n out_path: str = None):\n \"\"\"\n original_text: 原始完整文本（直接传字符串）\n srt_path: 输入字幕文件路径\n overwrite: 是否覆盖原文件（默认 True）\n backup: 覆盖时是否先生成 .bak 文件（默认 True）\n out_path: 如果不覆盖，可以指定输出文件路径\n \"\"\"\n original_full = original_text.replace(\"\\r\", \"\").replace(\"\\n\", \"\").strip()\n entries = read_srt(srt_path)\n\n recognized_lines = [flatten_for_align(txt) for _, _, txt in entries]\n recognized_full = \"\".join(recognized_lines)\n\n corrected_full = correct_text_with_pinyin(original_full, recognized_full)\n\n line_lengths = [len(s) for s in recognized_lines]\n corrected_lines = segment_corrected_by_recognized_boundaries(\n recognized_full, corrected_full, line_lengths\n )\n\n corrected_entries = []\n for (idx, ts, _), line_text in zip(entries, corrected_lines):\n corrected_entries.append((idx, ts, line_text))\n\n # 目标路径\n if overwrite:\n if backup:\n shutil.copy(srt_path, srt_path + \".bak\")\n logging.info(\"已生成备份文件：%s.bak\", srt_path)\n target_path = srt_path\n else:\n target_path = out_path or (srt_path + \".corrected.srt\")\n\n write_srt(target_path, corrected_entries)\n logging.info(\"已生成 %s （逐行对齐修正完成）\", target_path)\n\nif __name__ == '__main__':\n generate_subtitle(\"C:\\\\Users\\\\lxc18\\\\SonicVale\\\\1\\\\1\\\\audio\\\\id_2.wav\",\"C:\\\\Users\\\\lxc18\\\\SonicVale\\\\1\\\\1\\\\audio\\\\id_1.srt\")\n\n\n# -------------------- LLM 字幕矫正 --------------------\n\ndef correct_srt_file_with_llm(\n original_text: str,\n srt_path: str,\n llm_engine: LLMEngine,\n batch_size: int = 20,\n overwrite: bool = True,\n backup: bool = False,\n out_path: str = None\n):\n \"\"\"\n 使用LLM进行字幕矫正，分批处理\n \n original_text: 原始完整文本\n srt_path: 输入字幕文件路径\n llm_engine: LLM引擎实例\n batch_size: 每批处理的字幕条数（默认20条）\n overwrite: 是否覆盖原文件\n backup: 覆盖时是否生成.bak文件\n out_path: 如果不覆盖，可以指定输出文件路径\n \"\"\"\n original_full = original_text.replace(\"\\r\", \"\").replace(\"\\n\", \"\").strip()\n entries = read_srt(srt_path)\n \n if not entries:\n logging.warning(\"字幕文件为空：%s\", srt_path)\n return\n \n # 分批处理\n corrected_entries = []\n total_batches = (len(entries) + batch_size - 1) // batch_size\n \n for batch_idx in range(total_batches):\n start_idx = batch_idx * batch_size\n end_idx = min(start_idx + batch_size, len(entries))\n batch_entries = entries[start_idx:end_idx]\n \n logging.info(\"处理字幕批次 %d/%d (第%d-%d条)\", \n batch_idx + 1, total_batches, start_idx + 1, end_idx)\n \n # 准备当前批次的字幕数据\n subtitle_lines = [\n {\"index\": idx, \"text\": txt.replace(\"\\n\", \" \").replace('\"', '\\\\\"')}\n for idx, ts, txt in batch_entries\n ]\n \n # 调用LLM进行矫正\n prompt = get_subtitle_correction_prompt(original_full, subtitle_lines)\n \n try:\n response = llm_engine.generate_text(prompt)\n corrected_batch = llm_engine.save_load_json(response)\n \n # 构建索引映射\n corrected_map = {item[\"index\"]: item[\"corrected_text\"] for item in corrected_batch}\n \n # 处理当前批次的结果\n for idx, ts, original_txt in batch_entries:\n corrected_text = corrected_map.get(idx, original_txt)\n # 清理文本\n corrected_text = clean_subtitle_text(corrected_text)\n corrected_entries.append((idx, ts, corrected_text))\n \n except Exception as e:\n logging.error(\"批次 %d 矫正失败，使用原始文本: %s\", batch_idx + 1, str(e))\n # 失败时保留原始文本\n for idx, ts, txt in batch_entries:\n corrected_entries.append((idx, ts, txt))\n \n # 确定目标路径\n if overwrite:\n if backup:\n shutil.copy(srt_path, srt_path + \".bak\")\n logging.info(\"已生成备份文件：%s.bak\", srt_path)\n target_path = srt_path\n else:\n target_path = out_path or (srt_path + \".corrected.srt\")\n \n write_srt(target_path, corrected_entries)\n logging.info(\"已生成 %s （LLM字幕矫正完成）\", target_path)\n\n\ndef clean_subtitle_text(text: str) -> str:\n \"\"\"清理字幕文本\"\"\"\n # 去除空白字符\n text = re.sub(r\"\\s+\", \"\", text)\n # 清理首尾标点\n text = re.sub(r'^(…{1,2}|\\.{3,}|[，,。.!！？?;；:：、\"“”\"\"])+', '', text)\n text = re.sub(r'(…{1,2}|\\.{3,}|[，,。.!！？?;；:：、\"“”\"\"])+$', '', text)\n return text\n"
},
{
"path": "SonicVale/app/core/text_correct_engine.py",
"content": "import re\nimport json\nimport difflib\nimport logging\nfrom typing import List, Dict, Tuple, Optional\n\n\nclass TextCorrectorFinal:\n # 默认配置参数\n DEFAULT_BASE_THRESHOLD = 0.65 # 基础相似度阈值\n DEFAULT_BASE_WINDOW = 30 # 基础搜索窗口\n DEFAULT_EXTENDED_WINDOW = 80 # 扩展搜索窗口（匹配失败时使用）\n\n def __init__(self, base_threshold: float = None, base_window: int = None):\n \"\"\"初始化文本校正器\n \n Args:\n base_threshold: 基础相似度阈值，默认0.65\n base_window: 基础搜索窗口大小，默认30\n \"\"\"\n self.base_threshold = base_threshold or self.DEFAULT_BASE_THRESHOLD\n self.base_window = base_window or self.DEFAULT_BASE_WINDOW\n self.extended_window = self.DEFAULT_EXTENDED_WINDOW\n\n def clean_text(self, text: str) -> str:\n \"\"\"清理文本用于最终输出，移除换行符和全角空格，保留引号。\"\"\"\n text = re.sub(r'[\\n\\r\\u3000]', '', text)\n text = re.sub(r'\\s+', ' ', text)\n return text.strip()\n\n def clean_for_compare(self, text: str) -> str:\n \"\"\"清理文本用于相似度比较，移除换行符、全角空格和引号。\"\"\"\n text = re.sub(r'[\\n\\r\\u3000]', '', text)\n text = re.sub(r'[\"\"\"「」『』]', '', text) # 仅在比较时移除引号\n text = re.sub(r'\\s+', ' ', text)\n return text.strip()\n\n def get_adaptive_threshold(self, sentence: str) -> float:\n \"\"\"根据句子长度自适应调整相似度阈值。\n \n 短句子容易误匹配，需要更高阈值；长句子可以适当降低阈值。\n \"\"\"\n length = len(self.clean_for_compare(sentence))\n \n if length <= 5:\n # 非常短的句子，需要很高的阈值防止误匹配\n return max(self.base_threshold, 0.85)\n elif length <= 10:\n # 短句子\n return max(self.base_threshold, 0.75)\n elif length <= 20:\n # 中等长度\n return self.base_threshold\n else:\n # 长句子可以适当降低阈值\n return max(self.base_threshold - 0.05, 0.55)\n\n def _looks_like_abbreviation(self, sentence_with_dot: str) -> bool:\n \"\"\"\n 判断当前这一个 '.' 更像是缩写的一部分，而不是句子结束。\n sentence_with_dot: 当前已经累积的句子（包含这个点）\n \"\"\"\n s = sentence_with_dot.rstrip()\n # 找到以 . 结尾的最后一个 token（字母/数字/点）\n m = re.search(r'([A-Za-z0-9\\.]+)\\.$', s)\n if not m:\n return False\n\n token = m.group(1) # 不包含最后这个点，但可能包含内部的 .\n\n # 1) 小写长度很短的缩写，例如 Mr. Dr. etc.\n # 这里简单认为：1~4 个字母，首字母大写\n if re.fullmatch(r'[A-Za-z]{1,4}', token) and token[0].isupper():\n return True\n\n # 2) 多点缩写：U.S.A / F.B.I 这种（至少 3 个字母、2 个点）\n # U.S.A -> token 为 'U.S.A'\n if re.fullmatch(r'[A-Za-z](?:\\.[A-Za-z]){2,}', token):\n return True\n\n # 3) 你如果有特殊缩写，可以在这里硬编码\n # if token in {\"etc\", \"e.g\", \"i.e\"}:\n # return True\n\n return False\n\n def split_sentences(self, text: str) -> List[str]:\n \"\"\"按照标点符号进行细粒度分句，同时尽量保护英文缩写和数字。\n 保留换行作为候选分句符。如果产生了“只有标点/引号”的句子，则直接丢弃。\n \"\"\"\n # 规范化换行：把 \\r\\n 和 \\r 统一为 \\n\n text = text.replace('\\r\\n', '\\n').replace('\\r', '\\n')\n\n # 替换全角空格为普通空格，保留换行\n text = text.replace('\\u3000', ' ').strip()\n\n # 分割：中文标点、特殊点号、或换行\n sentences = re.split(r'([。！？!?：；]|(? Tuple[Optional[int], float]:\n \"\"\"在原文句子列表中找到与AI句子最匹配的单个句子。\n \n Args:\n ai_sentence: AI生成的句子\n original_sentences: 原文句子列表\n start_index: 搜索起始位置\n use_extended: 是否使用扩展搜索窗口\n \n Returns:\n (匹配索引, 相似度) 或 (None, 最高相似度)\n \"\"\"\n # 预处理 - 使用专门的比较清理方法\n processed_ai_sentence = self.clean_for_compare(ai_sentence)\n if not processed_ai_sentence:\n return None, 0\n\n # 根据句子长度获取自适应阈值\n threshold = self.get_adaptive_threshold(ai_sentence)\n \n # 选择搜索窗口大小\n search_window = self.extended_window if use_extended else self.base_window\n\n best_match_index = None\n best_similarity = 0\n\n # 向前搜索\n end_index = min(start_index + search_window, len(original_sentences))\n for i in range(start_index, end_index):\n original_sentence = original_sentences[i]\n processed_original_sentence = self.clean_for_compare(original_sentence)\n\n if not processed_original_sentence:\n continue\n\n matcher = difflib.SequenceMatcher(None, processed_ai_sentence, processed_original_sentence)\n similarity = matcher.ratio()\n\n if similarity > best_similarity:\n best_similarity = similarity\n best_match_index = i\n\n # 如果没找到匹配且未使用扩展窗口，尝试向后搜索（处理乱序情况）\n if best_similarity < threshold and not use_extended and start_index > 0:\n backward_start = max(0, start_index - 10)\n for i in range(backward_start, start_index):\n original_sentence = original_sentences[i]\n processed_original_sentence = self.clean_for_compare(original_sentence)\n\n if not processed_original_sentence:\n continue\n\n matcher = difflib.SequenceMatcher(None, processed_ai_sentence, processed_original_sentence)\n similarity = matcher.ratio()\n\n if similarity > best_similarity:\n best_similarity = similarity\n best_match_index = i\n\n if best_similarity < threshold:\n return None, best_similarity\n\n return best_match_index, best_similarity\n\n def correct_ai_text(self, original_text: str, ai_data: List[Dict]) -> List[Dict]:\n \"\"\"使用分句匹配 + difflib 的方式校正AI文本。\n \n 改进的算法：\n 1. 记录每个校正后item对应的原文句子索引范围\n 2. 基于实际索引位置插入遗漏句子\n 3. 支持自适应搜索窗口\n \"\"\"\n original_sentences = self.split_sentences(original_text)\n\n # 存储校正后的数据，以及每个item对应的原文索引范围\n corrected_data = [] # List of (item_dict, matched_indices_list)\n used_original_indices = set()\n current_original_index = 0\n\n for ai_item in ai_data:\n ai_text = ai_item.get('text_content', '')\n ai_sentences = self.split_sentences(ai_text)\n\n corrected_sentences_for_item = []\n matched_indices_for_item = [] # 记录这个item匹配到的所有原文索引\n\n logging.info(\"处理角色: %s (AI原文: '%s')\", ai_item.get('role_name', '未知'), ai_text[:50] if ai_text else '')\n\n for ai_sentence in ai_sentences:\n # 首先尝试基础窗口搜索\n match_index, similarity = self.find_best_sentence_match(\n ai_sentence, original_sentences, current_original_index, use_extended=False\n )\n\n # 如果基础窗口没找到，尝试扩展窗口\n if match_index is None:\n match_index, similarity = self.find_best_sentence_match(\n ai_sentence, original_sentences, current_original_index, use_extended=True\n )\n\n if match_index is not None:\n original_match = original_sentences[match_index]\n corrected_sentences_for_item.append(original_match)\n matched_indices_for_item.append(match_index)\n used_original_indices.add(match_index)\n current_original_index = match_index + 1\n logging.info(\"匹配成功 (相似度: %.2f): AI='%s' -> 原文='%s'\", similarity, ai_sentence, original_match)\n else:\n corrected_sentences_for_item.append(ai_sentence)\n logging.warning(\"匹配失败 (最高相似度: %.2f)，保留AI原句: '%s'\", similarity, ai_sentence)\n\n # 最终清理 - 保留原始格式（包括引号）\n corrected_text = self.clean_text(\" \".join(corrected_sentences_for_item))\n\n if corrected_text:\n corrected_item = ai_item.copy()\n corrected_item['text_content'] = corrected_text\n corrected_data.append((corrected_item, matched_indices_for_item))\n\n # 处理遗漏的原文句子 - 改进的插入逻辑\n missing_indices = set(range(len(original_sentences))) - used_original_indices\n \n if not missing_indices:\n # 没有遗漏，直接返回校正数据\n return [item for item, _ in corrected_data]\n\n logging.info(\"发现 %d 个遗漏句子，正在插入...\", len(missing_indices))\n\n # 构建原文索引到校正item的映射\n # index_to_item_map: {原文索引: (corrected_item, item在corrected_data中的位置)}\n index_to_item = {}\n for item_idx, (item, matched_indices) in enumerate(corrected_data):\n for orig_idx in matched_indices:\n index_to_item[orig_idx] = (item, item_idx)\n\n # 按原文顺序构建最终结果\n final_data = []\n inserted_items = set() # 记录已插入的item索引，避免重复插入\n\n for orig_idx in range(len(original_sentences)):\n if orig_idx in missing_indices:\n # 插入遗漏的句子\n missing_sentence = self.clean_text(original_sentences[orig_idx])\n if missing_sentence:\n logging.info(\"插入遗漏句子 (位置%d): '%s'\", orig_idx, missing_sentence)\n final_data.append({\n 'role_name': '旁白',\n 'text_content': missing_sentence,\n 'emotion_name': '',\n 'strength_name': ''\n })\n elif orig_idx in index_to_item:\n item, item_idx = index_to_item[orig_idx]\n # 只在第一次遇到这个item的匹配索引时插入\n if item_idx not in inserted_items:\n final_data.append(item)\n inserted_items.add(item_idx)\n\n # 处理可能没有匹配到任何原文索引但仍需要保留的item（纯AI生成内容）\n for item_idx, (item, matched_indices) in enumerate(corrected_data):\n if item_idx not in inserted_items:\n final_data.append(item)\n logging.warning(\"Item未匹配到原文，追加到末尾: %s\", item.get('text_content', '')[:30])\n\n return final_data\n\n\ndef read_files():\n \"\"\"读取原文和AI输出文件\"\"\"\n try:\n with open('原文3.txt', 'r', encoding='utf-8') as f:\n original_text = f.read()\n with open('AI输出的包含错误的文本3.json', 'r', encoding='utf-8') as f:\n ai_data = json.load(f)\n return original_text, ai_data\n except FileNotFoundError as e:\n logging.error(\"文件读取错误: %s\", e)\n return None, None\n except json.JSONDecodeError as e:\n logging.error(\"JSON解析错误: %s\", e)\n return None, None\n\n\ndef save_corrected_data(corrected_data: List[Dict]):\n \"\"\"保存校正后的数据\"\"\"\n try:\n with open('校正后的文本_final.json', 'w', encoding='utf-8') as f:\n json.dump(corrected_data, f, ensure_ascii=False, indent=4)\n logging.info(\"校正结果已保存到: 校正后的文本_final.json\")\n except Exception as e:\n logging.error(\"保存文件时出错: %s\", e)\n\n\ndef main():\n original_text, ai_data = read_files()\n if original_text is None or ai_data is None:\n return\n\n logging.info(\"文件读取成功！开始校正...\")\n\n corrector = TextCorrectorFinal()\n corrected_data = corrector.correct_ai_text(original_text, ai_data)\n\n save_corrected_data(corrected_data)\n\n logging.info(\"校正完成！\")\n\n\nif __name__ == \"__main__\":\n main()\n"
},
{
"path": "SonicVale/app/core/tts_engine.py",
"content": "import requests\nfrom typing import Optional, List\nimport os\nimport logging\n\nclass TTSEngine:\n def __init__(self, base_url: str):\n \"\"\"\n 初始化 TTS 引擎\n :param base_url: TTS 服务的基础 URL，如 http://127.0.0.1:8000\n \"\"\"\n self.base_url = base_url.rstrip(\"/\")\n\n def synthesize(\n self,\n text: str,\n filename: str,\n emo_text: Optional[str] = None,\n emo_vector: Optional[List[float]] = None,\n save_path: Optional[str] = None\n ) -> bytes:\n \"\"\"\n 调用 /v2/synthesize 接口进行语音合成\n :param text: 要合成的文本\n :param filename: 参考音频文件名（服务端已存在）\n :param emo_text: 情绪文本（可选）\n :param emo_vector: 8维情绪向量（可选，优先级高于 emo_text）\n :param save_path: 如果指定，将保存生成的音频文件到本地\n :return: 音频二进制数据\n \"\"\"\n url = f\"{self.base_url}/v2/synthesize\"\n\n payload = {\"text\": text, \"audio_path\": filename}\n\n if emo_vector is not None:\n payload[\"emo_vector\"] = emo_vector\n elif emo_text:\n payload[\"emo_text\"] = emo_text\n\n try:\n resp = requests.post(url, json=payload, timeout=120)\n if resp.status_code != 200:\n # 尝试解析错误信息\n try:\n error_data = resp.json()\n error_msg = error_data.get('detail') or error_data.get('message') or error_data.get('msg') or resp.text\n except:\n error_msg = resp.text\n raise Exception(f\"TTS服务返回错误({resp.status_code}): {error_msg}\")\n\n audio_bytes = resp.content\n \n # 检查返回的内容是否为有效音频\n if len(audio_bytes) < 100:\n raise Exception(f\"TTS服务返回的音频数据无效，大小: {len(audio_bytes)} 字节\")\n\n if save_path:\n with open(save_path, \"wb\") as f:\n f.write(audio_bytes)\n\n return audio_bytes\n \n except requests.exceptions.ConnectionError:\n raise Exception(f\"TTS服务连接失败，请检查TTS服务是否已启动 ({self.base_url})\")\n except requests.exceptions.Timeout:\n raise Exception(f\"TTS服务请求超时，请检查TTS服务是否正常运行\")\n except requests.exceptions.RequestException as e:\n raise Exception(f\"TTS服务请求异常: {str(e)}\")\n\n def get_models(self) -> dict:\n \"\"\"\n 调用 /v1/models 获取模型列表\n :return: 模型信息\n \"\"\"\n url = f\"{self.base_url}/v1/models\"\n resp = requests.get(url)\n resp.raise_for_status()\n return resp.json()\n\n def check_audio_exists(self, filename: str) -> bool:\n \"\"\"\n 调用 /v1/check/audio 检查参考音频是否存在\n :param filename: 原始文件名\n :return: True or False\n \"\"\"\n url = f\"{self.base_url}/v1/check/audio\"\n params = {\"file_name\": filename}\n resp = requests.get(url, params=params)\n resp.raise_for_status()\n return resp.json().get(\"exists\", False)\n\n def upload_audio(self, file_path: str,full_path=None) -> dict:\n \"\"\"\n 调用 /v1/upload_audio 上传音频\n :param file_path: 本地音频文件路径\n :param full_path: 用于唯一标识的全路径（可选，如果不传则使用 file_path）\n :return: 服务端响应 JSON\n \"\"\"\n if not os.path.isfile(file_path):\n return {\"code\": 400, \"msg\": f\"文件不存在: {file_path}\"}\n\n url = f\"{self.base_url}/v1/upload_audio\"\n try:\n with open(file_path, \"rb\") as f:\n files = {\n \"audio\": (os.path.basename(file_path), f, \"audio/wav\")\n }\n # 如果需要额外传 fullpath 参数\n data = {}\n if full_path:\n data[\"full_path\"] = full_path\n\n resp = requests.post(url, files=files, data=data, timeout=30)\n resp.raise_for_status()\n return resp.json()\n except requests.exceptions.RequestException as e:\n return {\"code\": 500, \"msg\": f\"请求失败: {str(e)}\"}\n except Exception as e:\n return {\"code\": 500, \"msg\": f\"上传异常: {str(e)}\"}\nif __name__ == \"__main__\":\n # 示例使用\n engine = TTSEngine(\"https://eihh5fmon4-8200.cnb.run/\")\n\n # 1. 上传音频\n upload_res = engine.upload_audio(\"C:\\\\Users\\\\lxc18\\\\Music\\\\多情绪\\\\吴泽\\\\解说\\\\中等.wav\",full_path=\"C:\\\\Users\\\\lxc18\\\\Music\\\\多情绪\\\\吴泽\\\\解说\\\\中等.wav\")\n # print(\"上传结果:\", upload_res)\n\n # 2. 检查音频是否存在\n exists = engine.check_audio_exists(\"C:\\\\Users\\\\lxc18\\\\Music\\\\多情绪\\\\吴泽\\\\解说\\\\中等.wav\")\n logging.info(\"音频存在: %s\", exists)\n\n # 3. 获取模型列表\n models = engine.get_models()\n logging.info(\"模型信息: %s\", models)\n\n # 4. 合成语音\n if exists:\n audio = engine.synthesize(\"萧炎，斗之力，三段！级别：低级！\", \"C:\\\\Users\\\\lxc18\\\\Music\\\\多情绪\\\\吴泽\\\\解说\\\\中等.wav\",emo_text=\"愤怒\", save_path=\"output.wav\")\n logging.info(\"语音已保存到 output.wav, 大小 %s 字节\", len(audio))\n"
},
{
"path": "SonicVale/app/core/tts_runtime.py",
"content": "# app/tts_worker.py\nimport asyncio\nfrom fastapi import FastAPI\nfrom markdown_it.rules_block import reference\n\nfrom app.core.ws_manager import manager\nfrom app.db.database import SessionLocal\nfrom app.routers.chapter_router import get_voice_service, get_emotion_service, get_strength_service\nfrom app.routers.multi_emotion_voice_router import get_multi_emotion_voice_service\nfrom app.routers.role_router import get_line_service, get_role_service, get_project_service\n\nTTS_TIMEOUT_SECONDS = 1200 # 可调\ndef emotion_text_to_vector(emotion: str, intensity: str) -> list[float]:\n \"\"\"\n 将情绪(文本) + 强度(文本) 转换成 8维向量\n 8维分别对应: [高兴, 生气, 伤心, 害怕, 厌恶, 低落, 惊喜, 平静]\n 基础情绪为 one-hot，复合情绪为多维加权混合\n :param emotion: 情绪名称\n :param intensity: \"微弱\" / \"稍弱\" / \"中等\" / \"较强\" / \"强烈\"\n :return: 长度为8的向量\n \"\"\"\n # 8维基础情绪索引: 高兴=0, 生气=1, 伤心=2, 害怕=3, 厌恶=4, 低落=5, 惊喜=6, 平静=7\n BASE_EMOTIONS = [\"高兴\", \"生气\", \"伤心\", \"害怕\", \"厌恶\", \"低落\", \"惊喜\", \"平静\"]\n\n # 复合情绪 → 基础情绪权重（各维度满强度，由 intensity 统一缩放）\n COMPOSITE_MAP = {\n \"嘲讽\": {\"高兴\": 0.5, \"厌恶\": 1.0}, # 讽刺语气\n \"悲愤\": {\"伤心\": 1.0, \"生气\": 1.0}, # 悲愤交加\n }\n\n INTENSITY_MAP = {\n \"微弱\": 0.2,\n \"稍弱\": 0.4,\n \"中等\": 0.6,\n \"较强\": 0.8,\n \"强烈\": 1.0\n }\n\n scale = INTENSITY_MAP.get(intensity, 0.5)\n vec = [0.0] * 8\n\n if emotion in BASE_EMOTIONS:\n # 基础情绪: one-hot\n vec[BASE_EMOTIONS.index(emotion)] = scale\n elif emotion in COMPOSITE_MAP:\n # 复合情绪: 多维加权混合\n for base_name, weight in COMPOSITE_MAP[emotion].items():\n vec[BASE_EMOTIONS.index(base_name)] = round(scale * weight, 4)\n # 未知情绪返回全零向量（静默降级）\n return vec\nasync def tts_worker(app: FastAPI):\n q = app.state.tts_queue\n ex = app.state.tts_executor\n while True:\n project_id, dto = await q.get()\n db = SessionLocal()\n try:\n line_service = get_line_service(db)\n role_service = get_role_service(db)\n voice_service = get_voice_service(db)\n multi_emotion_service = get_multi_emotion_voice_service(db)\n project_service = get_project_service(db)\n emotion_service = get_emotion_service(db)\n strength_service = get_strength_service(db)\n\n\n # line_service.update_line(dto.id, {\"status\": \"processing\"})\n await manager.broadcast({\n \"event\": \"line_update\",\n \"line_id\": dto.id,\n \"status\": \"processing\",\n \"progress\": q.qsize() + 1, # +1 包含当前正在处理的任务\n \"meta\": f\"角色 {dto.role_id} 开始生成\"\n })\n\n role = role_service.get_role(dto.role_id)\n voice = voice_service.get_voice(role.default_voice_id)\n reference_path = voice.reference_path\n\n\n # if voice.is_multi_emotion == 1:\n # # 使用多音色\n # multi_emotion = multi_emotion_service.get_multi_emotion_voice_by_voice_id_emotion_id_strength_id(voice.id, dto.emotion_id, dto.strength_id)\n # if multi_emotion is not None:\n # reference_path = multi_emotion.reference_path\n\n # 9.13\n emotion = emotion_service.get_emotion(dto.emotion_id)\n strength = strength_service.get_strength(dto.strength_id)\n # 拼接\n # emo_text = f\"{strength.name}的{emotion.name} \"\n # if emotion.name is \"解说\":\n # emo_text = None\n emo_text = None\n emo_vector = emotion_text_to_vector(emotion.name, strength.name)\n\n project = project_service.get_project(project_id)\n\n loop = asyncio.get_running_loop()\n await asyncio.wait_for(\n loop.run_in_executor(\n ex,\n line_service.generate_audio,\n reference_path,\n project.tts_provider_id,\n dto.text_content,\n emo_text,\n emo_vector,\n dto.audio_path\n ),\n timeout=TTS_TIMEOUT_SECONDS\n )\n\n line_service.update_line(dto.id, {\"status\": \"done\"})\n await manager.broadcast({\n \"event\": \"line_update\",\n \"line_id\": dto.id,\n \"status\": \"done\",\n \"progress\": q.qsize(),\n \"meta\": \"生成完成\",\n \"audio_path\": dto.audio_path\n })\n # 发送给前端，队列中剩余的数量\n await manager.broadcast({\n \"event\": \"tts_queue_rest\",\n \"queue_rest\": q.qsize(),\n \"project_id\": project_id\n })\n\n except Exception as e:\n try:\n line_service.update_line(dto.id, {\"status\": \"failed\"})\n except Exception:\n pass\n await manager.broadcast({\n \"event\": \"line_update\",\n \"line_id\": dto.id,\n \"status\": \"failed\",\n \"progress\": q.qsize(),\n \"meta\": f\"失败: {e}\"\n })\n\n finally:\n\n db.close()\n q.task_done()\n"
},
{
"path": "SonicVale/app/core/ws_manager.py",
"content": "# ws_manager.py\nfrom fastapi import WebSocket\nfrom typing import List\n\nclass WSManager:\n def __init__(self):\n self.conns: List[WebSocket] = []\n\n async def connect(self, ws: WebSocket):\n await ws.accept()\n self.conns.append(ws)\n\n def disconnect(self, ws: WebSocket):\n if ws in self.conns:\n self.conns.remove(ws)\n\n async def broadcast(self, data: dict):\n dead = []\n for ws in self.conns:\n try:\n await ws.send_json(data)\n except:\n dead.append(ws)\n for d in dead:\n self.disconnect(d)\n\nmanager = WSManager()\n"
},
{
"path": "SonicVale/app/db/database.py",
"content": "from typing import Any, Generator\n\nfrom sqlalchemy import create_engine\nfrom sqlalchemy.orm import sessionmaker, declarative_base, Session\nfrom app.core.config import *\n\nconfig_path = getConfigPath()\n# SQLite 数据库文件，存储在用户目录下的 SonicVale\nSQLALCHEMY_DATABASE_URL = f\"sqlite:///{os.path.join(config_path, 'app_test.db')}\"\n\n\n\n# echo=True 会打印执行的 SQL 语句，调试用\nengine = create_engine(\n SQLALCHEMY_DATABASE_URL, connect_args={\"check_same_thread\": False}, echo=False\n)\n\n# SessionLocal 用于依赖注入\nSessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)\n\n# Base 类，所有 ORM 模型继承它\nBase = declarative_base()\n\n# 依赖函数\ndef get_db() -> Generator[Session, Any, None]:\n db = SessionLocal()\n try:\n yield db\n finally:\n db.close()"
},
{
"path": "SonicVale/app/dto/chapter_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass ChapterCreateDTO(BaseModel):\n title: str\n project_id: int\n order_index: Optional[int] = None\n id: Optional[int] = None\n text_content : Optional[str] = None\n\nclass ChapterResponseDTO(BaseModel):\n title: str\n project_id: int\n order_index: Optional[int] = None\n id: Optional[int] = None\n text_content: Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n"
},
{
"path": "SonicVale/app/dto/emotion_dto.py",
"content": "\nfrom datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass EmotionCreateDTO(BaseModel):\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n\n\nclass EmotionResponseDTO(BaseModel):\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n\n"
},
{
"path": "SonicVale/app/dto/line_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\nclass LineInitDTO(BaseModel):\n role_name: Optional[str] = None\n text_content: str\n emotion_name: Optional[str] = None\n strength_name: Optional[str] = None\n\n\nclass LineOrderDTO(BaseModel):\n id: int\n line_order: int\nclass LineAudioProcessDTO(BaseModel):\n # 默认是1\n speed: Optional[float] = 1.0\n # 默认是1\n volume: Optional[float] = 1.0\n start_ms: Optional[int] = None\n end_ms: Optional[int] = None\n# 静止时间\n silence_sec: Optional[float] = 0.0\n current_ms: Optional[int] = None\n\nclass LineCreateDTO(BaseModel):\n chapter_id: int\n role_id:Optional[int] = None\n voice_id : Optional[int] = None\n line_order: Optional[int] = None\n id: Optional[int] = None\n text_content: Optional[str] = None\n\n emotion_id: Optional[int] = None\n strength_id: Optional[int] = None\n\n audio_path : Optional[str] = None\n status : Optional[str] = None\n is_done : Optional[int] = 0\n subtitle_path : Optional[str] = None\n\nclass LineResponseDTO(BaseModel):\n chapter_id: int\n role_id:Optional[int] = None\n voice_id : Optional[int] = None\n line_order: Optional[int] = None\n id: Optional[int] = None\n text_content: Optional[str] = None\n\n emotion_id: Optional[int] = None\n strength_id: Optional[int] = None\n\n audio_path : Optional[str] = None\n status : Optional[str] = None\n is_done: Optional[int] = 0\n subtitle_path : Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n\n"
},
{
"path": "SonicVale/app/dto/llm_provider_dto.py",
"content": "from dataclasses import Field\nfrom datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional, Dict, Any\n\n\nfrom pydantic import BaseModel, Field as PydField\n\n\nclass LLMProviderCreateDTO(BaseModel):\n \"\"\"业务实体：LLM\"\"\"\n name: str\n id: Optional[int] = None\n api_base_url : Optional[str] = None\n api_key: Optional[str] = None\n model_list: Optional[str] = None\n status : Optional[int] = None\n\n # ✅ 默认自定义参数\n custom_params: Optional[str] = None\n\n\nclass LLMProviderResponseDTO(BaseModel):\n \"\"\"业务实体：LLM\"\"\"\n name: str\n id: Optional[int] = None\n api_base_url : Optional[str] = None\n api_key: Optional[str] = None\n model_list: Optional[str] = None\n status : Optional[int] = None\n updated_at: Optional[datetime] = None\n created_at: Optional[datetime] = None\n\n # ✅ 默认自定义参数\n custom_params: Optional[str] = None"
},
{
"path": "SonicVale/app/dto/multi_emotion_voice_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass MultiEmotionVoiceCreateDTO(BaseModel):\n emotion_id: int\n voice_id: int\n strength_id: int\n id: Optional[int] = None\n reference_path: Optional[str] = None\n\n\n\nclass MultiEmotionVoiceResponseDTO(BaseModel):\n emotion_id: int\n voice_id: int\n strength_id: int\n id: Optional[int] = None\n reference_path: Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/dto/project_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass ProjectCreateDTO(BaseModel):\n name: str\n description: Optional[str] = None\n llm_provider_id: Optional[int] = None\n llm_model: Optional[str] = None\n tts_provider_id: Optional[int] = None\n prompt_id: Optional[int] = None\n # 精准填充\n is_precise_fill: Optional[int] = None\n # 项目路径\n project_root_path : Optional[str] = None\n\nclass ProjectResponseDTO(BaseModel):\n id: int\n name: str\n description: Optional[str] = None\n llm_provider_id: Optional[int] = None\n llm_model: Optional[str] = None\n tts_provider_id: Optional[int] = None\n prompt_id: Optional[int] = None\n # 精准填充\n is_precise_fill : Optional[int] = None\n # 项目路径\n project_root_path : Optional[str] = None\n created_at: datetime\n updated_at: datetime\n\n\nclass ProjectImportDTO(BaseModel):\n id : int\n content: str\n"
},
{
"path": "SonicVale/app/dto/prompt_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass PromptCreateDTO(BaseModel):\n\n \"\"\"业务实体：提示词\"\"\"\n name: str\n task: str\n description: Optional[str] = None\n content: Optional[str] = None\n id: Optional[int] = None\n\n\nclass PromptResponseDTO(BaseModel):\n\n \"\"\"业务实体：提示词\"\"\"\n name: str\n task: str\n description: Optional[str] = None\n content: Optional[str] = None\n id: Optional[int] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None"
},
{
"path": "SonicVale/app/dto/role_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass RoleCreateDTO(BaseModel):\n name: str\n project_id: int\n id: Optional[int] = None\n default_voice_id: Optional[int] = None\n\nclass RoleResponseDTO(BaseModel):\n name: str\n project_id: int\n id: Optional[int] = None\n default_voice_id: Optional[int] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/dto/strength_dto.py",
"content": "\nfrom datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\n\nclass StrengthCreateDTO(BaseModel):\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n\n\n\nclass StrengthResponseDTO(BaseModel):\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None"
},
{
"path": "SonicVale/app/dto/tts_provider_dto.py",
"content": "from datetime import datetime\n\nfrom pydantic import BaseModel\nfrom typing import Optional\n\n\nclass TTSProviderCreateDTO(BaseModel):\n name: Optional[str] = None\n id: Optional[int] = None\n api_base_url: Optional[str] = None\n api_key: Optional[str] = None\n status: Optional[int] = None\n\n\n\nclass TTSProviderResponseDTO(BaseModel):\n \"\"\"业务实体：tts_provider\"\"\"\n name: str\n id: Optional[int] = None\n api_base_url : Optional[str] = None\n api_key: Optional[str] = None\n status : Optional[int] = None\n updated_at: Optional[datetime] = None\n created_at: Optional[datetime] = None"
},
{
"path": "SonicVale/app/dto/voice_dto.py",
"content": "from datetime import datetime\nfrom typing import Optional, List\n\nfrom pydantic import BaseModel, Field, AliasChoices\n\n\nclass VoiceCreateDTO(BaseModel):\n name: str\n tts_provider_id: int\n id: Optional[int] = None\n reference_path: Optional[str] = None\n description: Optional[str] = None\n is_multi_emotion: Optional[int] = 0\n\n\nclass VoiceResponseDTO(BaseModel):\n name: str\n tts_provider_id: int\n id: Optional[int] = None\n reference_path : Optional[str] = None\n description : Optional[str] = None\n is_multi_emotion : Optional[int] = 0\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n\nclass VoiceExportDTO(BaseModel):\n \"\"\"导出音色库请求DTO\"\"\"\n tts_provider_id: int\n export_path: str\n ids: Optional[List[int]] = Field(default=None, validation_alias=AliasChoices(\"ids\", \"voice_ids\"))\n\n\nclass VoiceImportDTO(BaseModel):\n \"\"\"导入音色库请求DTO\"\"\"\n tts_provider_id: int\n zip_path: str\n target_dir: str\n\n\nclass VoiceImportResultDTO(BaseModel):\n \"\"\"导入音色库结果DTO\"\"\"\n success_count: int\n skipped_count: int\n skipped_names: List[str]\n\n\nclass VoiceAudioProcessDTO(BaseModel):\n \"\"\"音色参考音频处理DTO\"\"\"\n audio_path: str\n speed: Optional[float] = 1.0\n volume: Optional[float] = 1.0\n start_ms: Optional[int] = None\n end_ms: Optional[int] = None\n silence_sec: Optional[float] = 0.0\n current_ms: Optional[int] = None\n\n\nclass VoiceCopyDTO(BaseModel):\n \"\"\"复制音色请求DTO\"\"\"\n source_voice_id: int\n new_name: str\n target_dir: Optional[str] = None # 为空则使用原音色同目录\n\n"
},
{
"path": "SonicVale/app/entity/chapter_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass ChapterEntity:\n \"\"\"业务实体：章节\"\"\"\n title: str\n project_id: int\n order_index: Optional[int] = None\n id: Optional[int] = None\n text_content : Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n\n"
},
{
"path": "SonicVale/app/entity/emotion_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n@dataclass\nclass EmotionEntity:\n \"\"\"业务实体：情绪枚举\"\"\"\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/entity/line_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass LineEntity:\n \"\"\"业务实体：台词\"\"\"\n chapter_id: int\n id: Optional[int] = None\n role_id : Optional[ int] = None\n voice_id : Optional[int] = None\n line_order : Optional[int] = None\n text_content : Optional[str] = None\n\n emotion_id : Optional[int] = None\n strength_id : Optional[int] = None\n\n audio_path : Optional[str] = None\n subtitle_path : Optional[str] = None\n status : Optional[str] = None\n # 是否完成\n is_done : Optional[int] = 0\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n"
},
{
"path": "SonicVale/app/entity/llm_provider_entity.py",
"content": "\nfrom dataclasses import dataclass, field\nfrom datetime import datetime\nfrom typing import Optional, Dict, Any\n\n\n@dataclass\nclass LLMProviderEntity:\n \"\"\"业务实体：LLM\"\"\"\n name: str\n id: Optional[int] = None\n api_base_url : Optional[str] = None\n api_key: Optional[str] = None\n model_list : Optional[str] = None\n status : Optional[int] = None\n updated_at: Optional[datetime] = None\n created_at: Optional[datetime] = None\n\n # ✅ 自定义参数字段（默认值与数据库一致）\n custom_params: Optional[str] = None\n"
},
{
"path": "SonicVale/app/entity/multi_emotion_voice_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n# class MultiEmotionVoicePO(Base):\n# __tablename__ = \"multi_emotion\"\n# id = Column(Integer, primary_key=True, autoincrement=True, index=True)\n# emotion_id = Column(Integer, nullable=False)\n# voice_id = Column(Integer, nullable=False)\n# strength_id = Column(Integer, nullable=True)\n# reference_path = Column(String(255), nullable=True)\n# created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n# updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n# nullable=False)\n\n@dataclass\nclass MultiEmotionVoiceEntity:\n \"\"\"业务实体：多情感音色\"\"\"\n\n emotion_id: int\n voice_id: int\n strength_id: int\n id: Optional[int] = None\n reference_path: Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/entity/project_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass ProjectEntity:\n \"\"\"业务实体：项目\"\"\"\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n llm_provider_id: Optional[int] = None\n llm_model: Optional[str] = None\n tts_provider_id: Optional[int] = None\n prompt_id: Optional[int] = None # 提示词\n # 精准填充\n is_precise_fill: Optional[int] = None\n # 项目保存地址\n project_root_path: Optional[str] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n\n\n\n\n"
},
{
"path": "SonicVale/app/entity/prompt_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n# class PromptPO(Base):\n# __tablename__ = \"prompt\"\n# id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n# name = Column(String(255), nullable=False)\n# description = Column(Text, nullable=True)\n# content = Column(Text, nullable=True)\n# created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n# updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),nullable=False)\n\n@dataclass\nclass PromptEntity:\n \"\"\"业务实体：提示词\"\"\"\n name: str\n task: str\n description: Optional[str] = None\n content: Optional[str] = None\n id: Optional[int] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n"
},
{
"path": "SonicVale/app/entity/role_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass RoleEntity:\n \"\"\"业务实体：角色\"\"\"\n name: str\n project_id: int\n id: Optional[int] = None\n default_voice_id : Optional[int] = None\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/entity/strength_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n \n@dataclass\nclass StrengthEntity:\n \"\"\"业务实体：情绪强弱枚举\"\"\"\n name: str\n id: Optional[int] = None\n description: Optional[str] = None\n is_active: Optional[int] = 1\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n"
},
{
"path": "SonicVale/app/entity/tts_provider_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass TTSProviderEntity:\n \"\"\"业务实体：TTS\"\"\"\n name: str\n id: Optional[int] = None\n api_base_url : Optional[str] = None\n api_key: Optional[str] = None\n status : Optional[int] = None\n updated_at: Optional[datetime] = None\n created_at: Optional[datetime] = None\n\n"
},
{
"path": "SonicVale/app/entity/voice_entity.py",
"content": "\nfrom dataclasses import dataclass\nfrom datetime import datetime\nfrom typing import Optional\n\n\n\n@dataclass\nclass VoiceEntity:\n \"\"\"业务实体：音色\"\"\"\n name: str\n tts_provider_id: int\n id: Optional[int] = None\n reference_path : Optional[str] = None\n description : Optional[str] = None\n is_multi_emotion : Optional[int] = 0\n created_at: Optional[datetime] = None\n updated_at: Optional[datetime] = None\n"
},
{
"path": "SonicVale/app/main.py",
"content": "# app/main.py\nimport asyncio\nimport logging\nfrom concurrent.futures import ThreadPoolExecutor\n\nimport uvicorn\nfrom fastapi import FastAPI, Depends\nfrom sqlalchemy.orm import Session\nfrom starlette.middleware.cors import CORSMiddleware\n\nfrom app.core.config import getConfigPath\nfrom app.core.prompts import get_prompt_str\nfrom app.core.tts_runtime import tts_worker\nfrom app.core.ws_manager import manager\nfrom app.db.database import Base, engine, SessionLocal, get_db\nfrom app.entity.emotion_entity import EmotionEntity\nfrom app.entity.strength_entity import StrengthEntity\nfrom app.models.po import *\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.routers import project_router, chapter_router, role_router, voice_router, llm_provider_router, \\\n tts_provider_router, line_router, emotion_router, strength_router, multi_emotion_voice_router, prompt_router\nfrom app.routers.chapter_router import get_strength_service, get_prompt_service, get_project_service\nfrom app.routers.emotion_router import get_emotion_service\nfrom app.routers.llm_provider_router import get_llm_service\nfrom app.services.llm_provider_service import LLMProviderService\n\nfrom app.services.tts_provider_service import TTSProviderService\n\nimport os\nimport sys\n\nroot_path = os.getcwd()\nsys.path.append(root_path)\n\n# =========================\n# 日志配置（同时输出到控制台和文件）\n# =========================\nlog_file_path = os.path.join(getConfigPath(), \"app.log\")\nlogging.basicConfig(\n level=logging.INFO,\n format='%(asctime)s - %(levelname)s - %(message)s',\n handlers=[\n logging.StreamHandler(), # 控制台输出\n logging.FileHandler(log_file_path, encoding='utf-8') # 文件输出\n ]\n)\nlogging.info(f\"日志文件路径: {log_file_path}\")\n\n# =========================\n# FastAPI 实例\n# =========================\napp = FastAPI(\n title=\"音墟 (YinXu) - AI多角色小说配音\",\n description=\"桌面端小说多角色配音系统，支持 TTS、GPT 提取角色、台词管理及字幕生成\",\n version=\"1.0.0\",\n)\n# 跨域\n# 允许的前端地址\norigins = [\n \"http://localhost:5173\", # Vue 开发服务器\n \"http://127.0.0.1:5173\" # 有些浏览器可能会用这个\n]\n\napp.add_middleware(\n CORSMiddleware,\n allow_origins=origins, # 允许的源\n allow_credentials=True,\n allow_methods=[\"*\"], # 允许所有方法（GET, POST, DELETE...）\n allow_headers=[\"*\"], # 允许所有请求头\n)\n\n\n\n# =========================\n# 数据库初始化（创建表）\n# =========================\n\n# 启动时创建表\n# @app.on_event(\"startup\")\n# def startup():\n# Base.metadata.create_all(bind=engine)\n\nWORKERS = 1\nQUEUE_CAPACITY = 0\n\nfrom sqlalchemy import text\n\ndef add_prompt_id_column():\n with engine.connect() as conn:\n # 检查 project 表是否已有 prompt_id\n result = conn.execute(text(\"PRAGMA table_info(projects)\"))\n columns = [row[1] for row in result.fetchall()]\n if \"prompt_id\" not in columns:\n conn.execute(text(\"ALTER TABLE projects ADD COLUMN prompt_id INTEGER\"))\n conn.commit()\n\n# 添加line表中is_done字段\ndef add_is_done_column():\n with engine.connect() as conn:\n result = conn.execute(text(\"PRAGMA table_info(lines)\"))\n columns = [row[1] for row in result.fetchall()]\n if \"is_done\" not in columns:\n # ✅ 添加列并设置默认值 0\n conn.execute(text(\"ALTER TABLE lines ADD COLUMN is_done INTEGER DEFAULT 0\"))\n conn.commit()\n\n# 添加LLM自定义参数字段\ndef add_custom_params_column():\n with engine.begin() as conn: # ✅ 用 begin() 自动提交事务\n result = conn.execute(text(\"PRAGMA table_info(llm_provider)\"))\n columns = [row[1] for row in result.fetchall()]\n if \"custom_params\" not in columns:\n # ✅ 添加列\n conn.execute(text(\"ALTER TABLE llm_provider ADD COLUMN custom_params TEXT\"))\n\n # ✅ 可选：为历史数据填入默认 JSON（推荐）\n import json\n default_json = json.dumps({\n \"response_format\": {\"type\": \"json_object\"},\n \"temperature\": 0.7,\n \"top_p\": 0.9\n }, ensure_ascii=False)\n conn.execute(\n text(\"UPDATE llm_provider SET custom_params = :val\"),\n {\"val\": default_json}\n )\n\n logging.info(\"已添加 custom_params 列并写入默认值。\")\n else:\n logging.info(\"custom_params 列已存在，跳过。\")\n\n# 添加精准填充字段】\ndef add_is_precise_fill_column():\n with engine.begin() as conn: # ✅ 用 begin() 自动提交事务\n result = conn.execute(text(\"PRAGMA table_info(projects)\"))\n columns = [row[1] for row in result.fetchall()]\n if \"is_precise_fill\" not in columns:\n # ✅ 添加列\n conn.execute(text(\"ALTER TABLE projects ADD COLUMN is_precise_fill INTEGER DEFAULT 0\"))\n\n conn.commit()\n\n# 添加项目保存路径字段（project_path）\ndef add_project_root_path_column():\n with engine.begin() as conn: # ✅ 用 begin() 自动提交事务\n result = conn.execute(text(\"PRAGMA table_info(projects)\"))\n columns = [row[1] for row in result.fetchall()]\n if \"project_root_path\" not in columns:\n # ✅ 添加列\n conn.execute(text(\"ALTER TABLE projects ADD COLUMN project_root_path TEXT\"))\n\n conn.commit()\n\ndef get_tts_service(db: Session = Depends(get_db)) -> TTSProviderService:\n return TTSProviderService(TTSProviderRepository(db))\n\n@app.on_event(\"startup\")\nasync def startup_event():\n # 1) 建表\n try:\n Base.metadata.create_all(bind=engine)\n except Exception as e:\n logging.exception(\"❌ 数据库建表失败: %s\", e)\n\n # 更改数据库表字段\n add_prompt_id_column()\n # v1.0.6添加字段 is_done\n add_is_done_column()\n # v1.0.7 添加字段 custom_params\n add_custom_params_column()\n # v1.0.7 添加项目的字段 is_precise_fill\n add_is_precise_fill_column()\n # v1.0.7 添加项目的字段 project_root_path\n add_project_root_path_column()\n\n # 2) 初始化共享运行时\n try:\n app.state.tts_queue = asyncio.Queue(maxsize=QUEUE_CAPACITY)\n app.state.tts_executor = ThreadPoolExecutor(max_workers=WORKERS)\n except Exception as e:\n logging.exception(\"❌ 初始化队列/线程池失败: %s\", e)\n\n # 3) 启动后台 worker\n try:\n app.state.tts_workers = [\n asyncio.create_task(tts_worker(app)) for _ in range(WORKERS)\n ]\n except Exception as e:\n logging.exception(\"❌ 启动 worker 失败: %s\", e)\n\n # 4) 初始化默认数据\n db = SessionLocal()\n try:\n try:\n tts_service = get_tts_service(db)\n tts_service.create_default_tts_provider()\n except Exception as e:\n logging.warning(\"⚠️ 默认 TTS provider 初始化失败: %s\", e)\n\n try:\n emotion_service = get_emotion_service(db)\n for name in [\n # 8种基础情绪\n \"高兴\", \"生气\", \"伤心\", \"害怕\", \"厌恶\", \"低落\", \"惊喜\", \"平静\",\n # 2种独特复合情绪\n \"嘲讽\", \"悲愤\",\n ]:\n try:\n emotion_service.create_emotion(EmotionEntity(name=name))\n except Exception as e:\n logging.debug(\"情绪 %s 已存在或创建失败: %s\", name, e)\n except Exception as e:\n logging.warning(\"⚠️ 情绪初始化失败: %s\", e)\n\n try:\n strength_service = get_strength_service(db)\n for name in [\"微弱\",\"稍弱\",\"中等\",\"较强\",\"强烈\"]:\n try:\n strength_service.create_strength(StrengthEntity(name=name))\n except Exception as e:\n logging.debug(\"强度 %s 已存在或创建失败: %s\", name, e)\n except Exception as e:\n logging.warning(\"⚠️ 强度初始化失败: %s\", e)\n\n # 创建默认提示词\n try:\n prompt_service = get_prompt_service(db)\n if not prompt_service.get_all_prompts():\n logging.info(\"创建默认提示词\")\n prompt_service.create_default_prompt()\n else:\n default_prompt = prompt_service.get_prompt_by_name(\"默认拆分台词提示词\")\n if not default_prompt:\n prompt_service.create_default_prompt()\n else:\n #修改默认提示词\n default_prompt_content = get_prompt_str()\n default_prompt.content = default_prompt_content\n prompt_service.update_prompt(default_prompt.id, default_prompt.__dict__)\n\n except Exception as e:\n logging.warning(\"⚠️ 默认提示词创建失败: %s\", e)\n # 兼容之前版本，已有的项目的project_root_path 为 getConfigPath()\n try:\n project_service = get_project_service(db)\n for project in project_service.get_all_projects():\n if not project.project_root_path:\n project.project_root_path = getConfigPath()\n project_service.update_project(project.id, project.__dict__)\n logging.info(\"项目 %s 默认项目路径已修改为 %s\", project.name, project.project_root_path)\n\n # todo:修改所有的保存路径，然后前端请求添加保存路径（利用electron读取文件夹路径）\n except Exception as e:\n logging.warning(\"⚠️ 项目默认项目路径初始化失败: %s\", e)\n\n except Exception as e:\n logging.exception(\"❌ 默认数据初始化异常: %s\", e)\n finally:\n db.close()\n\n@app.on_event(\"shutdown\")\nasync def shutdown_event():\n # 优雅退出\n for t in getattr(app.state, \"tts_workers\", []):\n t.cancel()\n ex = getattr(app.state, \"tts_executor\", None)\n if ex:\n ex.shutdown(wait=False, cancel_futures=True)\n# =========================\n# 注册路由\n# =========================\napp.include_router(project_router.router)\napp.include_router(chapter_router.router)\napp.include_router(role_router.router)\napp.include_router(voice_router.router)\napp.include_router(llm_provider_router.router)\napp.include_router(tts_provider_router.router)\napp.include_router(line_router.router)\napp.include_router(emotion_router.router)\napp.include_router(strength_router.router)\napp.include_router(multi_emotion_voice_router.router)\napp.include_router(prompt_router.router)\n# =========================\n# 健康检查接口\n# =========================\n@app.get(\"/\")\ndef read_root():\n return {\"msg\": \"音墟 (YinXu) 后端服务运行中！\"}\n\n# =========================\n# 小测试接口：插入并查询 ProjectPO\n# =========================\n@app.get(\"/test-db\")\ndef test_db():\n session: Session = SessionLocal()\n try:\n # 使用时间戳生成唯一名称，避免 UNIQUE 冲突\n name = f\"测试项目_{int(datetime.now().timestamp())}\"\n\n test_project = ProjectPO(name=name, description=\"测试用项目\")\n session.add(test_project)\n session.commit()\n session.refresh(test_project)\n\n return {\n \"msg\": \"插入成功\",\n \"id\": test_project.id,\n \"name\": test_project.name,\n \"created_at\": test_project.created_at,\n \"updated_at\": test_project.updated_at\n }\n\n except Exception as e:\n session.rollback()\n return {\"error\": str(e)}\n\n finally:\n session.close()\n\n\nimport json\nfrom fastapi import WebSocket, WebSocketDisconnect\n\n@app.websocket(\"/ws\")\nasync def ws_endpoint(ws: WebSocket):\n await manager.connect(ws)\n logging.info(\"WebSocket 客户端已连接\")\n try:\n while True:\n msg_text = await ws.receive_text()\n try:\n data = json.loads(msg_text)\n except json.JSONDecodeError:\n data = {}\n\n # 👇 心跳处理：收到 ping 立即回复 pong\n if data.get(\"type\") == \"ping\":\n logging.debug(\"receive ping\")\n await ws.send_text(json.dumps({\"type\": \"pong\"}))\n continue\n\n # 这里可以扩展处理订阅/其他消息\n\n except WebSocketDisconnect:\n logging.info(\"WebSocket 客户端主动断开\")\n manager.disconnect(ws)\n except Exception as e:\n logging.warning(f\"WebSocket 连接异常: {e}\")\n manager.disconnect(ws)\n\n\n\nif __name__ == \"__main__\":\n\n # uvicorn.run(app, host=\"127.0.0.1\", port=8200)\n # 使用自定义 logger，避免 uvicorn 自动配置失败\n # logging.basicConfig(level=logging.INFO)\n uvicorn.run(\"app.main:app\", host=\"127.0.0.1\", port=8200, log_config=None)\n"
},
{
"path": "SonicVale/app/models/po.py",
"content": "\nfrom sqlalchemy import Column, Integer, Integer, String, Text, Enum, ForeignKey, DateTime, JSON, Index\nfrom datetime import datetime, timezone\n\nfrom app.db.database import Base\n\n\n# ------------------------------\n# 1. 项目表 projects\n# ------------------------------\nclass ProjectPO(Base):\n __tablename__ = \"projects\"\n\n id = Column(Integer, primary_key=True, autoincrement=True,index=True)\n name = Column(String(255), nullable=False, unique=True, index=True)\n description = Column(Text, nullable=True)\n llm_provider_id = Column(Integer, nullable=True) # LLM提供商\n llm_model = Column(String(255), nullable=True) # 指定模型\n tts_provider_id = Column(Integer, nullable=True) # TTS提供商\n prompt_id = Column(Integer, nullable=True) # 关联的prompt\n # 是否开启精准填充\n is_precise_fill = Column(Integer, default=0, nullable=False)\n # 项目根地址\n project_root_path = Column(String(255), nullable=True)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc), nullable=False)\n\n\n# ------------------------------\n# 2. 项目的全局角色表 roles\n# ------------------------------\nclass RolePO(Base):\n __tablename__ = \"roles\"\n\n id = Column(Integer, primary_key=True, autoincrement=True,index=True)\n project_id = Column(Integer, nullable=False)\n name = Column(String(100), nullable=False)\n default_voice_id = Column(Integer, ForeignKey(\"voices.id\"), nullable=True)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc), nullable=False)\n\n\n# ------------------------------\n# 3. 音色表 voices\n# ------------------------------\nclass VoicePO(Base):\n __tablename__ = \"voices\"\n\n id = Column(Integer, primary_key=True, autoincrement=True, index=True)\n tts_provider_id = Column(Integer, nullable=True)\n name = Column(String(100), nullable=False)\n reference_path = Column(String(255), nullable=True)\n description = Column(Text, nullable=True)\n # 是否包含多情绪\n is_multi_emotion = Column(Integer, default=0, nullable=False)\n\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n nullable=False)\n\n# 多情绪表\nclass MultiEmotionVoicePO(Base):\n __tablename__ = \"multi_emotion\"\n id = Column(Integer, primary_key=True, autoincrement=True, index=True)\n voice_id = Column(Integer, nullable=False)\n emotion_id = Column(Integer, nullable=False)\n strength_id = Column(Integer, nullable=True)\n reference_path = Column(String(255), nullable=True)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n nullable=False)\n\n# ------------------------------\n# 4. 章节表 chapters\n# ------------------------------\nclass ChapterPO(Base):\n __tablename__ = \"chapters\"\n\n id = Column(Integer, primary_key=True, autoincrement=True,index=True)\n project_id = Column(Integer, nullable=False)\n title = Column(String(255), nullable=False)\n order_index = Column(Integer, nullable=True)\n text_content = Column(Text, nullable=True) # SQLite 没有 LongText，用 Text 替代\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n nullable=False)\n\n\n\n# ------------------------------\n# 5. 台词表 lines\n# ------------------------------\n# 情绪枚举表\nclass EmotionPO(Base):\n __tablename__ = \"emotions\"\n\n id = Column(Integer, primary_key=True, autoincrement=True, index=True)\n name = Column(String(100), nullable=False)\n description = Column(Text, nullable=True)\n is_active = Column(Integer, default=1, nullable=False)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now())\n\n# 情绪强弱枚举表\nclass StrengthPO(Base):\n __tablename__ = \"strengths\"\n\n id = Column(Integer, primary_key=True, autoincrement=True, index=True)\n name = Column(String(100), nullable=False)\n description = Column(Text, nullable=True)\n is_active = Column(Integer, default=1, nullable=False)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now())\n\n\nclass LinePO(Base):\n __tablename__ = \"lines\"\n\n id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n\n # 外键\n chapter_id = Column(Integer, nullable=False, index=True)\n role_id = Column(Integer, nullable=True)\n voice_id = Column(Integer, nullable=True)\n\n # 核心信息\n line_order = Column(Integer, nullable=True, index=True)\n text_content = Column(Text, nullable=True)\n # 情绪 和 强弱\n emotion_id = Column(Integer, nullable=True)\n strength_id = Column(Integer, nullable=True)\n\n # 9.1 新增\n\n\n # 输出资源\n audio_path = Column(String(500), nullable=True)\n subtitle_path = Column(String(500), nullable=True)\n\n # 间隔停留时间（秒）\n # wait_time = Column(Integer, default=0, nullable=True)\n\n # 状态\n status = Column(\n Enum(\"pending\", \"processing\", \"done\", \"failed\", name=\"line_status\"),\n default=\"pending\",\n nullable=False\n )\n # 是否完成\n is_done = Column(Integer, default=0, nullable=False)\n\n # 时间戳\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc), nullable=False)\n __table_args__ = (\n Index(\"idx_chapter_order\", \"chapter_id\", \"line_order\"),\n )\n\n# -------------------------\n# LLMProviderPO\n# -------------------------\nclass LLMProviderPO(Base):\n __tablename__ = \"llm_provider\"\n\n id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n name = Column(String(255), nullable=False, unique=True) # 提供商名称\n api_base_url = Column(String(500), nullable=False)\n api_key = Column(String(500), nullable=True) # 可加密存储\n model_list = Column(JSON, nullable=True) # 支持的模型列表\n status = Column(Integer, default=1, nullable=False) # 启用/禁用\n\n # ✅ 自定义参数（默认包含 response_format、temperature、top_p）\n custom_params = Column(\n Text,\n nullable=False,\n default=lambda: {\n \"response_format\": {\"type\": \"json_object\"},\n \"temperature\": 0.7,\n \"top_p\": 0.9\n\n }\n )\n # 时间戳\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n nullable=False)\n\n\n# -------------------------\n# TTSProviderPO\n# -------------------------\nclass TTSProviderPO(Base):\n __tablename__ = \"tts_provider\"\n\n id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n name = Column(String(255), nullable=False, unique=True)\n api_base_url = Column(String(500), nullable=False)\n api_key = Column(String(500), nullable=True)\n # voice_list = Column(JSON, nullable=True)\n status = Column(Integer, default=1, nullable=False)\n\n # 时间戳\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n nullable=False)\n\n\nclass PromptPO(Base):\n __tablename__ = \"prompts\"\n id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n name = Column(String(255), nullable=False)\n task = Column(String(255), nullable=False)\n description = Column(Text, nullable=True)\n content = Column(Text, nullable=True)\n created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),nullable=False)\n\n\n# -------------------------\n# ProjectSettings\n# -------------------------\n# class ProjectSettings(Base):\n# __tablename__ = \"project_settings\"\n#\n# id = Column(Integer, primary_key=True, index=True, autoincrement=True)\n# project_id = Column(Integer, nullable=False) # 所属项目\n# llm_provider_id = Column(Integer, nullable=True) # LLM提供商\n# llm_model = Column(String(255), nullable=True) # 指定模型\n# tts_provider_id = Column(Integer, nullable=True) # TTS提供商\n#\n# # 时间戳\n# created_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), nullable=False)\n# updated_at = Column(DateTime, default=lambda: datetime.now(timezone.utc), onupdate=lambda: datetime.now(timezone.utc),\n# nullable=False)\n"
},
{
"path": "SonicVale/app/repositories/chapter_repository.py",
"content": "from typing import Optional, Sequence\n\nfrom sqlalchemy import select\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import ChapterPO\n\n\nclass ChapterRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, chapter_id: int) -> Optional[ChapterPO]:\n \"\"\"根据 ID 查询项目\"\"\"\n return self.db.get(ChapterPO, chapter_id)\n\n def get_all(self, project_id: int) -> Sequence[ChapterPO]:\n \"\"\"获取指定项目下的所有章节\"\"\"\n stmt = select(ChapterPO).where(ChapterPO.project_id == project_id)\n return self.db.execute(stmt).scalars().all()\n\n def create(self, chapter_data: ChapterPO) -> ChapterPO:\n \"\"\"新建项目\"\"\"\n self.db.add(chapter_data)\n self.db.commit()\n self.db.refresh(chapter_data)\n return chapter_data\n\n def update(self, chapter_id: int, chapter_data: dict) -> Optional[ChapterPO]:\n \"\"\"更新项目\"\"\"\n chapter = self.get_by_id(chapter_id)\n if not chapter:\n return None\n for key, value in chapter_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(chapter, key, value)\n\n self.db.commit()\n self.db.refresh(chapter)\n return chapter\n\n def delete(self, chapter_id: int) -> bool:\n \"\"\"删除章节\"\"\"\n project = self.get_by_id(chapter_id)\n if not project:\n return False\n self.db.delete(project)\n self.db.commit()\n return True\n # def delete_all_by_project_id(self, project_id: int) -> bool:\n # \"\"\"删除指定项目下的所有章节\"\"\"\n # pos = self.get_all(project_id)\n # for po in pos:\n # self.db.delete(po)\n # self.db.commit()\n # return True\n\n def get_by_name(self, name: str, project_id: int) -> Optional[ChapterPO]:\n \"\"\"根据项目ID和章节名称查找章节\"\"\"\n stmt = (\n select(ChapterPO)\n .where(ChapterPO.title == name)\n .where(ChapterPO.project_id == project_id)\n )\n return self.db.execute(stmt).scalar_one_or_none()\n\n def search(self, keyword: str) -> Sequence[ChapterPO]:\n \"\"\"模糊搜索\"\"\"\n stmt = select(ChapterPO).where(ChapterPO.title.ilike(f\"%{keyword}%\"))\n return self.db.execute(stmt).scalars().all()"
},
{
"path": "SonicVale/app/repositories/emotion_repository.py",
"content": "from typing import Optional, Sequence\n\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import EmotionPO\n\n\nclass EmotionRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[EmotionPO]:\n \"\"\"通过id获取情绪\"\"\"\n return self.db.query(EmotionPO).filter(EmotionPO.id == id).first()\n\n def get_by_name(self, name: str) -> Optional[EmotionPO]:\n \"\"\"通过名称获取情绪\"\"\"\n return self.db.query(EmotionPO).filter(EmotionPO.name == name).first()\n\n def get_all(self) -> list[type[EmotionPO]]:\n \"\"\"获取所有情绪\"\"\"\n return self.db.query(EmotionPO).all()\n\n def create(self, emotion: EmotionPO) -> EmotionPO:\n \"\"\"创建情绪\"\"\"\n\n self.db.add(emotion)\n self.db.commit()\n self.db.refresh(emotion)\n return emotion\n\n def update(self, id: int, data: dict) -> Optional[EmotionPO]:\n \"\"\"更新情绪\"\"\"\n emotion = self.get_by_id(id)\n if not emotion:\n return None\n for key, value in data.items():\n if value is not None:\n setattr(emotion, key, value)\n self.db.commit()\n self.db.refresh(emotion)\n return emotion\n\n def delete(self, id: int) -> bool:\n \"\"\"删除情绪\"\"\"\n emotion = self.get_by_id(id)\n if not emotion:\n return False\n self.db.delete(emotion)\n self.db.commit()\n return True\n\n\n"
},
{
"path": "SonicVale/app/repositories/line_repository.py",
"content": "from typing import Optional, List\n\nfrom sqlalchemy import Sequence, select, update\nfrom sqlalchemy.orm import Session\n\nfrom app.dto.line_dto import LineOrderDTO\nfrom app.models.po import LinePO\n\n\nclass LineRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[LinePO]:\n \"\"\"根据 ID 查询单行台词\"\"\"\n return self.db.get(LinePO, id)\n\n def get_all(self, chapter_id: int) -> Sequence[LinePO]:\n \"\"\"获取章节下所有单行台词，按 line_order 排序\"\"\"\n stmt = (\n select(LinePO)\n .where(LinePO.chapter_id == chapter_id)\n .order_by(LinePO.line_order.asc()) # 升序\n )\n return self.db.execute(stmt).scalars().all()\n\n\n def create(self, data: LinePO) -> LinePO:\n \"\"\"新增单行台词\"\"\"\n self.db.add(data)\n self.db.commit()\n self.db.refresh(data)\n return data\n\n\n def update(self, line_id: int, line_data: dict) -> Optional[LinePO]:\n \"\"\"更新单行台词信息\"\"\"\n line = self.get_by_id(line_id)\n if not line:\n return None\n for key, value in line_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(line, key, value)\n\n self.db.commit()\n self.db.refresh(line)\n return line\n\n def delete(self, line_id: int) -> bool:\n \"\"\"删除台词\"\"\"\n line = self.get_by_id(line_id)\n if not line:\n return False\n self.db.delete(line)\n self.db.commit()\n return True\n def delete_all_by_chapter_id(self, chapter_id: int) -> bool:\n \"\"\"删除章节下的所有台词\"\"\"\n lines = self.get_all(chapter_id)\n for line in lines:\n self.db.delete(line)\n self.db.commit()\n return True\n\n def get_lines_by_role_id(self, role_id: int):\n return self.db.execute(select(LinePO).where(LinePO.role_id == role_id)).scalars().all()\n\n def batch_update_line_order(self, line_orders:List[LineOrderDTO])-> int:\n \"\"\"批量更新台词的顺序\"\"\"\n if not line_orders:\n return 0\n\n from sqlalchemy import bindparam\n stmt = (\n update(LinePO)\n .where(LinePO.id == bindparam(\"id\"))\n .values(line_order=bindparam(\"line_order\"))\n )\n params = [{\"id\": it.id, \"line_order\": it.line_order} for it in line_orders]\n res = self.db.execute(stmt, params) # executemany\n self.db.commit()\n return res.rowcount if res.rowcount not in (None, -1) else len(params)\n"
},
{
"path": "SonicVale/app/repositories/llm_provider_repository.py",
"content": "from typing import List, Optional, Sequence, Any\nfrom sqlalchemy.orm import Session\nfrom sqlalchemy import select, Row, RowMapping\nfrom app.models.po import LLMProviderPO\n\n\nclass LLMProviderRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, llm_provider_id: int) -> Optional[LLMProviderPO]:\n \"\"\"根据 ID 查询LLM供应商\"\"\"\n return self.db.get(LLMProviderPO, llm_provider_id)\n\n def get_all(self) -> Sequence[LLMProviderPO]:\n \"\"\"获取所有LLM供应商\"\"\"\n return self.db.execute(select(LLMProviderPO)).scalars().all()\n\n def create(self, llm_provider_data: LLMProviderPO) -> LLMProviderPO:\n \"\"\"新建LLM供应商\"\"\"\n self.db.add(llm_provider_data)\n self.db.commit()\n self.db.refresh(llm_provider_data)\n return llm_provider_data\n\n def update(self, llm_provider_id: int, llm_provider_data: dict) -> Optional[LLMProviderPO]:\n \"\"\"更新LLM供应商\"\"\"\n llm_provider = self.get_by_id(llm_provider_id)\n if not llm_provider:\n return None\n for key, value in llm_provider_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(llm_provider, key, value)\n self.db.commit()\n self.db.refresh(llm_provider)\n return llm_provider\n\n def delete(self, llm_provider_id: int) -> bool:\n \"\"\"删除LLM供应商\"\"\"\n llm_provider = self.get_by_id(llm_provider_id)\n if not llm_provider:\n return False\n self.db.delete(llm_provider)\n self.db.commit()\n return True\n\n def get_by_name(self, name: str) -> Optional[LLMProviderPO]:\n \"\"\"根据名称查找LLM供应商\"\"\"\n stmt = select(LLMProviderPO).where(LLMProviderPO.name == name)\n return self.db.execute(stmt).scalar_one_or_none()\n\n def search(self, keyword: str) -> Sequence[LLMProviderPO]:\n \"\"\"模糊搜索\"\"\"\n stmt = select(LLMProviderPO).where(LLMProviderPO.name.ilike(f\"%{keyword}%\"))\n return self.db.execute(stmt).scalars().all()\n"
},
{
"path": "SonicVale/app/repositories/multi_emotion_voice_repository.py",
"content": "from typing import Optional, Sequence, Any\n\nfrom sqlalchemy.orm import Session, Query\n\nfrom app.models.po import MultiEmotionVoicePO\n\n\nclass MultiEmotionVoiceRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[MultiEmotionVoicePO]:\n \"\"\"通过id获取多情绪音色\"\"\"\n return self.db.query(MultiEmotionVoicePO).filter(MultiEmotionVoicePO.id == id).first()\n\n # 根据voice_id,emotion_id,strength_id获取多情绪音色\n def get_by_voice_id_emotion_id_strength_id(self, voice_id: int, emotion_id: int, strength_id: int) -> type[MultiEmotionVoicePO] | None:\n \"\"\"根据voice_id,emotion_id,strength_id获取多情绪音色\"\"\"\n return self.db.query(MultiEmotionVoicePO).filter(MultiEmotionVoicePO.voice_id == voice_id,\n MultiEmotionVoicePO.emotion_id == emotion_id,\n MultiEmotionVoicePO.strength_id == strength_id).one_or_none()\n # 根据voice_id获取多情绪音色\n def get_by_voice_id(self, voice_id: int) -> Sequence[type[MultiEmotionVoicePO]]:\n \"\"\"根据voice_id获取多情绪音色\"\"\"\n return self.db.query(MultiEmotionVoicePO).filter(MultiEmotionVoicePO.voice_id == voice_id).all()\n\n def get_all(self) -> list[type[MultiEmotionVoicePO]]:\n \"\"\"获取所有多情绪音色\"\"\"\n return self.db.query(MultiEmotionVoicePO).all()\n\n def create(self, multi_emotion_voice: MultiEmotionVoicePO) -> MultiEmotionVoicePO:\n \"\"\"创建多情绪音色\"\"\"\n\n self.db.add(multi_emotion_voice)\n self.db.commit()\n self.db.refresh(multi_emotion_voice)\n return multi_emotion_voice\n\n def update(self, id: int, data: dict) -> Optional[MultiEmotionVoicePO]:\n \"\"\"更新多情绪音色\"\"\"\n multi_emotion_voice = self.get_by_id(id)\n if not multi_emotion_voice:\n return None\n for key, value in data.items():\n if value is not None:\n setattr(multi_emotion_voice, key, value)\n self.db.commit()\n self.db.refresh(multi_emotion_voice)\n return multi_emotion_voice\n\n def delete(self, id: int) -> bool:\n \"\"\"删除多情绪音色\"\"\"\n multi_emotion_voice = self.get_by_id(id)\n if not multi_emotion_voice:\n return False\n self.db.delete(multi_emotion_voice)\n self.db.commit()\n return True\n\n def delete_multi_emotion_voice_by_voice_id(self, voice_id):\n \"\"\"通过音色id删除所有的多音色\"\"\"\n multi_voices = self.get_by_voice_id(voice_id)\n for multi_voice in multi_voices:\n self.db.delete(multi_voice)\n self.db.commit()\n return True\n\n\n"
},
{
"path": "SonicVale/app/repositories/project_repository.py",
"content": "from typing import List, Optional, Sequence, Any\nfrom sqlalchemy.orm import Session\nfrom sqlalchemy import select, Row, RowMapping\nfrom app.models.po import ProjectPO\n\n\nclass ProjectRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, project_id: int) -> Optional[ProjectPO]:\n \"\"\"根据 ID 查询项目\"\"\"\n return self.db.get(ProjectPO, project_id)\n\n def get_all(self) -> Sequence[ProjectPO]:\n \"\"\"获取所有项目\"\"\"\n return self.db.execute(select(ProjectPO)).scalars().all()\n\n def create(self, project_data: ProjectPO) -> ProjectPO:\n \"\"\"新建项目\"\"\"\n self.db.add(project_data)\n self.db.commit()\n self.db.refresh(project_data)\n return project_data\n\n def update(self, project_id: int, project_data: dict) -> Optional[ProjectPO]:\n \"\"\"更新项目\"\"\"\n project = self.get_by_id(project_id)\n if not project:\n return None\n for key, value in project_data.items():\n setattr(project, key, value)\n self.db.commit()\n self.db.refresh(project)\n return project\n\n def delete(self, project_id: int) -> bool:\n \"\"\"删除项目\"\"\"\n project = self.get_by_id(project_id)\n if not project:\n return False\n self.db.delete(project)\n self.db.commit()\n return True\n\n def get_by_name(self, name: str) -> Optional[ProjectPO]:\n \"\"\"根据名称查找项目\"\"\"\n stmt = select(ProjectPO).where(ProjectPO.name == name)\n return self.db.execute(stmt).scalar_one_or_none()\n\n def search(self, keyword: str) -> Sequence[ProjectPO]:\n \"\"\"模糊搜索\"\"\"\n stmt = select(ProjectPO).where(ProjectPO.name.ilike(f\"%{keyword}%\"))\n return self.db.execute(stmt).scalars().all()\n"
},
{
"path": "SonicVale/app/repositories/prompt_repository.py",
"content": "from typing import List, Optional, Sequence, Any\nfrom sqlalchemy.orm import Session\nfrom sqlalchemy import select, Row, RowMapping\nfrom app.models.po import PromptPO\n\n\nclass PromptRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, prompt_id: int) -> Optional[PromptPO]:\n \"\"\"根据 ID 查询提示词\"\"\"\n return self.db.get(PromptPO, prompt_id)\n\n def get_all(self) -> Sequence[PromptPO]:\n \"\"\"获取所有提示词\"\"\"\n return self.db.execute(select(PromptPO)).scalars().all()\n\n def create(self, prompt_data: PromptPO) -> PromptPO:\n \"\"\"新建提示词\"\"\"\n self.db.add(prompt_data)\n self.db.commit()\n self.db.refresh(prompt_data)\n return prompt_data\n\n def update(self, prompt_id: int, prompt_data: dict) -> Optional[PromptPO]:\n \"\"\"更新提示词\"\"\"\n prompt = self.get_by_id(prompt_id)\n if not prompt:\n return None\n for key, value in prompt_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(prompt, key, value)\n self.db.commit()\n self.db.refresh(prompt)\n return prompt\n\n def delete(self, prompt_id: int) -> bool:\n \"\"\"删除提示词\"\"\"\n prompt = self.get_by_id(prompt_id)\n if not prompt:\n return False\n self.db.delete(prompt)\n self.db.commit()\n return True\n\n def get_by_name(self, name: str) -> Optional[PromptPO]:\n \"\"\"根据名称查找提示词\"\"\"\n stmt = select(PromptPO).where(PromptPO.name == name)\n return self.db.execute(stmt).scalar_one_or_none()\n\n # 根据任务查询，返回多个提示词\n def get_by_task(self, task: str) -> Sequence[PromptPO]:\n stmt = select(PromptPO).where(PromptPO.task == task)\n return self.db.execute(stmt).scalars().all()\n\n\n def search(self, keyword: str) -> Sequence[PromptPO]:\n \"\"\"模糊搜索\"\"\"\n stmt = select(PromptPO).where(PromptPO.name.ilike(f\"%{keyword}%\"))\n return self.db.execute(stmt).scalars().all()\n"
},
{
"path": "SonicVale/app/repositories/role_repository.py",
"content": "from typing import Optional\n\nfrom sqlalchemy import Sequence, select\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import RolePO\n\n\nclass RoleRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[RolePO]:\n \"\"\"根据 ID 查询角色\"\"\"\n return self.db.get(RolePO, id)\n\n def get_all(self,project_id: int) -> Sequence[RolePO]:\n \"\"\"获取项目下所有角色\"\"\"\n return self.db.execute(select(RolePO).where(RolePO.project_id == project_id)).scalars().all()\n\n\n def create(self, data: RolePO) -> RolePO:\n \"\"\"新增角色\"\"\"\n self.db.add(data)\n self.db.commit()\n self.db.refresh(data)\n return data\n\n\n def update(self, role_id: int, role_data: dict) -> Optional[RolePO]:\n \"\"\"更新角色信息\"\"\"\n role = self.get_by_id(role_id)\n if not role:\n return None\n for key, value in role_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(role, key, value)\n\n self.db.commit()\n self.db.refresh(role)\n return role\n\n def delete(self, role_id: int) -> bool:\n \"\"\"删除项目\"\"\"\n role = self.get_by_id(role_id)\n if not role:\n return False\n self.db.delete(role)\n self.db.commit()\n return True\n\n\n def get_by_name(self, name: str,project_id: int) -> Optional[RolePO]:\n \"\"\"根据名称查找项目下的角色信息\"\"\"\n return self.db.execute(select(RolePO).where(RolePO.name == name,RolePO.project_id == project_id)).scalars().first()\n\n\n"
},
{
"path": "SonicVale/app/repositories/strength_repository.py",
"content": "from typing import Optional, Sequence\n\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import StrengthPO\n\n\nclass StrengthRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[StrengthPO]:\n \"\"\"通过id获取情绪强弱\"\"\"\n return self.db.query(StrengthPO).filter(StrengthPO.id == id).first()\n\n def get_by_name(self, name: str) -> Optional[StrengthPO]:\n \"\"\"通过名称获取情绪强弱\"\"\"\n return self.db.query(StrengthPO).filter(StrengthPO.name == name).first()\n\n def get_all(self) -> list[type[StrengthPO]]:\n \"\"\"获取所有情绪强弱\"\"\"\n return self.db.query(StrengthPO).all()\n\n def create(self, strength: StrengthPO) -> StrengthPO:\n \"\"\"创建情绪强弱\"\"\"\n self.db.add(strength)\n self.db.commit()\n self.db.refresh(strength)\n return strength\n\n def update(self, id: int, data: dict) -> Optional[StrengthPO]:\n \"\"\"更新情绪强弱\"\"\"\n strength = self.get_by_id(id)\n if not strength:\n return None\n for key, value in data.items():\n if value is not None:\n setattr(strength, key, value)\n self.db.commit()\n self.db.refresh(strength)\n return strength\n\n def delete(self, id: int) -> bool:\n \"\"\"删除情绪强弱\"\"\"\n strength = self.get_by_id(id)\n if not strength:\n return False\n self.db.delete(strength)\n self.db.commit()\n return True\n\n\n"
},
{
"path": "SonicVale/app/repositories/tts_provider_repository.py",
"content": "from typing import Optional\n\nfrom sqlalchemy import Sequence, select\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import TTSProviderPO\n\n\nclass TTSProviderRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[TTSProviderPO]:\n \"\"\"根据 ID 查询tts供应商\"\"\"\n return self.db.get(TTSProviderPO, id)\n\n def get_all(self) -> Sequence[TTSProviderPO]:\n \"\"\"获取tts下所有tts供应商\"\"\"\n return self.db.execute(select(TTSProviderPO)).scalars().all()\n\n\n def create(self, data: TTSProviderPO) -> TTSProviderPO:\n\n \"\"\"新增tts供应商\"\"\"\n self.db.add(data)\n self.db.commit()\n self.db.refresh(data)\n return data\n\n\n def update(self, tts_provider_id: int, voice_data: dict) -> Optional[TTSProviderPO]:\n \"\"\"更新tts供应商信息\"\"\"\n voice = self.get_by_id(tts_provider_id)\n if not voice:\n return None\n for key, value in voice_data.items():\n if value is not None: # 只更新不为空的字段\n setattr(voice, key, value)\n\n self.db.commit()\n self.db.refresh(voice)\n return voice\n\n # def delete(self, voice_id: int) -> bool:\n # \"\"\"删除项目\"\"\"\n # voice = self.get_by_id(voice_id)\n # if not voice:\n # return False\n # self.db.delete(voice)\n # self.db.commit()\n # return True\n #\n #\n def get_by_name(self, name: str) -> Optional[TTSProviderPO]:\n \"\"\"根据名称查找项目下的tts供应商信息\"\"\"\n return self.db.execute(select(TTSProviderPO).where(TTSProviderPO.name == name)).scalars().first()\n\n\n"
},
{
"path": "SonicVale/app/repositories/voice_repository.py",
"content": "from typing import Optional\n\nfrom sqlalchemy import Sequence, select\nfrom sqlalchemy.orm import Session\n\nfrom app.models.po import VoicePO\n\n\nclass VoiceRepository:\n def __init__(self, db: Session):\n self.db = db\n\n def get_by_id(self, id: int) -> Optional[VoicePO]:\n \"\"\"根据 ID 查询音色\"\"\"\n return self.db.get(VoicePO, id)\n\n def get_all(self,tts_id: int) -> Sequence[VoicePO]:\n \"\"\"获取tts下所有音色\"\"\"\n return self.db.execute(select(VoicePO).where(VoicePO.tts_provider_id == tts_id)).scalars().all()\n\n def get_by_ids(self, tts_id: int, ids: list[int]) -> Sequence[VoicePO]:\n \"\"\"根据ids获取tts下的音色\"\"\"\n if not ids:\n return []\n return self.db.execute(\n select(VoicePO).where(VoicePO.tts_provider_id == tts_id, VoicePO.id.in_(ids))\n ).scalars().all()\n\n\n def create(self, data: VoicePO) -> VoicePO:\n \"\"\"新增音色\"\"\"\n self.db.add(data)\n self.db.commit()\n self.db.refresh(data)\n return data\n\n\n def update(self, voice_id: int, voice_data: dict) -> Optional[VoicePO]:\n \"\"\"更新音色信息\"\"\"\n voice = self.get_by_id(voice_id)\n if not voice:\n return None\n for key, value in voice_data.items():\n setattr(voice, key, value)\n\n self.db.commit()\n self.db.refresh(voice)\n return voice\n\n def delete(self, voice_id: int) -> bool:\n \"\"\"删除项目\"\"\"\n voice = self.get_by_id(voice_id)\n if not voice:\n return False\n self.db.delete(voice)\n self.db.commit()\n return True\n\n\n def get_by_name(self, name: str,tts_id: int) -> Optional[VoicePO]:\n \"\"\"根据名称查找项目下的音色信息\"\"\"\n return self.db.execute(select(VoicePO).where(VoicePO.name == name,VoicePO.tts_provider_id == tts_id)).scalars().first()\n\n\n"
},
{
"path": "SonicVale/app/routers/chapter_router.py",
"content": "# 初始化 router\nimport asyncio\nimport io\nimport json\nimport logging\nimport os\nimport traceback\n\nfrom typing import List\n\n\nfrom sqlalchemy.orm import Session\nfrom fastapi import APIRouter, Depends, HTTPException, Form\n\n\nfrom app.core.response import Res\nfrom app.core.text_correct_engine import TextCorrectorFinal\nfrom app.core.ws_manager import manager\nfrom app.db.database import get_db, SessionLocal\nfrom app.dto.chapter_dto import ChapterResponseDTO, ChapterCreateDTO\nfrom app.dto.line_dto import LineInitDTO, LineCreateDTO, LineResponseDTO\nfrom app.entity.chapter_entity import ChapterEntity\nfrom app.repositories.chapter_repository import ChapterRepository\nfrom app.repositories.emotion_repository import EmotionRepository\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.repositories.prompt_repository import PromptRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.repositories.strength_repository import StrengthRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.repositories.voice_repository import VoiceRepository\n\nfrom app.services.chapter_service import ChapterService\nfrom app.services.emotion_service import EmotionService\nfrom app.services.line_service import LineService\nfrom app.services.multi_emotion_voice_service import MultiEmotionVoiceService\nfrom app.services.project_service import ProjectService\nfrom app.services.prompt_service import PromptService\nfrom app.services.role_service import RoleService\nfrom app.services.strength_service import StrengthService\nfrom app.services.voice_service import VoiceService\n\nrouter = APIRouter(prefix=\"/chapters\", tags=[\"Chapters\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_chapter_service(db: Session = Depends(get_db)) -> ChapterService:\n repository = ChapterRepository(db) # ✅ 传入 db\n return ChapterService(repository)\n\ndef get_line_service(db: Session = Depends(get_db)) -> LineService:\n repository = LineRepository(db)\n role_repository = RoleRepository(db)\n tts_provider_repository = TTSProviderRepository(db)\n llm_provider_repository = LLMProviderRepository(db)\n return LineService(repository, role_repository, tts_provider_repository, llm_provider_repository)\n\ndef get_project_service(db: Session = Depends(get_db)) -> ProjectService:\n repository = ProjectRepository(db)\n return ProjectService(repository)\n\ndef get_voice_service(db: Session = Depends(get_db)) -> VoiceService:\n repository = VoiceRepository(db)\n multi_emotion_voice_repository = MultiEmotionVoiceRepository(db)\n return VoiceService(repository,multi_emotion_voice_repository)\n\ndef get_role_service(db: Session = Depends(get_db)) -> RoleService:\n repository = RoleRepository(db)\n return RoleService(repository)\n\ndef get_emotion_service(db: Session = Depends(get_db)) -> EmotionService:\n repository = EmotionRepository(db)\n return EmotionService(repository)\n\ndef get_strength_service(db: Session = Depends(get_db)) -> StrengthService:\n repository = StrengthRepository(db)\n return StrengthService(repository)\n\ndef get_multi_emotion_voice_service(db: Session = Depends(get_db)) -> MultiEmotionVoiceService:\n repository = MultiEmotionVoiceRepository(db)\n return MultiEmotionVoiceService(repository)\n\ndef get_prompt_service(db: Session = Depends(get_db)) -> PromptService:\n repository = PromptRepository(db)\n return PromptService(repository)\n\n@router.post(\"\", response_model=Res[ChapterResponseDTO],\n summary=\"创建章节\",\n description=\"根据项目ID创建章节，章节名称在同一项目下不可重复\" )\nasync def create_chapter(dto: ChapterCreateDTO, chapter_service: ChapterService = Depends(get_chapter_service),\n project_service: ProjectService = Depends(get_project_service)):\n \"\"\"创建章节\"\"\"\n try:\n # DTO → Entity\n entity = ChapterEntity(**dto.__dict__)\n # 判断project_id是否存在\n project = project_service.get_project(dto.project_id)\n if project is None:\n return Res(data=None, code=400, message=f\"项目 '{dto.project_id}' 不存在\")\n # 调用 Service 创建项目（返回 True/False）\n entityRes = chapter_service.create_chapter(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = ChapterResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"章节 '{entity.title}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n@router.get(\"/{chapter_id}\", response_model=Res[ChapterResponseDTO],\n summary=\"查询章节\",\n description=\"根据章节id查询章节信息\")\nasync def get_chapter(chapter_id: int, chapter_service: ChapterService = Depends(get_chapter_service)):\n entity = chapter_service.get_chapter(chapter_id)\n if entity:\n res = ChapterResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"项目不存在\")\n\n@router.get(\"/project/{project_id}\", response_model=Res[List[ChapterResponseDTO]],\n summary=\"查询项目下的所有章节\",\n description=\"根据项目id查询项目下的所有章节信息\")\nasync def get_all_chapters(project_id: int, chapter_service: ChapterService = Depends(get_chapter_service)):\n entities = chapter_service.get_all_chapters(project_id)\n if entities:\n res = [ChapterResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=404, message=\"项目不存在章节\")\n\n# 修改，传入的参数是id\n@router.put(\"/{chapter_id}\", response_model=Res[ChapterCreateDTO],\n summary=\"修改章节\",\n description=\"根据章节id修改章节信息,并且不能修改项目id\")\nasync def update_chapter(chapter_id: int, dto: ChapterCreateDTO, chapter_service: ChapterService = Depends(get_chapter_service)):\n chapter = chapter_service.get_chapter(chapter_id)\n if chapter is None:\n return Res(data=None, code=404, message=\"章节不存在\")\n res = chapter_service.update_chapter(chapter_id, dto.dict(exclude_unset=True))\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败\")\n\n\n# 根据id，删除\n@router.delete(\"/{chapter_id}\", response_model=Res,\n summary=\"删除章节\",\n description=\"根据章节id删除章节信息,并且级联删除\")\nasync def delete_chapter(chapter_id: int, chapter_service: ChapterService = Depends(get_chapter_service)):\n success = chapter_service.delete_chapter(chapter_id)\n\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或章节不存在\")\n\n\n# 根据内容进行解析得到json,初次解析，然后可编辑角色昵称以及内容，以及可以合并上下或者增加。（json都是多条，角色+台词）\n@router.get(\n \"/get-lines/{project_id}/{chapter_id}\",\n response_model=Res[str],\n summary=\"根据内容进行解析得到json\",\n description=\"根据内容进行解析得到json\"\n)\nasync def get_lines(\n project_id: int,\n chapter_id: int,\n chapter_service: ChapterService = Depends(get_chapter_service),\n line_service: LineService = Depends(get_line_service),\n role_service: RoleService = Depends(get_role_service),\n emotion_service: EmotionService = Depends(get_emotion_service),\n strength_service: StrengthService = Depends(get_strength_service),\n prompt_service: PromptService = Depends(get_prompt_service),\n project_service: ProjectService = Depends(get_project_service)\n):\n # 判断章节内容是否存在\n chapter = chapter_service.get_chapter(chapter_id)\n if chapter.text_content is None:\n return Res(data=None, code=400, message=\"章节内容不存在\")\n try:\n contents = chapter_service.split_text(chapter_id, 1500)\n logging.info(\"内容划分为 %s 段\", len(contents))\n except Exception as e:\n logging.error(f\"章节拆分失败: {e}\\n{traceback.format_exc()}\")\n return Res(data=None, code=500, message=\"章节拆分失败\")\n\n all_line_data = []\n\n try:\n roles = role_service.get_all_roles(project_id)\n roles = set(role.name for role in roles)\n emotions = emotion_service.get_all_emotions()\n strengths = strength_service.get_all_strengths()\n\n emotion_names = [emotion.name for emotion in emotions]\n strength_names = [strength.name for strength in strengths]\n emotions_dict = {emotion.name: emotion.id for emotion in emotions}\n strengths_dict = {strength.name: strength.id for strength in strengths}\n except Exception as e:\n logging.error(f\"初始化角色/情绪/强度失败: {e}\\n{traceback.format_exc()}\")\n return Res(data=None, code=500, message=\"初始化角色/情绪/强度失败\")\n\n project = project_service.get_project(project_id)\n # 精准填充\n is_precise_fill = project.is_precise_fill\n # 判断tts，llm，model是否存在\n if project.tts_provider_id is None or project.llm_provider_id is None or project.llm_model is None:\n return Res(data=None, code=500, message=\"tts/llm/model不存在\")\n\n\n prompt = prompt_service.get_prompt(project.prompt_id) if project else None\n if prompt is None:\n return Res(data=None, code=500, message=\"提示词不存在\")\n\n for idx, content in enumerate(contents):\n logging.info(f\"解析第 {idx + 1}/{len(contents)} 段...\")\n\n try:\n roles_list = list(roles)\n result = chapter_service.para_content(\n prompt.content, chapter_id, content,\n roles_list, emotion_names, strength_names,is_precise_fill\n )\n\n if not result[\"success\"]:\n return Res(\n data=None,\n code=500,\n message=result[\"message\"]\n )\n\n # 提取lines_data中的角色\n lines_data = result[\"data\"]\n for line_data in lines_data:\n roles.add(line_data.role_name)\n\n all_line_data.extend(lines_data)\n\n except Exception as e:\n logging.error(\n f\"解析第 {idx + 1} 段失败: {e}\\n{traceback.format_exc()}\"\n )\n return Res(data=None, code=500, message=f\"解析失败：第 {idx + 1} 段处理出错，错误信息：{e}\")\n\n try:\n audio_path = os.path.join(project.project_root_path,str(project_id),str(chapter_id),\"audio\")\n os.makedirs(audio_path, exist_ok=True)\n line_service.update_init_lines(\n all_line_data, project_id, chapter_id, emotions_dict, strengths_dict,audio_path\n )\n except Exception as e:\n logging.error(f\"写入数据库失败: {e}\\n{traceback.format_exc()}\")\n return Res(data=None, code=500, message=\"写入数据库失败\")\n\n return Res(data=None, code=200, message=\"解析成功\")\n\n\n# 导出LLM prompt指令\n@router.get(\"/export-llm-prompt/{project_id}/{chapter_id}\",response_model=Res[str],summary=\"导出LLM prompt指令\",description=\"导出LLM prompt指令\")\nasync def export_llm_prompt(project_id:int,chapter_id: int, chapter_service: ChapterService = Depends(get_chapter_service),\n project_service = Depends(get_project_service),\n prompt_service: PromptService = Depends(get_prompt_service),\n role_service: RoleService = Depends(get_role_service),\n emotion_service: EmotionService = Depends(get_emotion_service),\n strength_service: StrengthService = Depends(get_strength_service)):\n try:\n roles = role_service.get_all_roles(project_id)\n roles = [role.name for role in roles]\n emotions = emotion_service.get_all_emotions()\n strengths = strength_service.get_all_strengths()\n\n emotion_names = [emotion.name for emotion in emotions]\n strength_names = [strength.name for strength in strengths]\n except Exception as e:\n return Res(data=None, code=500, message=\"初始化角色/情绪/强度失败\")\n\n project = project_service.get_project(project_id)\n prompt = prompt_service.get_prompt(project.prompt_id) if project else None\n chapter = chapter_service.get_chapter(chapter_id)\n content = chapter.text_content\n res = chapter_service.fill_prompt(prompt.content, roles, emotion_names, strength_names, content)\n # record\n return Res(data=res, code=200, message=\"导出成功\")\n\n# 解析第三方的json\n@router.post(\"/import-lines/{project_id}/{chapter_id}\",response_model=Res[str],summary=\"导入第三方json\",description=\"导入第三方json\")\nasync def import_lines(project_id: int,chapter_id: int,data:str=Form( ...),line_service: LineService = Depends(get_line_service),\n emotion_service: EmotionService = Depends(get_emotion_service),\n strength_service: StrengthService = Depends(get_strength_service),\n project_service: ProjectService = Depends(get_project_service),\n chapter_service: ChapterService = Depends(get_chapter_service)):\n # 解析data\n lines_data = json.loads(data)\n # 转化成List[LineInitDTO]\n emotions = emotion_service.get_all_emotions()\n strengths = strength_service.get_all_strengths()\n\n emotions_dict = {emotion.name: emotion.id for emotion in emotions}\n strengths_dict = {strength.name: strength.id for strength in strengths}\n # 精准填充\n project = project_service.get_project(project_id)\n is_precise_fill = project.is_precise_fill\n \n if is_precise_fill == 1:\n # 获取章节内容\n content = chapter_service.get_chapter(chapter_id).text_content\n if not content:\n return Res(data=None, code=500, message=\"章节内容为空\")\n corrector = TextCorrectorFinal()\n lines_data = corrector.correct_ai_text(content, lines_data)\n lines_data = [LineInitDTO(**line) for line in lines_data]\n\n\n audio_path = os.path.join(project.project_root_path,str(project_id),str(chapter_id),\"audio\")\n os.makedirs(audio_path, exist_ok=True)\n line_service.update_init_lines(lines_data, project_id, chapter_id, emotions_dict, strengths_dict,audio_path)\n return Res(data=None, code=200, message=\"导入成功\")\n\n\n\n# @router.post(\"/save-init-lines/{project_id}/{chapter_id}\",response_model=Res[str],summary=\"保存初始化调整后的解析内容\",description=\"保存初始化调整后的解析内容\")\n# async def update_init_lines(project_id: int,chapter_id: int,lines: List[LineInitDTO], chapter_service: ChapterService = Depends(get_chapter_service)):\n# chapter_service.update_init_lines(lines,project_id,chapter_id)\n# return Res(data=None, code=200, message=\"保存成功\")\n\n# 绑定音色就是采用的修改角色信息\n\n# 获取章节下所有台词\n\n\n\n# 传入台词实体，然后生成音频\n# @router.post(\"/generate-audio/{project_id}/{chapter_id}\",response_model=Res[str],summary=\"生成音频\",description=\"生成音频\")\n# async def generate_audio(project_id: int,chapter_id: int,\n# dto: LineCreateDTO, chapter_service: ChapterService = Depends(get_chapter_service),\n# voice_service: VoiceService = Depends(get_voice_service),\n# role_service: RoleService = Depends(get_role_service),\n# project_service: ProjectService = Depends(get_project_service)):\n# \"\"\"生成音频\"\"\"\n# # 获取角色绑定的音色的reference_path\n# role = role_service.get_role(dto.role_id)\n# voice = voice_service.get_voice(role.default_voice_id)\n# project = project_service.get_project(project_id)\n# save_path = dto.audio_path\n# res = chapter_service.generate_audio(voice.reference_path,project.tts_provider_id,dto.text_content,save_path=save_path)\n# return Res(data=None, code=200, message=\"生成成功\")\n\n# 合并结果并导出\n# @router.get(\"/export-audio/{project_id}/{chapter_id}\",response_model=Res[str],summary=\"合并结果并导出\",description=\"合并结果并导出\")\n# async def export_audio(project_id: int,chapter_id: int, chapter_service: ChapterService = Depends(get_chapter_service))\n# res = chapter_service.export_audio(project_id,chapter_id)\n\n# 添加智能匹配角色和音色的功能\n@router.post(\"/add-smart-role-and-voice/{project_id}/{chapter_id}\",response_model=Res[List],summary=\"添加智能匹配角色和音色的功能\",description=\"添加智能匹配角色和音色的功能\")\nasync def add_smart_role_and_voice(project_id: int,chapter_id: int,\n chapter_service: ChapterService = Depends(get_chapter_service),\n project_service: ProjectService = Depends(get_project_service),\n voice_service: VoiceService = Depends(get_voice_service),\n role_service: RoleService = Depends(get_role_service)):\n # 获取项目信息\n project = project_service.get_project(project_id)\n # 首先获取项目下所有角色\n roles = role_service.get_all_roles(project_id)\n# 将所有角色未绑定音色的角色提取出来\n roles_no_voice = [role for role in roles if role.default_voice_id is None]\n # 只要角色name\n role_names = [role.name for role in roles_no_voice]\n # 获取所有音色\n voices = voice_service.get_all_voices(project.tts_provider_id)\n # 只要音色的名字和描述\n voice_names = [\n {\n \"name\": voice.name,\n \"description\": voice.description\n }\n for voice in voices\n ]\n # 获取原文内容\n content = chapter_service.get_chapter(chapter_id).text_content\n res,data = chapter_service.add_smart_role_and_voice(project,content,role_names,voice_names)\n # 将data中的每一个元素转化为RoleBindVoiceDTO\n # data = [RoleBindVoiceDTO(**item) for item in data]\n if res:\n return Res(data=data, code=200, message=\"智能匹配成功\")\n else:\n return Res(data=None, code=500, message=\"智能匹配失败\")\n"
},
{
"path": "SonicVale/app/routers/emotion_router.py",
"content": "from typing import List\n\nfrom fastapi import APIRouter, Depends, HTTPException\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.emotion_dto import EmotionResponseDTO, EmotionCreateDTO\nfrom app.entity.emotion_entity import EmotionEntity\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.repositories.emotion_repository import EmotionRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.services.line_service import LineService\nfrom app.services.project_service import ProjectService\nfrom app.services.emotion_service import EmotionService\n\nrouter = APIRouter(prefix=\"/emotions\", tags=[\"Emotions\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_emotion_service(db: Session = Depends(get_db)) -> EmotionService:\n repository = EmotionRepository(db)\n return EmotionService(repository)\n\n@router.post(\"\", response_model=Res[EmotionResponseDTO],\n summary=\"创建情绪枚举\",\n description=\"根据项目ID创建情绪枚举，情绪枚举名称在同一项目下不可重复\" )\ndef create_emotion(dto: EmotionCreateDTO, emotion_service: EmotionService = Depends(get_emotion_service)):\n \"\"\"创建情绪枚举\"\"\"\n try:\n # DTO → Entity\n entity = EmotionEntity(**dto.__dict__)\n\n # 调用 Service 创建项目（返回 True/False）\n entityRes = emotion_service.create_emotion(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = EmotionResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"情绪枚举 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n@router.get(\"/{emotion_id}\", response_model=Res[EmotionResponseDTO],\n summary=\"查询情绪枚举\",\n description=\"根据情绪枚举id查询情绪枚举信息\")\ndef get_emotion(emotion_id: int, emotion_service: EmotionService = Depends(get_emotion_service)):\n entity = emotion_service.get_emotion(emotion_id)\n if entity:\n res = EmotionResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"情绪枚举不存在\")\n\n@router.get(\"\", response_model=Res[List[EmotionResponseDTO]],\n summary=\"查询所有情绪枚举\",\n description=\"根据所有情绪枚举信息\")\ndef get_all_emotions(emotion_service: EmotionService = Depends(get_emotion_service)):\n entities = emotion_service.get_all_emotions()\n if entities:\n res = [EmotionResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=404, message=\"项目不存在情绪枚举\")\n\n# 修改，传入的参数是id\n@router.put(\"/{emotion_id}\", response_model=Res[EmotionCreateDTO],\n summary=\"修改情绪枚举信息\",\n description=\"根据情绪枚举id修改情绪枚举信息,并且不能修改项目id\")\ndef update_emotion(emotion_id: int, dto: EmotionCreateDTO, emotion_service: EmotionService = Depends(get_emotion_service)):\n emotion = emotion_service.get_emotion(emotion_id)\n if emotion is None:\n return Res(data=None, code=404, message=\"情绪枚举不存在\")\n res = emotion_service.update_emotion(emotion_id, dto.dict(exclude_unset=True))\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败,情绪枚举已存在\")\n\n\n# 根据id，删除，不开放\n@router.delete(\"/{emotion_id}\", response_model=Res,\n summary=\"删除情绪枚举\",\n description=\"根据情绪枚举id删除情绪枚举信息\")\ndef delete_emotion(emotion_id: int, emotion_service: EmotionService = Depends(get_emotion_service)):\n success = emotion_service.delete_emotion(emotion_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或情绪枚举不存在\")\n\n\n\n"
},
{
"path": "SonicVale/app/routers/line_router.py",
"content": "import asyncio\nimport os\nimport logging\nimport shutil\nfrom concurrent.futures import ThreadPoolExecutor\nfrom typing import List, Optional\n\nfrom fastapi import APIRouter, Depends, HTTPException, Body, Request, Query\nfrom sqlalchemy.orm import Session\n\nfrom app.core.config import getConfigPath\nfrom app.core.response import Res\nfrom app.core.ws_manager import manager\nfrom app.db.database import get_db, SessionLocal\nfrom app.dto.line_dto import LineResponseDTO, LineCreateDTO, LineOrderDTO, LineAudioProcessDTO\nfrom app.entity.line_entity import LineEntity\nfrom app.repositories.chapter_repository import ChapterRepository\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.repositories.voice_repository import VoiceRepository\nfrom app.services.chapter_service import ChapterService\nfrom app.services.project_service import ProjectService\nfrom app.services.line_service import LineService\nfrom app.services.role_service import RoleService\nfrom app.services.voice_service import VoiceService\n\nrouter = APIRouter(prefix=\"/lines\", tags=[\"Lines\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_line_service(db: Session = Depends(get_db)) -> LineService:\n repository = LineRepository(db)\n role_repository = RoleRepository(db)\n tts_repository = TTSProviderRepository(db)\n llm_repository = LLMProviderRepository(db)\n return LineService(repository, role_repository, tts_repository, llm_repository)\ndef get_project_service(db: Session = Depends(get_db)) -> ProjectService:\n repository = ProjectRepository(db)\n return ProjectService(repository)\n\ndef get_chapter_service(db: Session = Depends(get_db)) -> ChapterService:\n repository = ChapterRepository(db)\n return ChapterService(repository)\n\ndef get_voice_service(db: Session = Depends(get_db)) -> VoiceService:\n repository = VoiceRepository(db)\n multi_emotion_voice_repository = MultiEmotionVoiceRepository(db)\n return VoiceService(repository, multi_emotion_voice_repository)\n\ndef get_role_service(db: Session = Depends(get_db)) -> RoleService:\n repository = RoleRepository(db)\n return RoleService(repository)\n@router.post(\"/{project_id}\", response_model=Res[LineResponseDTO],\n summary=\"创建台词\",\n description=\"根据项目ID创建台词\" )\ndef create_line(project_id:int,dto: LineCreateDTO, line_service: LineService = Depends(get_line_service),\n project_service: ProjectService = Depends(get_project_service),\n chapter_service : ChapterService = Depends(get_chapter_service)):\n \"\"\"创建台词\"\"\"\n try:\n # DTO → Entity\n entity = LineEntity(**dto.__dict__)\n # 判断project_id是否存在\n project = project_service.get_project(project_id)\n if project is None:\n return Res(data=None, code=400, message=f\"项目 '{project_id}' 不存在\")\n\n chapter = chapter_service.get_chapter(dto.chapter_id)\n if chapter is None:\n return Res(data=None, code=400, message=f\"章节 '{dto.chapter_id}' 不存在\")\n # 调用 Service 创建项目（返回 True/False）\n\n entityRes = line_service.create_line(entity)\n\n # 新增台词,这里搞个audio_path\n audio_path = os.path.join(project.project_root_path, str(project_id), str(dto.chapter_id), \"audio\")\n os.makedirs(audio_path, exist_ok=True)\n res_path = os.path.join(audio_path, \"id_\" + str(entityRes.id) + \".wav\")\n line_service.update_line(entityRes.id, {\"audio_path\": res_path})\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = LineResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"台词 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n@router.get(\"/{line_id}\", response_model=Res[LineResponseDTO],\n summary=\"查询台词\",\n description=\"根据台词id查询台词信息\")\ndef get_line(line_id: int, line_service: LineService = Depends(get_line_service)):\n entity = line_service.get_line(line_id)\n if entity:\n res = LineResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"项目不存在\")\n\n@router.get(\"/lines/{chapter_id}\", response_model=Res[List[LineResponseDTO]],\n summary=\"查询章节下的所有台词\",\n description=\"根据章节id查询章节下的所有台词信息\")\ndef get_all_lines(chapter_id: int, line_service: LineService = Depends(get_line_service)):\n entities = line_service.get_all_lines(chapter_id)\n if entities:\n res = [LineResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=200, message=\"章节不存在台词\")\n\n# 修改，传入的参数是id\n@router.put(\"/{line_id}\", response_model=Res[LineCreateDTO],\n summary=\"修改台词信息\",\n description=\"根据台词id修改台词信息,并且不能修改章节id\")\ndef update_line(line_id: int, dto: LineCreateDTO, line_service: LineService = Depends(get_line_service)):\n line = line_service.get_line(line_id)\n if line is None:\n return Res(data=None, code=404, message=\"台词不存在\")\n res = line_service.update_line(line_id, dto.dict(exclude_unset=True))\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败\")\n\n\n# 根据id，删除\n@router.delete(\"/{line_id}\", response_model=Res,\n summary=\"删除台词\",\n description=\"根据台词id删除台词信息\")\ndef delete_line(line_id: int, line_service: LineService = Depends(get_line_service)):\n success = line_service.delete_line(line_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或台词不存在\")\n\n# 删除章节下所有台词\n@router.delete(\"/lines/{chapter_id}\", response_model=Res,summary=\"删除章节下所有台词\",description=\"根据章节id删除章节下的所有台词信息\")\ndef delete_all_lines(chapter_id: int, line_service: LineService = Depends(get_line_service)):\n success = line_service.delete_all_lines(chapter_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或台词不存在\")\n\n\n\n\n\n@router.put(\"/batch/orders\", response_model=Res[bool])\ndef batch_update_line_order(\n line_orders: List[LineOrderDTO] = Body(...), # 关键：明确从 body 读取“数组”\n line_service: LineService = Depends(get_line_service),\n):\n res = line_service.batch_update_line_order(line_orders)\n return Res(data=res, code=200, message=\"更新成功\")\n\n# 完成配音时候，更新音频路径，保证顺序一致\n@router.put(\"/{line_id}/audio_path\", response_model=Res[bool])\ndef update_line_audio_path(\n line_id: int,\n dto: LineCreateDTO, # 关键：明确从 body 读取“数组”\n line_service: LineService = Depends(get_line_service),\n):\n res = line_service.update_audio_path(line_id,dto)\n if not res:\n return Res(data=None, code=400, message=\"更新失败\")\n return Res(data=res, code=200, message=\"更新成功\")\n\n\n\n@router.post(\"/generate-audio/{project_id}/{chapter_id}\")\nasync def generate_audio(request: Request, project_id: int, dto: LineCreateDTO,line_service: LineService = Depends(get_line_service)):\n q = request.app.state.tts_queue # 👈 永远拿到已初始化的同一份队列\n if q.full():\n # 可选：带上 Retry-After 头\n raise HTTPException(status_code=429, detail=\"队列已满，请稍后重试\")\n q.put_nowait((project_id, dto))\n queue_size = q.qsize() # 入队后的队列大小\n line_service.update_line(dto.id, {\"status\": \"processing\"})\n \n # 入队后立即广播队列大小，让前端实时看到更新\n await manager.broadcast({\n \"event\": \"line_update\",\n \"line_id\": dto.id,\n \"status\": \"queued\",\n \"progress\": queue_size,\n \"meta\": f\"已入队，等待生成\"\n })\n \n logging.info(\"队列剩余数量: %s\", queue_size)\n return {\"code\": 200, \"message\": \"已入队\", \"data\": {\"line_id\": dto.id}}\n\n\n# 改为异步任务\n\n# @router.post(\"/generate-audio/{project_id}/{chapter_id}\")\n# async def generate_audio(project_id : int, chapter_id: int, dto: LineCreateDTO):\n# # 立即返回，不阻塞\n# asyncio.create_task(_run_line_tts(project_id,dto))\n# return {\"code\": 200, \"message\": \"已入队\", \"data\": {\"line_id\": dto.id}}\n#\n#\n# TTS_EXECUTOR = ThreadPoolExecutor(max_workers=4) # 线程池大小\n# TTS_SEMAPHORE = asyncio.Semaphore(1) # 最多 4 个并行 TTS\n# async def _run_line_tts(project_id:int,dto: LineCreateDTO):\n# db = SessionLocal()\n# line_service = get_line_service(db)\n# role_service = get_role_service( db)\n# voice_service = get_voice_service(db)\n# project_service = get_project_service(db)\n# try:\n# # 1) 更新为 running\n# line_service.update_line(dto.id, {\"status\": \"processing\"})\n# print(\"开始生成\")\n# await manager.broadcast({\n# \"event\": \"line_update\",\n# \"line_id\": dto.id,\n# \"status\": \"processing\",\n# \"progress\": 0,\n# \"meta\": f\"角色 {dto.role_id} 开始生成\"\n# })\n#\n# # 2) 模拟进度\n# # 获取角色绑定的音色的reference_path\n# role = role_service.get_role(dto.role_id)\n# voice = voice_service.get_voice(role.default_voice_id)\n# project = project_service.get_project(project_id)\n# save_path = dto.audio_path\n# loop = asyncio.get_running_loop()\n# async with TTS_SEMAPHORE:\n# # 可选：设置超时，防挂死\n# try:\n# res = await asyncio.wait_for(\n# loop.run_in_executor(\n# TTS_EXECUTOR, # ✅ 用自建线程池\n# line_service.generate_audio,\n# voice.reference_path,\n# project.tts_provider_id, # 若引擎需要 base_url，就换成 project.tts_base_url\n# dto.text_content,\n# save_path\n# ),\n# timeout=120 # 例：最多等 5 分钟\n# )\n# except asyncio.TimeoutError:\n# raise RuntimeError(\"TTS 超时\")\n#\n# # res = chapter_service.generate_audio(voice.reference_path,project.tts_provider_id,dto.text_content,save_path=save_path)\n# # 3) 真正合成\n# line_service.update_line(dto.id, {\"status\": \"done\"})\n#\n# # 4) 广播完成\n# await manager.broadcast({\n# \"event\": \"line_update\",\n# \"line_id\": dto.id,\n# \"status\": \"done\",\n# \"progress\": 100,\n# \"meta\": \"生成完成\",\n# \"audio_path\": dto.audio_path\n# })\n# except Exception as e:\n# line_service.update_line(dto.id, {\"status\": \"failed\"})\n# await manager.broadcast({\n# \"event\": \"line_update\",\n# \"line_id\": dto.id,\n# \"status\": \"failed\",\n# \"progress\": 0,\n# \"meta\": f\"失败: {e}\"\n# })\n# finally:\n# db.close()\n#\n#\n# # 批量更新line_order\n\n# 处理音频文件，传入倍速，音量大小，以及line_id\n@router.post(\"/process-audio/{line_id}\")\nasync def process_audio(line_id: int, dto: LineAudioProcessDTO, line_service: LineService = Depends(get_line_service)):\n res = line_service.process_audio(line_id,dto)\n if not res:\n return Res(data=None, code=400, message=\"处理失败\")\n return Res(data=res, code=200, message=\"处理成功\")\n\n# 导出音频与字幕\n@router.get(\"/export-audio/{chapter_id}\")\nasync def export_audio(chapter_id: int,\n single: bool = Query(False, description=\"是否导出单条音频字幕\"),\n line_service: LineService = Depends(get_line_service)):\n res = line_service.export_audio(chapter_id, single)\n # res 现在返回 dict，包含 success, message, audio_path 等字段\n if isinstance(res, dict):\n if res.get(\"success\"):\n return Res(data=res, code=200, message=res.get(\"message\", \"导出成功\"))\n else:\n return Res(data=res, code=400, message=res.get(\"message\", \"导出失败\"))\n # 兼容旧的返回格式\n if not res:\n return Res(data=None, code=400, message=\"导出失败\")\n return Res(data=res, code=200, message=\"导出成功\")\n\n\n# 生成单条音频的字幕（已经有音频）\n#\n\n# 矫正字幕 - 拼音匹配矫正\n@router.post(\"/correct-subtitle-pinyin/{chapter_id}\")\nasync def correct_subtitle_pinyin(\n chapter_id: int, \n line_service: LineService = Depends(get_line_service)\n):\n \"\"\"使用拼音匹配算法矫正字幕\"\"\"\n lines = line_service.get_all_lines(chapter_id)\n if not lines:\n logging.info(\"无台词记录\")\n return Res(data=None, code=400, message=\"无台词记录\")\n paths = [line.audio_path for line in lines]\n if not paths or not paths[0]:\n logging.info(\"未找到有效音频路径\")\n return Res(data=None, code=400, message=\"未找到有效音频路径\")\n \n # 读取所有台词，组成一个文本\n text = \"\\n\".join([line.text_content for line in lines])\n output_dir_path = os.path.join(os.path.dirname(paths[0]), \"result\")\n output_subtitle_path = os.path.join(output_dir_path, \"result.srt\")\n \n if not os.path.exists(output_subtitle_path):\n logging.info(\"请先导出音频\")\n return Res(data=None, code=400, message=\"请先导出音频\")\n \n # 拼音矫正输出到独立文件\n pinyin_subtitle_path = os.path.join(output_dir_path, \"result_pinyin.srt\")\n shutil.copy(output_subtitle_path, pinyin_subtitle_path)\n line_service.correct_subtitle_pinyin(text, pinyin_subtitle_path)\n logging.info(\"整体字幕矫正完成（拼音匹配）：%s\", pinyin_subtitle_path)\n\n # 将单条字幕也进行矫正\n logging.info(\"开始对单条字幕进行矫正\")\n for line in lines:\n subtitle_path = line.subtitle_path\n line_text = line.text_content\n if subtitle_path is not None and line_text is not None and os.path.exists(subtitle_path):\n # 单条字幕也输出到 _pinyin 文件\n base, ext = os.path.splitext(subtitle_path)\n pinyin_single_path = f\"{base}_pinyin{ext}\"\n shutil.copy(subtitle_path, pinyin_single_path)\n line_service.correct_subtitle_pinyin(line_text, pinyin_single_path)\n logging.info(\"单条字幕矫正完成：%s\", line.id)\n \n return Res(data=None, code=200, message=\"拼音匹配矫正完成\")\n\n\n# 矫正字幕 - LLM矫正\n@router.post(\"/correct-subtitle-llm/{chapter_id}\")\nasync def correct_subtitle_llm(\n chapter_id: int,\n batch_size: int = Query(20, description=\"LLM分批处理时每批的条数\"),\n line_service: LineService = Depends(get_line_service),\n chapter_service: ChapterService = Depends(get_chapter_service),\n project_service: ProjectService = Depends(get_project_service)\n):\n \"\"\"使用LLM矫正字幕，自动从项目配置获取LLM信息\"\"\"\n # 获取章节信息\n chapter = chapter_service.get_chapter(chapter_id)\n if not chapter:\n return Res(data=None, code=400, message=\"章节不存在\")\n \n # 获取项目信息，从中读取LLM配置\n project = project_service.get_project(chapter.project_id)\n if not project:\n return Res(data=None, code=400, message=\"项目不存在\")\n \n if not project.llm_provider_id:\n return Res(data=None, code=400, message=\"项目未配置LLM提供商，请在项目设置中配置\")\n \n if not project.llm_model:\n return Res(data=None, code=400, message=\"项目未配置LLM模型，请在项目设置中选择模型\")\n \n lines = line_service.get_all_lines(chapter_id)\n if not lines:\n logging.info(\"无台词记录\")\n return Res(data=None, code=400, message=\"无台词记录\")\n paths = [line.audio_path for line in lines]\n if not paths or not paths[0]:\n logging.info(\"未找到有效音频路径\")\n return Res(data=None, code=400, message=\"未找到有效音频路径\")\n \n # 读取所有台词，组成一个文本\n text = \"\\n\".join([line.text_content for line in lines])\n output_dir_path = os.path.join(os.path.dirname(paths[0]), \"result\")\n output_subtitle_path = os.path.join(output_dir_path, \"result.srt\")\n \n if not os.path.exists(output_subtitle_path):\n logging.info(\"请先导出音频\")\n return Res(data=None, code=400, message=\"请先导出音频\")\n \n # LLM矫正输出到独立文件\n llm_subtitle_path = os.path.join(output_dir_path, \"result_llm.srt\")\n shutil.copy(output_subtitle_path, llm_subtitle_path)\n line_service.correct_subtitle_llm(\n text, llm_subtitle_path, \n llm_provider_id=project.llm_provider_id, \n llm_model=project.llm_model, \n batch_size=batch_size\n )\n logging.info(\"整体字幕矫正完成（LLM）：%s\", llm_subtitle_path)\n\n # 将单条字幕也进行矫正\n logging.info(\"开始对单条字幕进行矫正\")\n for line in lines:\n subtitle_path = line.subtitle_path\n line_text = line.text_content\n if subtitle_path is not None and line_text is not None and os.path.exists(subtitle_path):\n # 单条字幕也输出到 _llm 文件\n base, ext = os.path.splitext(subtitle_path)\n llm_single_path = f\"{base}_llm{ext}\"\n shutil.copy(subtitle_path, llm_single_path)\n line_service.correct_subtitle_llm(\n line_text, llm_single_path,\n llm_provider_id=project.llm_provider_id,\n llm_model=project.llm_model,\n batch_size=batch_size\n )\n logging.info(\"单条字幕矫正完成：%s\", line.id)\n \n return Res(data=None, code=200, message=\"LLM矫正完成\")\n\n"
},
{
"path": "SonicVale/app/routers/llm_provider_router.py",
"content": "from fastapi import APIRouter, Depends, HTTPException\nfrom typing import List\n\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.llm_provider_dto import LLMProviderCreateDTO, LLMProviderResponseDTO\nfrom app.entity.llm_provider_entity import LLMProviderEntity\nfrom app.services.llm_provider_service import LLMProviderService\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\n\n# 初始化 router\nrouter = APIRouter(prefix=\"/llm_providers\", tags=[\"LLMProviders\"])\n\n# 依赖注入（实际LLM供应商可用 DI 容器）\n\ndef get_llm_service(db: Session = Depends(get_db)) -> LLMProviderService:\n repository = LLMProviderRepository(db) # ✅ 传入 db\n return LLMProviderService(repository)\n\n\n\n\n@router.post(\"/\", response_model=Res[LLMProviderResponseDTO],\n summary=\"创建LLM供应商\",\n description=\"根据LLM供应商信息创建LLM供应商，LLM供应商名称不可重复\")\ndef create_llm_provider(dto: LLMProviderCreateDTO, service: LLMProviderService = Depends(get_llm_service)):\n \"\"\"\n 创建LLM供应商\n - dto: 前端 POST JSON 传入参数\n - service: Service 层注入\n \"\"\"\n try:\n # DTO → Entity\n entity = LLMProviderEntity(**dto.__dict__)\n\n # 调用 Service 创建LLM供应商（返回 True/False）\n entityRes = service.create_llm_provider(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = LLMProviderResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"LLM供应商 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n# 按id查找\n@router.get(\"/{llm_provider_id}\", response_model=Res[LLMProviderResponseDTO],\n summary=\"查询LLM供应商\",\n description=\"根据LLM供应商ID查询LLM供应商信息\")\ndef get_llm_provider(llm_provider_id: int, service: LLMProviderService = Depends(get_llm_service)):\n entity = service.get_llm_provider(llm_provider_id)\n if entity:\n res = LLMProviderResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"LLM供应商不存在\")\n\n@router.get(\"/\", response_model=Res[List[LLMProviderResponseDTO]],\n summary=\"查询所有LLM供应商\",\n description=\"查询所有LLM供应商信息\")\ndef get_all_llm_providers(service: LLMProviderService = Depends(get_llm_service)):\n entities = service.get_all_llm_providers()\n dtos = [LLMProviderResponseDTO(**e.__dict__) for e in entities]\n return Res(data=dtos, code=200, message=\"查询成功\")\n\n\n# ------------------- 修改LLM供应商 -------------------\n@router.put(\"/{llm_provider_id}\", response_model=Res[LLMProviderCreateDTO],\n summary=\"修改LLM供应商\",\n description=\"根据LLM供应商ID修改LLM供应商信息\")\ndef update_llm_provider(llm_provider_id: int, dto: LLMProviderCreateDTO, service: LLMProviderService = Depends(get_llm_service)):\n\n # 先根据id进行查找\n llm_provider = service.get_llm_provider(llm_provider_id)\n if not llm_provider:\n return Res(data=None, code=400, message=\"LLM供应商不存在\")\n\n success = service.update_llm_provider(llm_provider_id,dto.dict(exclude_unset=True))\n if success:\n return Res(data=dto, code=200, message=\"更新成功\")\n else:\n return Res(data=None, code=400, message=\"更新失败\")\n\n\n# ------------------- 删除LLM供应商 -------------------\n@router.delete(\"/{llm_provider_id}\", response_model=Res,\n summary=\"删除LLM供应商\",\n description=\"根据LLM供应商ID删除LLM供应商,并且级联删除LLM供应商下所有章节以及内容\")\ndef delete_llm_provider(llm_provider_id: int, service: LLMProviderService = Depends(get_llm_service)):\n success = service.delete_llm_provider(llm_provider_id)\n # todo 级联删除LLM供应商所有相关内容，比如LLM供应商下所有章节以及内容\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或LLM供应商不存在\")\n\n\n# 测试供应商\n@router.post(\"/test\", response_model=Res)\ndef test_llm_provider(dto: LLMProviderCreateDTO, service: LLMProviderService = Depends(get_llm_service)):\n \"\"\"\n 测试供应商\n \"\"\"\n entity = LLMProviderEntity(**dto.__dict__)\n res,msg = service.test_llm_provider(entity)\n if res == True:\n return Res(data=None, code=200, message=\"测试成功\")\n else:\n return Res(data=None, code=400, message= msg)"
},
{
"path": "SonicVale/app/routers/multi_emotion_voice_router.py",
"content": "from typing import List\n\nfrom fastapi import APIRouter, Depends\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.multi_emotion_voice_dto import MultiEmotionVoiceCreateDTO, MultiEmotionVoiceResponseDTO\nfrom app.entity.multi_emotion_voice_entity import MultiEmotionVoiceEntity\nfrom app.repositories.emotion_repository import EmotionRepository\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\nfrom app.repositories.strength_repository import StrengthRepository\nfrom app.repositories.voice_repository import VoiceRepository\nfrom app.services.emotion_service import EmotionService\nfrom app.services.multi_emotion_voice_service import MultiEmotionVoiceService\nfrom app.services.strength_service import StrengthService\nfrom app.services.voice_service import VoiceService\n\nrouter = APIRouter(prefix=\"/multi_emotion_voices\", tags=[\"MultiEmotionVoice\"])\n\ndef get_multi_emotion_voice_service(db: Session = Depends(get_db)) -> MultiEmotionVoiceService:\n repository = MultiEmotionVoiceRepository(db)\n return MultiEmotionVoiceService(repository)\ndef get_voice_service(db: Session = Depends(get_db)) -> VoiceService:\n repository = VoiceRepository(db)\n multi_emotion_voice_repository = MultiEmotionVoiceRepository(db)\n return VoiceService(repository, multi_emotion_voice_repository)\n\ndef get_emotion_service(db: Session = Depends(get_db)) -> EmotionService:\n repository = EmotionRepository(db)\n return EmotionService(repository)\n\ndef get_strength_service(db: Session = Depends(get_db)) -> StrengthService:\n repository = StrengthRepository(db)\n return StrengthService(repository)\n\n# 根据voice_id获取多音色\n@router.get(\"/voice_id/{voice_id}\", response_model=Res[List[MultiEmotionVoiceResponseDTO]],summary=\"根据voice_id获取多音色\", description=\"根据voice_id获取多音色\")\ndef get_multi_emotion_voice_by_voice_id(voice_id: int, multi_emotion_voice_service: MultiEmotionVoiceService = Depends(get_multi_emotion_voice_service),\n voice_service: VoiceService = Depends(get_voice_service)):\n # 应该查询voice\n voice = voice_service.get_voice(voice_id)\n if voice is None:\n return Res(code=404, message=\"音色不存在\")\n entities = multi_emotion_voice_service.get_multi_emotion_voice_by_voice_id(voice_id)\n if entities is None:\n return Res(code=404, message=\"多音色不存在\")\n else:\n res = [MultiEmotionVoiceResponseDTO(**entity.__dict__) for entity in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n\n# 查询所有多音色\n@router.get(\"\", response_model=Res[List[MultiEmotionVoiceResponseDTO]],summary=\"查询所有多音色\", description=\"查询所有多音色\")\ndef get_all_multi_emotion_voice(multi_emotion_voice_service: MultiEmotionVoiceService = Depends(get_multi_emotion_voice_service)):\n entities = multi_emotion_voice_service.get_all_multi_emotion_voices()\n if not entities:\n return Res(data=[], code=200, message=\"查询成功\")\n entities = [MultiEmotionVoiceResponseDTO(**e.__dict__) for e in entities]\n return Res(data=entities, code=200, message=\"查询成功\")\n# 创建\n@router.post(\"\", response_model=Res[MultiEmotionVoiceResponseDTO],summary=\"创建多情绪音色\", description=\"创建多情绪音色\")\ndef create_multi_emotion_voice(dto: MultiEmotionVoiceCreateDTO, multi_emotion_voice_service: MultiEmotionVoiceService = Depends(get_multi_emotion_voice_service),\n voice_service: VoiceService = Depends(get_voice_service),\n emotion_service: EmotionService = Depends(get_emotion_service),\n strength_service: StrengthService = Depends(get_strength_service)):\n \"\"\"创建多音色\"\"\"\n # 先要判断voice是否存在\n voice = voice_service.get_voice(dto.voice_id)\n # 判断情绪枚举是否存在\n emotion = emotion_service.get_emotion(dto.emotion_id)\n # 判断强度枚举是否存在\n strength = strength_service.get_strength(dto.strength_id)\n if voice is None or emotion is None or strength is None:\n return Res(code=500, message=\"创建失败,音色或者情绪枚举或者情绪强弱枚举不存在，不能创建多情绪音色\")\n # DTO → Entity\n entity = MultiEmotionVoiceEntity(**dto.__dict__)\n entity = multi_emotion_voice_service.create_multi_emotion_voice(entity)\n if entity is None:\n return Res(code=500, message=\"创建失败,已存在多情绪音色\")\n else :\n entity = MultiEmotionVoiceResponseDTO(**entity.__dict__)\n return Res(data=entity, code=200, message=\"创建成功\")\n\n\n# 修改\n@router.put(\"/{multi_emotion_voice_id}\", response_model=Res[MultiEmotionVoiceCreateDTO],summary=\"修改多情绪音色\", description=\"修改多情绪音色\")\ndef update_multi_emotion_voice(multi_emotion_voice_id: int, dto: MultiEmotionVoiceCreateDTO, multi_emotion_voice_service: MultiEmotionVoiceService = Depends(get_multi_emotion_voice_service)):\n \"\"\"修改多音色\"\"\"\n entity = multi_emotion_voice_service.get_multi_emotion_voice_by_id(multi_emotion_voice_id)\n if entity is None:\n return Res(code=404, message=\"多音色不存在\")\n res = multi_emotion_voice_service.update_multi_emotion_voice(multi_emotion_voice_id, dto.dict(exclude_unset=True))\n if res is None:\n return Res(code=500, message=\"修改失败\")\n else:\n entityRes = MultiEmotionVoiceResponseDTO(**entity.__dict__)\n return Res(data=entityRes, code=200, message=\"修改成功\")\n\n# 删除\n@router.delete(\"/{multi_emotion_voice_id}\", response_model=Res[MultiEmotionVoiceResponseDTO],summary=\"删除多情绪音色\", description=\"删除多情绪音色\")\ndef delete_multi_emotion_voice(multi_emotion_voice_id: int, multi_emotion_voice_service: MultiEmotionVoiceService = Depends(get_multi_emotion_voice_service)):\n \"\"\"删除多音色\"\"\"\n res = multi_emotion_voice_service.delete_multi_emotion_voice(multi_emotion_voice_id)\n if res:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败\")\n\n#\n"
},
{
"path": "SonicVale/app/routers/project_router.py",
"content": "import os\nimport shutil\nimport logging\n\nfrom fastapi import APIRouter, Depends, HTTPException\nfrom typing import List\n\nfrom sqlalchemy.orm import Session\n\nfrom app.core.config import getConfigPath\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.project_dto import ProjectCreateDTO, ProjectResponseDTO, ProjectImportDTO\nfrom app.entity.chapter_entity import ChapterEntity\nfrom app.entity.project_entity import ProjectEntity\nfrom app.models.po import ChapterPO\nfrom app.repositories.chapter_repository import ChapterRepository\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.services.chapter_service import ChapterService\nfrom app.services.project_service import ProjectService\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.services.role_service import RoleService\n\n# 初始化 router\nrouter = APIRouter(prefix=\"/projects\", tags=[\"Projects\"])\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_service(db: Session = Depends(get_db)) -> ProjectService:\n repository = ProjectRepository(db) # ✅ 传入 db\n return ProjectService(repository)\n\ndef get_chapter_service(db: Session = Depends(get_db)) -> ChapterService:\n repository = ChapterRepository(db) # ✅ 传入 db\n return ChapterService(repository)\n\ndef get_role_service(db: Session = Depends(get_db)) -> RoleService:\n repository = RoleRepository(db) # ✅ 传入 db\n return RoleService(repository)\n\n\n@router.post(\"/\", response_model=Res[ProjectResponseDTO],\n summary=\"创建项目\",\n description=\"根据项目信息创建项目，项目名称不可重复\")\ndef create_project(dto: ProjectCreateDTO, service: ProjectService = Depends(get_service)):\n \"\"\"\n 创建项目\n - dto: 前端 POST JSON 传入参数\n - service: Service 层注入\n \"\"\"\n try:\n # DTO → Entity\n entity = ProjectEntity(**dto.__dict__)\n\n # 调用 Service 创建项目（返回 True/False）\n entityRes,message = service.create_project(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = ProjectResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=message)\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n# 按id查找\n@router.get(\"/{project_id}\", response_model=Res[ProjectResponseDTO],\n summary=\"查询项目\",\n description=\"根据项目ID查询项目信息\")\ndef get_project(project_id: int, service: ProjectService = Depends(get_service)):\n entity = service.get_project(project_id)\n if entity:\n res = ProjectResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"项目不存在\")\n\n@router.get(\"/\", response_model=Res[List[ProjectResponseDTO]],\n summary=\"查询所有项目\",\n description=\"查询所有项目信息\")\ndef get_all_projects(service: ProjectService = Depends(get_service)):\n entities = service.get_all_projects()\n dtos = [ProjectResponseDTO(**e.__dict__) for e in entities]\n return Res(data=dtos, code=200, message=\"查询成功\")\n\n\n# ------------------- 修改项目 -------------------\n@router.put(\"/{project_id}\", response_model=Res[ProjectCreateDTO],\n summary=\"修改项目\",\n description=\"根据项目ID修改项目信息\")\ndef update_project(project_id: int, dto: ProjectCreateDTO, service: ProjectService = Depends(get_service)):\n\n # 先根据id进行查找\n project = service.get_project(project_id)\n if not project:\n return Res(data=None, code=400, message=\"项目不存在\")\n\n success = service.update_project(project_id,dto.dict())\n if success:\n return Res(data=dto, code=200, message=\"更新成功\")\n else:\n return Res(data=None, code=400, message=\"更新失败\")\n\n\n# ------------------- 删除项目 -------------------\n@router.delete(\"/{project_id}\", response_model=Res,\n summary=\"删除项目\",\n description=\"根据项目ID删除项目,并且级联删除项目下所有章节以及内容\")\ndef delete_project(project_id: int, service: ProjectService = Depends(get_service), chapter_service: ChapterService = Depends(get_chapter_service),role_service: RoleService = Depends(get_role_service)):\n\n # 级联删除项目所有相关内容，比如项目下所有章节以及内容\n entities = chapter_service.get_all_chapters(project_id)\n for entity in entities:\n chapter_service.delete_chapter(entity.id)\n # 删除project目录\n project = service.get_project(project_id)\n\n project_path = os.path.join(project.project_root_path, str(project_id))\n if os.path.exists(project_path):\n shutil.rmtree(project_path) # 删除整个文件夹及其所有内容\n logging.info(\"已删除目录及内容: %s\", project_path)\n else:\n logging.info(\"目录不存在: %s\", project_path)\n\n # 还要删除角色库中projet下的所有角色\n roles = role_service.get_all_roles(project_id)\n for role in roles:\n role_service.delete_role(role.id)\n success = service.delete_project(project_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或项目不存在\")\n\n# 直接导入整本小说内容，然后解析，创建章节\n@router.post(\"/{project_id}/import\")\ndef import_project(project_id: int, dto: ProjectImportDTO,service: ProjectService = Depends(get_service),\n chapter_service: ChapterService = Depends(get_chapter_service)):\n\n content = dto.content\n # 删除该项目下的所有章节\n # chapters = chapter_service.get_all_chapters(project_id)\n # for chapter in chapters:\n # chapter_service.delete_chapter(chapter.id)\n # 解析content\n chapter_contents = service.parse_content(content)\n if len(chapter_contents) == 0:\n return Res(code=400, message=\"导入失败\")\n\n # 批量创建章节\n for chapter_content in chapter_contents:\n name = chapter_content[\"chapter_name\"]\n content = chapter_content[\"content\"]\n logging.info(\"批量创建章节 %s\", name)\n chapter_service.create_chapter(ChapterEntity(project_id=project_id, title=name, text_content=content))\n return Res(code=200, message=\"导入成功\")\n"
},
{
"path": "SonicVale/app/routers/prompt_router.py",
"content": "from fastapi import APIRouter, Depends, HTTPException\nfrom typing import List\n\nfrom sqlalchemy.orm import Session\n\nfrom app.core.enums import TaskEnum\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.prompt_dto import PromptCreateDTO, PromptResponseDTO\nfrom app.entity.prompt_entity import PromptEntity\nfrom app.services.prompt_service import PromptService\nfrom app.repositories.prompt_repository import PromptRepository\n\n# 初始化 router\nrouter = APIRouter(prefix=\"/prompts\", tags=[\"Prompts\"])\n\n# 依赖注入（实际提示词可用 DI 容器）\n\ndef get_service(db: Session = Depends(get_db)) -> PromptService:\n repository = PromptRepository(db) # ✅ 传入 db\n return PromptService(repository)\n\n\n\n\n@router.post(\"/\", response_model=Res[PromptResponseDTO],\n summary=\"创建提示词\",\n description=\"根据提示词信息创建提示词，提示词名称不可重复\")\ndef create_prompt(dto: PromptCreateDTO, service: PromptService = Depends(get_service)):\n \"\"\"\n 创建提示词\n - dto: 前端 POST JSON 传入参数\n - service: Service 层注入\n \"\"\"\n try:\n # DTO → Entity\n entity = PromptEntity(**dto.__dict__)\n\n # 调用 Service 创建提示词（返回 True/False）\n entityRes = service.create_prompt(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = PromptResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"创建失败,可能是不存在该任务或提示词数据不完整\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n# 按id查找\n@router.get(\"/{prompt_id}\", response_model=Res[PromptResponseDTO],\n summary=\"查询提示词\",\n description=\"根据提示词ID查询提示词信息\")\ndef get_prompt(prompt_id: int, service: PromptService = Depends(get_service)):\n entity = service.get_prompt(prompt_id)\n if entity:\n res = PromptResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"提示词不存在\")\n\n@router.get(\"/\", response_model=Res[List[PromptResponseDTO]],\n summary=\"查询所有提示词\",\n description=\"查询所有提示词信息\")\ndef get_all_prompts(service: PromptService = Depends(get_service)):\n entities = service.get_all_prompts()\n dtos = [PromptResponseDTO(**e.__dict__) for e in entities]\n return Res(data=dtos, code=200, message=\"查询成功\")\n\n\n# ------------------- 修改提示词 -------------------\n@router.put(\"/{prompt_id}\", response_model=Res[PromptCreateDTO],\n summary=\"修改提示词\",\n description=\"根据提示词ID修改提示词信息\")\ndef update_prompt(prompt_id: int, dto: PromptCreateDTO, service: PromptService = Depends(get_service)):\n\n # 先根据id进行查找\n prompt = service.get_prompt(prompt_id)\n if not prompt:\n return Res(data=None, code=400, message=\"提示词不存在\")\n\n success = service.update_prompt(prompt_id,dto.dict(exclude_unset=True))\n if success:\n return Res(data=dto, code=200, message=\"更新成功\")\n else:\n return Res(data=None, code=400, message=\"更新失败,可能是不存在该任务或提示词数据不完整\")\n\n\n# ------------------- 删除提示词 -------------------\n@router.delete(\"/{prompt_id}\", response_model=Res,\n summary=\"删除提示词\",\n description=\"根据提示词ID删除提示词,并且级联删除提示词下所有章节以及内容\")\ndef delete_prompt(prompt_id: int, service: PromptService = Depends(get_service)):\n success = service.delete_prompt(prompt_id)\n # todo 级联删除提示词所有相关内容，比如提示词下所有章节以及内容\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或提示词不存在\")\n\n# 获取所有的任务列表\n@router.get(\"/tasks/all\", response_model=Res[List[str]])\ndef get_all_tasks(service: PromptService = Depends(get_service)):\n tasks = service.get_all_tasks()\n return Res(data=tasks, code=200, message=\"查询成功\")\n\n# 根据任务列表获取对应的提示词\n@router.get(\"/tasks/by\", response_model=Res[List[PromptResponseDTO]])\ndef get_prompt_by_task(task: TaskEnum, service: PromptService = Depends(get_service)):\n prompts = service.get_prompt_by_task(task.value) # 取枚举的值\n dtos = [PromptResponseDTO(**e.__dict__) for e in prompts]\n return Res(data=dtos, code=200, message=\"查询成功\")\n\n\n\n# 测试供应商\n# @router.post(\"/test\", response_model=Res)\n# def test_prompt(dto: PromptCreateDTO, service: PromptService = Depends(get_service)):\n# \"\"\"\n# 测试供应商\n# \"\"\"\n# entity = PromptEntity(**dto.__dict__)\n# success = service.test_prompt(entity)\n# if success:\n# return Res(data=None, code=200, message=\"测试成功\")\n# else:\n# return Res(data=None, code=400, message=\"测试失败\")"
},
{
"path": "SonicVale/app/routers/role_router.py",
"content": "from typing import List\n\nfrom fastapi import APIRouter, Depends, HTTPException\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.role_dto import RoleResponseDTO, RoleCreateDTO\nfrom app.entity.role_entity import RoleEntity\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.services.line_service import LineService\nfrom app.services.project_service import ProjectService\nfrom app.services.role_service import RoleService\n\nrouter = APIRouter(prefix=\"/roles\", tags=[\"Roles\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_role_service(db: Session = Depends(get_db)) -> RoleService:\n repository = RoleRepository(db)\n return RoleService(repository)\ndef get_project_service(db: Session = Depends(get_db)) -> ProjectService:\n repository = ProjectRepository(db)\n return ProjectService(repository)\n\ndef get_line_service(db: Session = Depends(get_db)) -> LineService:\n repository = LineRepository(db)\n role_repository = RoleRepository(db)\n tts_provider_repository = TTSProviderRepository(db)\n llm_provider_repository = LLMProviderRepository(db)\n return LineService(repository, role_repository, tts_provider_repository, llm_provider_repository)\n@router.post(\"\", response_model=Res[RoleResponseDTO],\n summary=\"创建角色\",\n description=\"根据项目ID创建角色，角色名称在同一项目下不可重复\" )\ndef create_role(dto: RoleCreateDTO, role_service: RoleService = Depends(get_role_service),\n project_service: ProjectService = Depends(get_project_service)):\n \"\"\"创建角色\"\"\"\n try:\n # DTO → Entity\n entity = RoleEntity(**dto.__dict__)\n # 判断project_id是否存在\n project = project_service.get_project(dto.project_id)\n if project is None:\n return Res(data=None, code=400, message=f\"项目 '{dto.project_id}' 不存在\")\n # 调用 Service 创建项目（返回 True/False）\n entityRes = role_service.create_role(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = RoleResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"角色 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n@router.get(\"/{role_id}\", response_model=Res[RoleResponseDTO],\n summary=\"查询角色\",\n description=\"根据角色id查询角色信息\")\ndef get_role(role_id: int, role_service: RoleService = Depends(get_role_service)):\n entity = role_service.get_role(role_id)\n if entity:\n res = RoleResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"项目不存在\")\n\n@router.get(\"/project/{project_id}\", response_model=Res[List[RoleResponseDTO]],\n summary=\"查询项目下的所有角色\",\n description=\"根据项目id查询项目下的所有角色信息\")\ndef get_all_roles(project_id: int, role_service: RoleService = Depends(get_role_service)):\n entities = role_service.get_all_roles(project_id)\n if entities:\n res = [RoleResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=404, message=\"项目不存在角色\")\n\n# 修改，传入的参数是id\n@router.put(\"/{role_id}\", response_model=Res[RoleCreateDTO],\n summary=\"修改角色信息\",\n description=\"根据角色id修改角色信息,并且不能修改项目id\")\ndef update_role(role_id: int, dto: RoleCreateDTO, role_service: RoleService = Depends(get_role_service)):\n role = role_service.get_role(role_id)\n if role is None:\n return Res(data=None, code=404, message=\"角色不存在\")\n res = role_service.update_role(role_id, dto.dict(exclude_unset=True))\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败\")\n\n\n# 根据id，删除\n@router.delete(\"/{role_id}\", response_model=Res,\n summary=\"删除角色\",\n description=\"根据角色id删除角色信息\")\ndef delete_role(role_id: int, role_service: RoleService = Depends(get_role_service),line_service: LineService = Depends(get_line_service)):\n success = role_service.delete_role(role_id)\n if success:\n # 获取改角色下所有的台词\n line_service.clear_role_id(role_id)\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或角色不存在\")\n\n\n# 根据内容进行解析\n\n"
},
{
"path": "SonicVale/app/routers/strength_router.py",
"content": "from typing import List\n\nfrom fastapi import APIRouter, Depends, HTTPException\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.strength_dto import StrengthResponseDTO, StrengthCreateDTO\nfrom app.entity.strength_entity import StrengthEntity\n\nfrom app.repositories.strength_repository import StrengthRepository\n\nfrom app.services.strength_service import StrengthService\n\nrouter = APIRouter(prefix=\"/strengths\", tags=[\"Strengths\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_strength_service(db: Session = Depends(get_db)) -> StrengthService:\n repository = StrengthRepository(db)\n return StrengthService(repository)\n\n@router.post(\"\", response_model=Res[StrengthResponseDTO],\n summary=\"创建情绪强弱枚举\",\n description=\"根据项目ID创建情绪强弱枚举，情绪强弱枚举名称在同一项目下不可重复\" )\ndef create_strength(dto: StrengthCreateDTO, strength_service: StrengthService = Depends(get_strength_service)):\n \"\"\"创建情绪强弱枚举\"\"\"\n try:\n # DTO → Entity\n entity = StrengthEntity(**dto.__dict__)\n\n # 调用 Service 创建项目（返回 True/False）\n entityRes = strength_service.create_strength(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = StrengthResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"情绪强弱枚举 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n@router.get(\"/{strength_id}\", response_model=Res[StrengthResponseDTO],\n summary=\"查询情绪强弱枚举\",\n description=\"根据情绪强弱枚举id查询情绪强弱枚举信息\")\ndef get_strength(strength_id: int, strength_service: StrengthService = Depends(get_strength_service)):\n entity = strength_service.get_strength(strength_id)\n if entity:\n res = StrengthResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"情绪强弱枚举不存在\")\n\n@router.get(\"\", response_model=Res[List[StrengthResponseDTO]],\n summary=\"查询所有情绪强弱枚举\",\n description=\"根据所有情绪强弱枚举信息\")\ndef get_all_strengths(strength_service: StrengthService = Depends(get_strength_service)):\n entities = strength_service.get_all_strengths()\n if entities:\n res = [StrengthResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=404, message=\"项目不存在情绪强弱枚举\")\n\n# 修改，传入的参数是id\n@router.put(\"/{strength_id}\", response_model=Res[StrengthCreateDTO],\n summary=\"修改情绪强弱枚举信息\",\n description=\"根据情绪强弱枚举id修改情绪强弱枚举信息,并且不能修改项目id\")\ndef update_strength(strength_id: int, dto: StrengthCreateDTO, strength_service: StrengthService = Depends(get_strength_service)):\n strength = strength_service.get_strength(strength_id)\n if strength is None:\n return Res(data=None, code=404, message=\"情绪强弱枚举不存在\")\n res = strength_service.update_strength(strength_id, dto.dict(exclude_unset=True))\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败，情绪强弱枚举名已存在\")\n\n\n# 根据id，删除，不开放\n@router.delete(\"/{strength_id}\", response_model=Res,\n summary=\"删除情绪强弱枚举\",\n description=\"根据情绪强弱枚举id删除情绪强弱枚举信息\")\ndef delete_strength(strength_id: int, strength_service: StrengthService = Depends(get_strength_service)):\n success = strength_service.delete_strength(strength_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或情绪强弱枚举不存在\")\n\n\n\n"
},
{
"path": "SonicVale/app/routers/tts_provider_router.py",
"content": "from fastapi import APIRouter, Depends, HTTPException\nfrom typing import List\n\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.tts_provider_dto import TTSProviderCreateDTO, TTSProviderResponseDTO\nfrom app.entity.tts_provider_entity import TTSProviderEntity\nfrom app.services.tts_provider_service import TTSProviderService\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\n\n# 初始化 router\nrouter = APIRouter(prefix=\"/tts_providers\", tags=[\"TTSProviders\"])\n\n# 依赖注入（实际TTS供应商可用 DI 容器）\n\ndef get_service(db: Session = Depends(get_db)) -> TTSProviderService:\n repository = TTSProviderRepository(db) # ✅ 传入 db\n return TTSProviderService(repository)\n\n\n# 按id查找\n@router.get(\"/{tts_provider_id}\", response_model=Res[TTSProviderResponseDTO],\n summary=\"查询TTS供应商\",\n description=\"根据TTS供应商ID查询TTS供应商信息\")\ndef get_tts_provider(tts_provider_id: int, service: TTSProviderService = Depends(get_service)):\n entity = service.get_tts_provider(tts_provider_id)\n if entity:\n res = TTSProviderResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"TTS供应商不存在\")\n\n@router.get(\"/\", response_model=Res[List[TTSProviderResponseDTO]],\n summary=\"查询所有TTS供应商\",\n description=\"查询所有TTS供应商信息\")\ndef get_all_tts_providers(service: TTSProviderService = Depends(get_service)):\n entities = service.get_all_tts_providers()\n dtos = [TTSProviderResponseDTO(**e.__dict__) for e in entities]\n return Res(data=dtos, code=200, message=\"查询成功\")\n\n\n# ------------------- 修改TTS供应商 -------------------\n@router.put(\"/{tts_provider_id}\", response_model=Res[TTSProviderCreateDTO],\n summary=\"修改TTS供应商\",\n description=\"根据TTS供应商ID修改TTS供应商信息\")\ndef update_tts_provider(tts_provider_id: int, dto: TTSProviderCreateDTO, service: TTSProviderService = Depends(get_service)):\n\n # 先根据id进行查找\n tts_provider = service.get_tts_provider(tts_provider_id)\n if not tts_provider:\n return Res(data=None, code=400, message=\"TTS供应商不存在\")\n\n success = service.update_tts_provider(tts_provider_id,dto.dict(exclude_unset=True))\n if success:\n return Res(data=dto, code=200, message=\"更新成功\")\n else:\n return Res(data=None, code=400, message=\"更新失败\")\n\n\n\n# 测试tts是否正常\n@router.post(\"/test\", response_model=Res)\ndef test_tts_provider(dto: TTSProviderCreateDTO, service: TTSProviderService = Depends(get_service)):\n \"\"\"\n 测试tts是否正常\n \"\"\"\n entity = TTSProviderEntity(**dto.dict())\n success = service.test_tts_provider(entity)\n if success:\n return Res(data=None, code=200, message=\"测试成功\")\n else:\n return Res(data=None, code=400, message=\"测试失败\")\n\n\n\n# ------------------- 删除TTS供应商 -------------------\n# @router.delete(\"/{tts_provider_id}\", response_model=Res,\n# summary=\"删除TTS供应商\",\n# description=\"根据TTS供应商ID删除TTS供应商,并且级联删除TTS供应商下所有章节以及内容\")\n# def delete_tts_provider(tts_provider_id: int, service: TTSProviderService = Depends(get_service)):\n# success = service.delete_tts_provider(tts_provider_id)\n# # todo 级联删除TTS供应商所有相关内容，比如TTS供应商下所有章节以及内容\n# if success:\n# return Res(data=None, code=200, message=\"删除成功\")\n# else:\n# return Res(data=None, code=400, message=\"删除失败或TTS供应商不存在\")"
},
{
"path": "SonicVale/app/routers/voice_router.py",
"content": "from typing import List\n\nfrom fastapi import APIRouter, Depends, HTTPException\nfrom fastapi.responses import FileResponse\nfrom sqlalchemy.orm import Session\n\nfrom app.core.response import Res\nfrom app.db.database import get_db\nfrom app.dto.tts_provider_dto import TTSProviderResponseDTO\nfrom app.dto.voice_dto import VoiceResponseDTO, VoiceCreateDTO, VoiceExportDTO, VoiceImportDTO, VoiceImportResultDTO, VoiceAudioProcessDTO, VoiceCopyDTO\nfrom app.entity.voice_entity import VoiceEntity\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\n\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.repositories.voice_repository import VoiceRepository\n\nfrom app.services.tts_provider_service import TTSProviderService\nfrom app.services.voice_service import VoiceService\n\nrouter = APIRouter(prefix=\"/voices\", tags=[\"Voices\"])\n\n\n# 依赖注入（实际项目可用 DI 容器）\n\ndef get_voice_service(db: Session = Depends(get_db)) -> VoiceService:\n repository = VoiceRepository(db)\n multi_emotion_voice_repository = MultiEmotionVoiceRepository(db)\n return VoiceService(repository, multi_emotion_voice_repository)\ndef get_tts_provider_service(db: Session = Depends(get_db)) -> TTSProviderService:\n repository = TTSProviderRepository(db)\n return TTSProviderService(repository)\n\n\n# ====== 静态路由放在动态路由之前，避免路径冲突 ======\n\n@router.post(\"/process-audio\", response_model=Res[str],\n summary=\"处理音色参考音频\",\n description=\"对音色的参考音频进行处理（变速、音量、裁剪等）\")\ndef process_voice_audio(dto: VoiceAudioProcessDTO, voice_service: VoiceService = Depends(get_voice_service)):\n \"\"\"处理音色参考音频\"\"\"\n try:\n result = voice_service.process_audio(dto)\n if result:\n return Res(data=dto.audio_path, code=200, message=\"处理成功\")\n else:\n return Res(data=None, code=400, message=\"处理失败\")\n except FileNotFoundError as e:\n return Res(data=None, code=404, message=f\"音频文件不存在: {str(e)}\")\n except Exception as e:\n return Res(data=None, code=500, message=f\"处理失败: {str(e)}\")\n\n\n@router.post(\"/export\", response_model=Res[str],\n summary=\"导出音色库\",\n description=\"将指定TTS供应商下的音色打包到zip文件（可选传ids仅导出选中）\")\ndef export_voices(dto: VoiceExportDTO, voice_service: VoiceService = Depends(get_voice_service)):\n \"\"\"导出音色库到zip文件\"\"\"\n try:\n result = voice_service.export_voices(dto.tts_provider_id, dto.export_path, dto.ids)\n if result:\n return Res(data=result, code=200, message=\"导出成功\")\n else:\n return Res(data=None, code=400, message=\"没有可导出的音色\")\n except Exception as e:\n return Res(data=None, code=500, message=f\"导出失败: {str(e)}\")\n\n\n@router.post(\"/import\", response_model=Res[VoiceImportResultDTO],\n summary=\"导入音色库\",\n description=\"从zip文件导入音色库，将音频文件复制到指定目录，已存在的音色会跳过\")\ndef import_voices(dto: VoiceImportDTO, voice_service: VoiceService = Depends(get_voice_service)):\n \"\"\"从zip文件导入音色库\"\"\"\n try:\n success_count, skipped_count, skipped_names = voice_service.import_voices(\n dto.tts_provider_id, dto.zip_path, dto.target_dir\n )\n result = VoiceImportResultDTO(\n success_count=success_count,\n skipped_count=skipped_count,\n skipped_names=skipped_names\n )\n return Res(data=result, code=200, message=f\"导入完成：成功{success_count}个，跳过{skipped_count}个\")\n except FileNotFoundError as e:\n return Res(data=None, code=404, message=str(e))\n except ValueError as e:\n return Res(data=None, code=400, message=str(e))\n except Exception as e:\n return Res(data=None, code=500, message=f\"导入失败: {str(e)}\")\n\n\n@router.post(\"/copy\", response_model=Res[VoiceResponseDTO],\n summary=\"复制音色\",\n description=\"复制现有音色，包括音频文件，生成新的音色记录\")\ndef copy_voice(dto: VoiceCopyDTO, voice_service: VoiceService = Depends(get_voice_service)):\n \"\"\"复制音色\"\"\"\n try:\n new_voice = voice_service.copy_voice(\n dto.source_voice_id, dto.new_name, dto.target_dir\n )\n res = VoiceResponseDTO(**new_voice.__dict__)\n return Res(data=res, code=200, message=\"复制成功\")\n except ValueError as e:\n return Res(data=None, code=400, message=str(e))\n except Exception as e:\n return Res(data=None, code=500, message=f\"复制失败: {str(e)}\")\n\n\n@router.get(\"/tts/{tts_provider_id}\", response_model=Res[List[VoiceResponseDTO]],\n summary=\"查询tts供应商下的所有音色\",\n description=\"根据tts供应商id,查询tts供应商下的所有音色信息\")\ndef get_all_voices(tts_provider_id: int, voice_service: VoiceService = Depends(get_voice_service)):\n entities = voice_service.get_all_voices(tts_provider_id)\n if entities:\n res = [VoiceResponseDTO(**e.__dict__) for e in entities]\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=[], code=404, message=\"项目不存在音色\")\n\n\n@router.post(\"\", response_model=Res[VoiceResponseDTO],\n summary=\"创建音色\",\n description=\"根据项目ID创建音色，音色名称在同一项目下不可重复\" )\ndef create_voice(dto: VoiceCreateDTO, voice_service: VoiceService = Depends(get_voice_service),\n tts_provider_service: TTSProviderService = Depends(get_tts_provider_service)):\n \"\"\"创建音色\"\"\"\n try:\n # DTO → Entity\n entity = VoiceEntity(**dto.__dict__)\n # 判断tts_id是否存在\n tts_provider = tts_provider_service.get_tts_provider(dto.tts_provider_id)\n\n if tts_provider is None:\n return Res(data=None, code=400, message=f\"tts服务提供商 '{dto.tts_provider_id}' 不存在\")\n # 调用 Service 创建项目（返回 True/False）\n entityRes = voice_service.create_voice(entity)\n\n # 返回统一 Response\n if entityRes is not None:\n # 创建成功，可以返回 DTO 或者部分字段\n res = VoiceResponseDTO(**entityRes.__dict__)\n return Res(data=res, code=200, message=\"创建成功\")\n else:\n return Res(data=None, code=400, message=f\"音色 '{entity.name}' 已存在\")\n\n except ValueError as e:\n raise HTTPException(status_code=400, detail=str(e))\n\n\n# ====== 动态路由放在最后 ======\n\n@router.get(\"/{voice_id}\", response_model=Res[VoiceResponseDTO],\n summary=\"查询音色\",\n description=\"根据音色id查询音色信息\")\ndef get_voice(voice_id: int, voice_service: VoiceService = Depends(get_voice_service)):\n entity = voice_service.get_voice(voice_id)\n if entity:\n res = VoiceResponseDTO(**entity.__dict__)\n return Res(data=res, code=200, message=\"查询成功\")\n else:\n return Res(data=None, code=404, message=\"项目不存在\")\n\n\n# 修改，传入的参数是id\n@router.put(\"/{voice_id}\", response_model=Res[VoiceCreateDTO],\n summary=\"修改音色信息\",\n description=\"根据音色id修改音色信息,并且不能修改项目id\")\ndef update_voice(voice_id: int, dto: VoiceCreateDTO, voice_service: VoiceService = Depends(get_voice_service)):\n voice = voice_service.get_voice(voice_id)\n if voice is None:\n return Res(data=None, code=404, message=\"音色不存在\")\n res = voice_service.update_voice(voice_id, dto.dict())\n if res:\n return Res(data=dto, code=200, message=\"修改成功\")\n else:\n return Res(data=None, code=400, message=\"修改失败\")\n\n\n# 根据 id，删除\n@router.delete(\"/{voice_id}\", response_model=Res,\n summary=\"删除音色\",\n description=\"根据音色id删除音色信息\")\ndef delete_voice(voice_id: int, voice_service: VoiceService = Depends(get_voice_service)):\n success = voice_service.delete_voice(voice_id)\n if success:\n return Res(data=None, code=200, message=\"删除成功\")\n else:\n return Res(data=None, code=400, message=\"删除失败或音色不存在\")\n\n\n# tts_provider的查询和修改\n# @router.get(\"/tts/provider/{tts_provider_id}\", response_model=Res[TTSProviderResponseDTO])\n# def get_tts_provider(tts_provider_id: int, tts_provider_service: TTSProviderService = Depends(get_tts_provider_service)):\n# tts_provider = tts_provider_service.get_tts_provider(tts_provider_id)\n# if tts_provider:\n# res = TTSProviderResponseDTO(**tts_provider.__dict__)\n# return Res(data=res, code=200, message=\"查询成功\")\n# else:\n# return Res(data=None, code=404, message=\"tts服务提供商不存在\")\n#\n# # tts_provider的修改\n# @router.put(\"/tts/provider/{tts_provider_id}\", response_model=Res[TTSProviderResponseDTO])\n# def update_tts_provider(tts_provider_id: int, dto: TTSProviderResponseDTO, tts_provider_service: TTSProviderService = Depends(get_tts_provider_service)):\n# # 先判断是否存在\n# tts_provider = tts_provider_service.get_tts_provider(tts_provider_id)\n# if tts_provider is None:\n# return Res(data=None, code=404, message=\"tts服务提供商不存在\")\n# tts_provider = tts_provider_service.update_tts_provider(tts_provider_id, dto.dict(exclude_unset=True))\n# if tts_provider:\n# return Res(data=None, code=200, message=\"修改成功\")\n# else:\n# return Res(data=None, code=400, message=\"修改失败\")\n\n"
},
{
"path": "SonicVale/app/services/chapter_service.py",
"content": "import json\nimport logging\nimport os\nimport re\nimport shutil\nimport threading\nfrom collections import defaultdict\nfrom typing import List\n\nfrom sqlalchemy import Sequence\n\nfrom app.core.config import getConfigPath\nfrom app.core.text_correct_engine import TextCorrectorFinal\nfrom app.core.tts_engine import TTSEngine\nfrom app.db.database import SessionLocal\nfrom app.dto.line_dto import LineInitDTO\nfrom app.entity.chapter_entity import ChapterEntity\nfrom app.entity.line_entity import LineEntity\nfrom app.models.po import ChapterPO, RolePO, LinePO\n\nfrom app.repositories.chapter_repository import ChapterRepository\nfrom app.repositories.line_repository import LineRepository\n\nfrom app.core.prompts import get_context2lines_prompt, get_add_smart_role_and_voice\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.repositories.project_repository import ProjectRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.core.llm_engine import LLMEngine\nfrom app.repositories.voice_repository import VoiceRepository\n\n\nclass ChapterService:\n\n def __init__(self, repository: ChapterRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_chapter(self, entity: ChapterEntity):\n \"\"\"创建新章节\n - 检查同名章节是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n\n chapter = self.repository.get_by_name(entity.title, entity.project_id)\n if chapter:\n logging.info(\"同名章节已存在\")\n return None\n # 手动将entity转化为po\n po = ChapterPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = ChapterEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_chapter(self, chapter_id: int) -> ChapterEntity | None:\n \"\"\"根据 ID 查询章节\"\"\"\n po = self.repository.get_by_id(chapter_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = ChapterEntity(**data)\n return res\n\n def get_all_chapters(self,project_id: int) -> Sequence[ChapterEntity]:\n \"\"\"获取所有章节列表\"\"\"\n pos = self.repository.get_all(project_id)\n # pos -> entities\n\n entities = [\n ChapterEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_chapter(self, chapter_id: int, data:dict) -> bool:\n \"\"\"更新章节\n - 可以只更新部分字段\n - 检查同名冲突\n - 检查project_id不能改变\n \"\"\"\n title = data[\"title\"]\n project_id = data[\"project_id\"]\n if self.repository.get_by_name(title, project_id) and self.repository.get_by_name(title,project_id).id != chapter_id:\n return False\n po = self.repository.get_by_id(chapter_id)\n # 防止改变project_id\n if po.project_id != project_id:\n return False\n self.repository.update(chapter_id, data)\n return True\n\n def delete_chapter(self, chapter_id: int) -> bool:\n \"\"\"删除章节\n \"\"\"\n db = SessionLocal()\n try :\n chapter = self.repository.get_by_id(chapter_id)\n\n # 移除资源内容\n # 删除该路径所有内容\n project_repository = ProjectRepository(db)\n project = project_repository.get_by_id(chapter.project_id)\n chapter_path = os.path.join(project.project_root_path, str(chapter.project_id), str(chapter_id))\n if os.path.exists(chapter_path):\n shutil.rmtree(chapter_path) # 删除整个文件夹及其所有内容\n logging.info(\"已删除目录及内容: %s\", chapter_path)\n else:\n logging.info(\"目录不存在: %s\", chapter_path)\n # 先删除资源，再删除记录\n res = self.repository.delete(chapter_id)\n # 删除章节下所有台词\n line_repository = LineRepository(db)\n line_res = line_repository.delete_all_by_chapter_id(chapter_id)\n finally:\n db.close()\n return res\n\n # 先获取章节内容\n def split_text(self, chapter_id: int, max_length: int = 1500) -> List[str]:\n \"\"\"\n 将文本按标点/换行断句，并按最大长度分组，确保每段以标点结束。\n 支持中英文标点和换行符。\n \"\"\"\n content = self.get_chapter(chapter_id).text_content\n # 去掉空行\n content = \"\\n\".join([line for line in content.split(\"\\n\") if line.strip()])\n\n # 如果最后没有句号/问号/感叹号/点号，自动补一个句号\n if not re.search(r'[。！？.!?]$', content):\n content += \"。\"\n\n # 使用正则分割，支持中英文标点 + 逗号 + 换行\n # [] 里列出所有可能的结束符号\n sentences = re.findall(r'[^。！？.!?,，\\n]*[。！？.!?,，\\n]', content, re.MULTILINE | re.DOTALL)\n\n chunks = []\n buffer = \"\"\n\n for sentence in sentences:\n if len(buffer) + len(sentence) <= max_length:\n buffer += sentence\n else:\n if buffer:\n chunks.append(buffer.strip())\n buffer = sentence\n\n if buffer:\n chunks.append(buffer.strip())\n\n return chunks\n\n # 然后进行划分\n\n # 然后循环解析，并保存\n def fill_prompt(self,template: str, characters: list[str], emotions: list[str], strengths: list[str],\n novel_content: str) -> str:\n result = template\n result = result.replace(\"{possible_characters}\", \", \".join(characters))\n result = result.replace(\"{possible_emotions}\", \", \".join(emotions))\n result = result.replace(\"{possible_strengths}\", \", \".join(strengths))\n result = result.replace(\"{novel_content}\", novel_content)\n return result\n\n def para_content(self, prompt:str,chapter_id: int,content: str = None,role_names: List[str] = None,emotion_names: List[str] = None,strength_names: List[str] = None,is_precise_fill: int = 0):\n db = SessionLocal()\n try :\n # 获取content\n chapter = self.repository.get_by_id(chapter_id)\n # content = chapter.text_content\n # 获取角色列表\n # role_repository = RoleRepository(db)\n # roles = role_repository.get_all(chapter.project_id)\n # role_names = [role.name for role in roles]\n # 组装prompt\n # prompt = get_context2lines_prompt(role_names, content,emotion_names,strength_names)\n prompt = self.fill_prompt(prompt, role_names, emotion_names, strength_names, content)\n\n # 获取llm_provider\n\n project_repository = ProjectRepository(db)\n project = project_repository.get_by_id(chapter.project_id)\n llm_provider_id = project.llm_provider_id\n #\n llm_provider_repository = LLMProviderRepository(db)\n llm_provider = llm_provider_repository.get_by_id(llm_provider_id)\n llm = LLMEngine(llm_provider.api_key, llm_provider.api_base_url, project.llm_model, llm_provider.custom_params)\n try:\n llm.generate_text_test(\"请输出一份用户信息，严格使用 JSON 格式，不要包含任何额外文字。字段包括：name, age, city\")\n logging.info(\"LLM可用\")\n except Exception as e:\n logging.warning(\"LLM不可用\")\n return {\n \"success\": False,\n \"message\": f\"LLM 不可用: {str(e)}\"\n }\n logging.info(\"开始内容解析\")\n try:\n result = llm.generate_text(prompt)\n # 解析json，并且构造为List[LineInitDTO]\n # 解析 JSON 字符串为 Python 对象\n parsed_data = llm.save_load_json(result)\n if not parsed_data:\n return {\n \"success\": False,\n \"message\": \"JSON 解析失败或返回空对象\",\n }\n \n # 验证 parsed_data 是否为有效的字典列表\n if not isinstance(parsed_data, list):\n logging.error(\"LLM返回的数据不是列表，实际类型: %s\", type(parsed_data))\n return {\n \"success\": False,\n \"message\": f\"LLM返回的数据格式不正确，期望列表但收到: {type(parsed_data).__name__}\",\n }\n \n # 验证列表中的每个元素是否为字典\n for idx, item in enumerate(parsed_data):\n if not isinstance(item, dict):\n logging.error(\"列表第 %d 项不是字典，实际类型: %s, 内容: %s\", idx, type(item), str(item)[:100])\n return {\n \"success\": False,\n \"message\": f\"LLM返回的数据格式不正确，列表第 {idx} 项应为字典但收到: {type(item).__name__}\",\n }\n \n # 这里进行自动填充\n if is_precise_fill == 1:\n logging.info(\"开始自动填充\")\n corrector = TextCorrectorFinal()\n parsed_data = corrector.correct_ai_text(content, parsed_data)\n\n # parsed_data = json.loads(result)\n # 构造 List[LineInitDTO]\n line_dtos: List[LineInitDTO] = [LineInitDTO(**item) for item in parsed_data]\n return {\n \"success\": True,\n \"data\": line_dtos\n }\n\n except Exception as e:\n logging.exception(\"调用 LLM 出错: %s\", e)\n return {\n \"success\": False,\n \"message\": f\"调用 LLM 出错: {str(e)}\"\n }\n finally:\n db.close()\n\n\n # 导出指令\n # def get_prompt_content(self,project_id, chapter_id,prompt):\n # db = SessionLocal()\n # try:\n # # 获取content\n # chapter = self.repository.get_by_id(chapter_id)\n # content = chapter.text_content\n # # 获取角色列表\n # role_repository = RoleRepository(db)\n # roles = role_repository.get_all(chapter.project_id)\n # role_names = [role.name for role in roles]\n # # 组装prompt\n # # 获取project\n #\n # prompt = self.fill_prompt(prompt, role_names, emotion_names, strength_names, content)\n # prompt = get_context2lines_prompt(role_names, content)\n # return prompt\n # finally:\n # db.close()\n def add_smart_role_and_voice(self,project,content, role_names, voice_names):\n # 智能匹配提示词，要写死吗？\n db = SessionLocal()\n try:\n llm_provider_id = project.llm_provider_id\n llm_provider_repository = LLMProviderRepository(db)\n llm_provider = llm_provider_repository.get_by_id(llm_provider_id)\n llm = LLMEngine(llm_provider.api_key, llm_provider.api_base_url, project.llm_model, llm_provider.custom_params)\n prompt = get_add_smart_role_and_voice(content,role_names, voice_names)\n result = llm.generate_smart_text(prompt)\n parse_data = llm.save_load_json(result)\n # 获取项目所有音色\n voice_repository = VoiceRepository(db)\n voices = voice_repository.get_all(project.tts_provider_id)\n # map name- id\n voice_id_map = {voice.name: voice.id for voice in voices}\n\n\n # 对角色进行update\n role_repository = RoleRepository(db)\n res = []\n for item in parse_data:\n role = role_repository.get_by_name( item[\"role_name\"],project.id)\n if role:\n if item[\"voice_name\"]:\n logging.info(\"更新角色音色：%s %s\", item[\"role_name\"], item[\"voice_name\"])\n role_repository.update(role.id, {\"default_voice_id\": voice_id_map.get(item[\"voice_name\"])})\n res.append({\"role_name\": item[\"role_name\"], \"voice_name\": item[\"voice_name\"]})\n\n return True,res\n except Exception as e:\n logging.exception(\"LLM智能匹配出错: %s\", e)\n return False, []\n finally:\n db.close()\n\n\n\n\n\n"
},
{
"path": "SonicVale/app/services/emotion_service.py",
"content": "from sqlalchemy import Sequence\n\nfrom app.entity.emotion_entity import EmotionEntity\nfrom app.models.po import EmotionPO\nfrom app.repositories.emotion_repository import EmotionRepository\n\n\nclass EmotionService:\n\n def __init__(self, repository: EmotionRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_emotion(self, entity: EmotionEntity):\n \"\"\"创建新情绪枚举\n - 检查同名情绪枚举是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n\n emotion = self.repository.get_by_name(entity.name)\n if emotion:\n return None\n # 手动将entity转化为po\n po = EmotionPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = EmotionEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_emotion(self, emotion_id: int) -> EmotionEntity | None:\n \"\"\"根据 ID 查询情绪枚举\"\"\"\n po = self.repository.get_by_id(emotion_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = EmotionEntity(**data)\n return res\n\n def get_all_emotions(self) -> Sequence[EmotionEntity]:\n \"\"\"获取所有情绪枚举列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n EmotionEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_emotion(self, emotion_id: int, data:dict) -> bool:\n \"\"\"更新情绪枚举\n - 可以只更新部分字段\n \"\"\"\n name = data.get(\"name\")\n if self.repository.get_by_name(name):\n return False\n self.repository.update(emotion_id, data)\n return True\n\n def delete_emotion(self, emotion_id: int) -> bool:\n \"\"\"删除情绪枚举\n \"\"\"\n res = self.repository.delete(emotion_id)\n return res\n\n def get_emotion_by_name(self, name: str) -> EmotionEntity | None:\n \"\"\"根据名称查询情绪枚举\"\"\"\n po = self.repository.get_by_name(name)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = EmotionEntity(**data)\n return res\n"
},
{
"path": "SonicVale/app/services/line_service.py",
"content": "import contextlib\nimport hashlib\nimport logging\n\nimport shutil\nimport subprocess\nimport sys\nimport tempfile\nimport threading\nfrom collections import defaultdict\nfrom typing import List\n\nfrom openpyxl import Workbook\nfrom sqlalchemy import Sequence\n\n\nfrom app.core.audio_engin import AudioProcessor\nfrom app.core.config import getConfigPath, getFfmpegPath\nfrom app.core.subtitle import subtitle_engine\nfrom app.core.tts_engine import TTSEngine\nfrom app.dto.line_dto import LineCreateDTO, LineOrderDTO, LineAudioProcessDTO\nfrom app.entity.line_entity import LineEntity\nfrom app.models.po import LinePO, RolePO\nfrom app.repositories.line_repository import LineRepository\nfrom app.repositories.role_repository import RoleRepository\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\nfrom app.core.llm_engine import LLMEngine\n\nimport os\n\nimport numpy as np\nimport soundfile as sf\n\ndef _lock_key(path: str) -> str:\n return hashlib.md5(path.encode(\"utf-8\")).hexdigest()\n_file_locks = defaultdict(threading.Lock)\nclass LineService:\n\n def __init__(self, repository: LineRepository,role_repository: RoleRepository,tts_provider_repository: TTSProviderRepository, llm_provider_repository: LLMProviderRepository = None):\n \"\"\"注入 repository\"\"\"\n\n self.tts_provider_repository = tts_provider_repository\n self.llm_provider_repository = llm_provider_repository\n self.role_repository = role_repository\n self.repository = repository\n\n def create_line(self, entity: LineEntity):\n \"\"\"创建新台词\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n # 手动将entity转化为po\n po = LinePO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = LineEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_line(self, line_id: int) -> LineEntity | None:\n \"\"\"根据 ID 查询台词\"\"\"\n po = self.repository.get_by_id(line_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = LineEntity(**data)\n return res\n\n def get_all_lines(self,chapter_id: int) -> Sequence[LineEntity]:\n \"\"\"获取所有台词列表\"\"\"\n pos = self.repository.get_all(chapter_id)\n # pos -> entities\n\n entities = [\n LineEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def delete_line(self, line_id: int) -> bool:\n \"\"\"删除台词\n \"\"\"\n # 还要把audio_path删除\n po = self.repository.get_by_id(line_id)\n if po and po.audio_path:\n with contextlib.suppress(FileNotFoundError):\n os.remove(po.audio_path)\n res = self.repository.delete(line_id)\n return res\n # 删除章节下所有台词\n def delete_all_lines(self, chapter_id: int) -> bool:\n \"\"\"删除章节下所有台词\n \"\"\"\n # 要移除所有的音频资源\n for line in self.get_all_lines(chapter_id):\n if line and line.audio_path:\n with contextlib.suppress(FileNotFoundError):\n os.remove(line.audio_path)\n return self.repository.delete_all_by_chapter_id(chapter_id)\n\n # 单个台词新增\n def add_new_line(self, line: LineCreateDTO,project_id,chapter_id,index,emotions_dict, strengths_dict,audio_path):\n # 先判断角色是否存在\n role = self.role_repository.get_by_name(line.role_name,project_id)\n if role is None:\n # 新增角色\n role = self.role_repository.create(RolePO(name=line.role_name, project_id=project_id))\n # 获取情绪id\n emotion_id = emotions_dict.get(line.emotion_name)\n # 获取强度id\n strength_id = strengths_dict.get(line.strength_name)\n res = self.repository.create(LinePO(text_content=line.text_content, role_id=role.id,\n chapter_id=chapter_id,line_order = index+1,emotion_id=emotion_id,strength_id=strength_id))\n\n # 新增台词,这里搞个audio_path\n\n # audio_path = os.path.join(getConfigPath(), str(project_id), str(chapter_id), \"audio\")\n # os.makedirs(audio_path, exist_ok=True)\n res_path = os.path.join(audio_path, \"id_\"+str(res.id) + \".wav\")\n self.repository.update(res.id, {\"audio_path\": res_path})\n\n\n def update_init_lines(self, lines: list, project_id: object, chapter_id: object,emotions_dict, strengths_dict,audio_path) -> None:\n for index, line in enumerate(lines):\n self.add_new_line(line,project_id,chapter_id,index,emotions_dict, strengths_dict,audio_path)\n\n # 获取章节下所有台词\n\n # 更新line\n def update_line(self, line_id: int, data: dict) -> bool:\n po = self.repository.get_by_id(line_id)\n if po is None:\n return False\n res = self.repository.update(line_id, data)\n if res is None:\n return False\n return True\n # 生成音频（服务器和本地两种方式）\n\n def generate_audio(self, reference_path: str,tts_provider_id,content,emo_text:str,emo_vector:list[float],save_path= None):\n #\n tts_provider = self.tts_provider_repository.get_by_id(tts_provider_id)\n if tts_provider is None:\n raise Exception(f\"TTS服务提供商不存在（ID: {tts_provider_id}）\")\n \n if not tts_provider.api_base_url:\n raise Exception(\"TTS服务地址未配置，请先在配置中心设置TTS服务\")\n \n tts_engine = TTSEngine(tts_provider.api_base_url)\n \n # 检查参考音频路径是否有效\n if not reference_path:\n raise Exception(\"参考音频路径未设置，请检查角色音色配置\")\n\n key = _lock_key(reference_path)\n lock = _file_locks[key]\n\n with lock:\n try:\n audio_exists = tts_engine.check_audio_exists(reference_path)\n except Exception as e:\n raise Exception(f\"检查参考音频失败: {str(e)}\")\n \n if not audio_exists:\n # 检查本地文件是否存在\n if not os.path.isfile(reference_path):\n raise Exception(f\"参考音频文件不存在: {reference_path}\")\n \n upload_result = tts_engine.upload_audio(reference_path, reference_path)\n if upload_result.get('code') and upload_result.get('code') != 200:\n raise Exception(f\"上传参考音频失败: {upload_result.get('msg', '未知错误')}\")\n \n # 合成音频\n return tts_engine.synthesize(content, reference_path, emo_text, emo_vector, save_path)\n\n # 将角色role_id下所有台词的role_id都置位空\n def clear_role_id(self, role_id: int):\n # 先获取role_id下所有台词实体\n pos = self.repository.get_lines_by_role_id(role_id)\n for po in pos:\n self.repository.update(po.id, {\"role_id\": None})\n\n def batch_update_line_order(self,line_orders:List[LineOrderDTO]):\n for line_order in line_orders:\n self.update_line(line_order.id,{\"line_order\":line_order.line_order})\n return True\n\n def update_audio_path(self, id, dto) -> bool:\n try:\n po = self.get_line(id)\n old_path = po.audio_path\n new_path = dto.audio_path\n\n if not old_path:\n return False # 原始路径为空\n\n if not os.path.exists(old_path):\n return False # 原始文件不存在\n\n if os.path.exists(new_path):\n return False # 目标文件已存在，避免覆盖\n\n # 确保目标目录存在\n os.makedirs(os.path.dirname(new_path), exist_ok=True)\n\n # 重命名文件\n shutil.move(old_path, new_path)\n\n # 更新数据库\n self.update_line(id, {\"audio_path\": new_path})\n return True\n\n except Exception as e:\n logging.exception(\"[update_audio_path] 失败: %s\", e)\n return False\n\n def process_audio_ffmpeg(\n self,\n audio_path: str,\n speed: float = 1.0,\n volume: float = 1.0,\n start_ms: int | None = None,\n end_ms: int | None = None,\n out_path: str | None = None,\n keep_format: bool = True, # 是否保持原文件采样率/声道\n default_sr: int = 44100,\n default_ch: int = 2\n ):\n \"\"\"\n 使用 ffmpeg 对音频进行变速 (0.5~2.0)、音量调整、可选裁剪。\n 输出 WAV PCM16。\n 如果 keep_format=True，则保持输入文件的 sr/ch 不变。\n \"\"\"\n ffmpeg_path = getFfmpegPath()\n if not os.path.exists(audio_path):\n raise FileNotFoundError(audio_path)\n\n # 获取原始参数\n info = sf.info(audio_path)\n target_sr = info.samplerate if keep_format else default_sr\n target_ch = info.channels if keep_format else default_ch\n\n # 参数规整\n speed = float(np.clip(speed or 1.0, 0.5, 2.0))\n volume = 1.0 if volume is None else max(0.0, float(volume))\n\n # 输出路径\n target_path = out_path or audio_path\n os.makedirs(os.path.dirname(target_path) or \".\", exist_ok=True)\n with tempfile.NamedTemporaryFile(delete=False, suffix=\".wav\",\n dir=os.path.dirname(target_path) or \".\") as tmp:\n tmp_path = tmp.name\n\n # 构建 ffmpeg 命令\n filter_chain = [f\"atempo={speed}\"]\n if abs(volume - 1.0) > 1e-6:\n filter_chain.append(f\"volume={volume}\")\n\n cmd = [ffmpeg_path, \"-y\"]\n if start_ms is not None:\n cmd.extend([\"-ss\", str(start_ms / 1000)])\n cmd.extend([\"-i\", audio_path])\n if end_ms is not None:\n cmd.extend([\"-to\", str(end_ms / 1000)])\n cmd.extend([\n \"-af\", \",\".join(filter_chain),\n \"-ar\", str(target_sr),\n \"-ac\", str(target_ch),\n \"-c:a\", \"pcm_s16le\",\n tmp_path\n ])\n\n subprocess.run(cmd, check=True,\n creationflags=subprocess.CREATE_NO_WINDOW if sys.platform == \"win32\" else 0)\n\n # 软限幅：避免 clipping\n data, sr = sf.read(tmp_path, dtype=\"float32\", always_2d=True)\n peak = float(np.max(np.abs(data)))\n if peak > 1.0:\n data = data / peak\n sf.write(tmp_path, data, sr, format=\"WAV\", subtype=\"PCM_16\")\n\n os.replace(tmp_path, target_path)\n return target_path\n\n\n # 删除区间进行拼接\n def process_audio_ffmpeg_cut(\n self,\n audio_path: str,\n speed: float = 1.0,\n volume: float = 1.0,\n start_ms: int | None = None,\n end_ms: int | None = None,\n silence_sec: float = 0.0, # 末尾静音时长，单位秒\n out_path: str | None = None,\n keep_format: bool = True, # 是否保持原文件采样率/声道\n default_sr: int = 44100,\n default_ch: int = 2\n ):\n \"\"\"\n 使用 ffmpeg 对音频进行变速 (0.5~2.0)、音量调整。\n 删除 [start_ms, end_ms] 区间，并拼接前后音频。\n 输出 WAV PCM16。\n 可在末尾附加 silence_sec 秒静音。\n \"\"\"\n ffmpeg_path = getFfmpegPath()\n if not os.path.exists(audio_path):\n raise FileNotFoundError(audio_path)\n\n # 获取原始参数\n info = sf.info(audio_path)\n target_sr = info.samplerate if keep_format else default_sr\n target_ch = info.channels if keep_format else default_ch\n\n # 参数规整\n speed = float(np.clip(speed or 1.0, 0.5, 2.0))\n volume = 1.0 if volume is None else max(0.0, float(volume))\n\n # 输出路径\n target_path = out_path or audio_path\n os.makedirs(os.path.dirname(target_path) or \".\", exist_ok=True)\n with tempfile.NamedTemporaryFile(delete=False, suffix=\".wav\",\n dir=os.path.dirname(target_path) or \".\") as tmp:\n tmp_path = tmp.name\n\n # 构建 ffmpeg 命令\n if start_ms is None or end_ms is None or end_ms <= start_ms:\n # 无剪切\n if silence_sec > 0:\n # 添加静音\n cmd = [\n ffmpeg_path, \"-y\",\n \"-i\", audio_path,\n \"-f\", \"lavfi\", \"-t\", str(silence_sec),\n \"-i\", f\"anullsrc=channel_layout={'stereo' if target_ch == 2 else 'mono'}:sample_rate={target_sr}\",\n \"-filter_complex\",\n f\"[0:a]atempo={speed},volume={volume}[main];\"\n f\"[main][1:a]concat=n=2:v=0:a=1[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(target_sr),\n \"-ac\", str(target_ch),\n \"-c:a\", \"pcm_s16le\",\n tmp_path\n ]\n elif silence_sec < 0:\n # 裁掉末尾 abs(silence_sec)\n cut_dur = info.duration + silence_sec\n if cut_dur <= 0:\n cut_dur = 0 # 整段裁掉\n\n cmd = [\n ffmpeg_path, \"-y\",\n \"-i\", audio_path,\n \"-filter_complex\",\n f\"[0:a]atempo={speed},volume={volume},atrim=0:{cut_dur}[out]\",\n \"-map\", \"[out]\",\n \"-ar\", str(target_sr),\n \"-ac\", str(target_ch),\n \"-c:a\", \"pcm_s16le\",\n tmp_path\n ]\n else:\n # 不处理末尾\n cmd = [\n ffmpeg_path, \"-y\", \"-i\", audio_path,\n \"-af\", f\"atempo={speed},volume={volume}\",\n \"-ar\", str(target_sr),\n \"-ac\", str(target_ch),\n \"-c:a\", \"pcm_s16le\",\n tmp_path\n ]\n\n\n else:\n\n # 剪切\n\n start_sec = start_ms / 1000\n\n end_sec = end_ms / 1000\n\n if silence_sec > 0:\n\n # 拼接 + 添加静音\n\n cmd = [\n\n ffmpeg_path, \"-y\",\n\n \"-i\", audio_path,\n\n \"-f\", \"lavfi\", \"-t\", str(silence_sec),\n\n \"-i\", f\"anullsrc=channel_layout={'stereo' if target_ch == 2 else 'mono'}:sample_rate={target_sr}\",\n\n \"-filter_complex\",\n\n f\"[0:a]atrim=0:{start_sec},asetpts=PTS-STARTPTS[first];\"\n\n f\"[0:a]atrim={end_sec},asetpts=PTS-STARTPTS[second];\"\n\n f\"[first][second]concat=n=2:v=0:a=1,atempo={speed},volume={volume}[main];\"\n\n f\"[main][1:a]concat=n=2:v=0:a=1[out]\",\n\n \"-map\", \"[out]\",\n\n \"-ar\", str(target_sr),\n\n \"-ac\", str(target_ch),\n\n \"-c:a\", \"pcm_s16le\",\n\n tmp_path\n\n ]\n\n elif silence_sec < 0:\n\n # 拼接后再裁掉末尾\n\n cut_dur = info.duration + silence_sec\n if cut_dur <= 0:\n cut_dur = 0 # 整段裁掉\n\n cmd = [\n\n ffmpeg_path, \"-y\", \"-i\", audio_path,\n\n \"-filter_complex\",\n\n f\"[0:a]atrim=0:{start_sec},asetpts=PTS-STARTPTS[first];\"\n\n f\"[0:a]atrim={end_sec},asetpts=PTS-STARTPTS[second];\"\n\n f\"[first][second]concat=n=2:v=0:a=1,atempo={speed},volume={volume},atrim=0:{cut_dur}[out]\",\n\n \"-map\", \"[out]\",\n\n \"-ar\", str(target_sr),\n\n \"-ac\", str(target_ch),\n\n \"-c:a\", \"pcm_s16le\",\n\n tmp_path\n\n ]\n\n else:\n\n # 拼接但不处理末尾\n\n cmd = [\n\n ffmpeg_path, \"-y\", \"-i\", audio_path,\n\n \"-filter_complex\",\n\n f\"[0:a]atrim=0:{start_sec},asetpts=PTS-STARTPTS[first];\"\n\n f\"[0:a]atrim={end_sec},asetpts=PTS-STARTPTS[second];\"\n\n f\"[first][second]concat=n=2:v=0:a=1,atempo={speed},volume={volume}[out]\",\n\n \"-map\", \"[out]\",\n\n \"-ar\", str(target_sr),\n\n \"-ac\", str(target_ch),\n\n \"-c:a\", \"pcm_s16le\",\n\n tmp_path\n\n ]\n\n # 执行 ffmpeg\n subprocess.run(\n cmd, check=True,\n creationflags=subprocess.CREATE_NO_WINDOW if sys.platform == \"win32\" else 0\n )\n\n # 软限幅：避免 clipping\n data, sr = sf.read(tmp_path, dtype=\"float32\", always_2d=True)\n peak = float(np.max(np.abs(data)))\n if peak > 1.0:\n data = data / peak\n sf.write(tmp_path, data, sr, format=\"WAV\", subtype=\"PCM_16\")\n\n os.replace(tmp_path, target_path)\n return target_path\n\n def process_audio(self, line_id, dto:LineAudioProcessDTO):\n line = self.get_line(line_id)\n if line:\n # 读取音频文件\n # audio_file =self.process_audio_ffmpeg(line.audio_path, dto.speed, dto.volume,dto.start_ms,dto.end_ms)\n # 删除拼接\n # audio_file = self.process_audio_ffmpeg_cut(line.audio_path, dto.speed, dto.volume, dto.start_ms, dto.end_ms, dto.tail_silence_sec,dto.current_ms)\n processor = AudioProcessor(line.audio_path)\n start_ms = dto.start_ms\n end_ms = dto.end_ms\n speed = dto.speed\n volume = dto.volume\n current_ms = dto.current_ms\n silence_sec = dto.silence_sec\n # ---------- (1) 优先裁剪 ----------\n if start_ms is not None and end_ms is not None and end_ms > start_ms:\n logging.info(\"裁剪\")\n processor.cut(start_ms, end_ms)\n\n # ---------- (2) 插入静音 ----------\n elif current_ms is not None and silence_sec is not None and silence_sec != 0:\n logging.info(\"插入静音\")\n processor.insert_silence(current_ms, silence_sec)\n\n # ---------- (3) 末尾静音/裁剪 ----------\n elif current_ms is None and silence_sec is not None and silence_sec != 0:\n logging.info(\"末尾静音/裁剪\")\n processor.append_silence(silence_sec)\n\n # ---------- (4) 音量 + 变速 ----------\n if speed != 1.0:\n processor.change_speed(speed)\n if volume != 1.0:\n processor.change_volume(volume)\n logging.info(\"音频处理完成\")\n return True\n\n else:\n return False\n\n # 导出音频,合并音频，并且导出字幕\n def concat_wav_files(self,paths, out_path, verify=True, block_frames=262144):\n \"\"\"\n 按顺序把若干 WAV 合并到 out_path。\n 假设：采样率与声道一致（如需更稳，可保留 verify=True 做轻校验）。\n \"\"\"\n assert paths and len(paths) >= 1, \"至少提供一个文件路径\"\n os.makedirs(os.path.dirname(out_path) or \".\", exist_ok=True)\n\n # 以首文件格式为准\n info0 = sf.info(paths[0])\n sr, ch, subtype = info0.samplerate, info0.channels, info0.subtype or \"PCM_16\"\n\n # 可选校验\n if verify:\n for p in paths[1:]:\n info = sf.info(p)\n if info.samplerate != sr or info.channels != ch:\n raise ValueError(\n f\"格式不一致：{p} (sr={info.samplerate}, ch={info.channels}) vs 首文件 (sr={sr}, ch={ch})\")\n\n # 流式写入\n with sf.SoundFile(out_path, mode='w', samplerate=sr, channels=ch, format='WAV', subtype=subtype) as fout:\n for p in paths:\n with sf.SoundFile(p, mode='r') as fin:\n if verify and (fin.samplerate != sr or fin.channels != ch):\n raise ValueError(f\"参数不一致：{p}\")\n while True:\n block = fin.read(block_frames, dtype='float32', always_2d=True)\n if len(block) == 0:\n break\n fout.write(block.astype(np.float32, copy=False))\n return out_path\n\n\n\n def export_lines_to_excel(self,lines, file_path=\"all_lines.xlsx\"):\n # 1) 取出所有数据\n # lines = self.repository.get_all(chapter_id)\n\n # 2) 创建 Excel 工作簿\n wb = Workbook()\n ws = wb.active\n ws.title = \"Lines\"\n\n # 3) 写表头（根据你的数据字段调整）\n headers = [\"序号\",\"角色\", \"台词\"]\n ws.append(headers)\n\n # 4) 写内容\n for line in lines:\n role = self.role_repository.get_by_id(line.role_id)\n role_name = role.name if role else \"未知角色\"\n ws.append([\n line.line_order,\n role_name,\n line.text_content\n ])\n # 5) 保存到文件\n wb.save(file_path)\n return file_path\n\n def export_audio(self, chapter_id, single=False):\n \"\"\"导出音频与字幕\n \n Returns:\n dict: 包含导出结果的详细信息\n - success: bool, 是否成功\n - message: str, 错误信息（如果失败）\n - audio_path: str, 合并后的音频路径\n - subtitle_path: str, 字幕路径\n - missing_files: list, 缺失的音频文件列表\n \"\"\"\n try:\n # 拿到所有的台词\n lines = self.repository.get_all(chapter_id)\n \n if not lines:\n return {\"success\": False, \"message\": \"该章节没有台词\"}\n \n # 过滤掉空路径和不存在的文件\n valid_lines = []\n missing_files = []\n for line in lines:\n if not line.audio_path:\n missing_files.append(f\"台词#{line.id}(序号{line.line_order}): 无音频路径\")\n elif not os.path.exists(line.audio_path):\n missing_files.append(f\"台词#{line.id}(序号{line.line_order}): 文件不存在 - {line.audio_path}\")\n else:\n valid_lines.append(line)\n \n if not valid_lines:\n return {\n \"success\": False, \n \"message\": \"没有有效的音频文件可导出\",\n \"missing_files\": missing_files\n }\n \n paths = [line.audio_path for line in valid_lines]\n \n # 把paths[0]的path去掉后面的文件名，得到文件夹路径\n output_dir_path = os.path.join(os.path.dirname(paths[0]), \"result\")\n # 不存在就创建\n os.makedirs(output_dir_path, exist_ok=True)\n \n # 放到result目录下\n output_path = os.path.join(output_dir_path, \"result.wav\")\n \n # 合并音频文件\n try:\n self.concat_wav_files(paths, output_path)\n except ValueError as e:\n return {\n \"success\": False,\n \"message\": f\"音频合并失败: {str(e)}\",\n \"missing_files\": missing_files\n }\n except Exception as e:\n logging.exception(\"[export_audio] concat_wav_files 失败\")\n return {\n \"success\": False,\n \"message\": f\"音频合并异常: {str(e)}\",\n \"missing_files\": missing_files\n }\n \n # 生成字幕\n output_subtitle_path = os.path.join(output_dir_path, \"result.srt\")\n try:\n subtitle_engine.generate_subtitle(output_path, output_subtitle_path)\n except Exception as e:\n logging.exception(\"[export_audio] 生成整体字幕失败\")\n # 字幕生成失败不影响音频导出，继续执行\n \n # 生成单条字幕（如果需要）\n if single:\n subtitle_dir_path = os.path.join(os.path.dirname(paths[0]), \"subtitles\")\n # 先清空这个文件夹\n shutil.rmtree(subtitle_dir_path, ignore_errors=True)\n os.makedirs(subtitle_dir_path, exist_ok=True)\n \n for line in valid_lines:\n try:\n path = line.audio_path\n base_name = os.path.splitext(os.path.basename(path))[0]\n subtitle_path = os.path.join(subtitle_dir_path, base_name + \".srt\")\n subtitle_engine.generate_subtitle(path, subtitle_path)\n # 将subtitle_path写进line.subtitle_path\n self.repository.update(line.id, {\"subtitle_path\": subtitle_path})\n except Exception as e:\n logging.warning(f\"[export_audio] 生成单条字幕失败 line#{line.id}: {e}\")\n # 单条字幕失败不影响整体导出\n \n # 导出所有数据到Excel\n try:\n self.export_lines_to_excel(lines, os.path.join(output_dir_path, \"all_lines.xlsx\"))\n except Exception as e:\n logging.warning(f\"[export_audio] 导出Excel失败: {e}\")\n # Excel导出失败不影响整体导出\n \n result = {\n \"success\": True,\n \"audio_path\": output_path,\n \"subtitle_path\": output_subtitle_path,\n \"exported_count\": len(valid_lines),\n \"total_count\": len(lines)\n }\n \n if missing_files:\n result[\"missing_files\"] = missing_files\n result[\"message\"] = f\"导出成功，但有{len(missing_files)}条台词缺少音频\"\n \n return result\n \n except Exception as e:\n logging.exception(\"[export_audio] 未预期的错误\")\n return {\"success\": False, \"message\": f\"导出失败: {str(e)}\"}\n\n\n\n\n def generate_subtitle(self, line_id, dto):\n # 获取台词\n line = self.get_line(line_id)\n if line:\n # 将音频文件路径的后缀改为.srt\n dto.subtitle_path = os.path.splitext(dto.subtitle_path)[0] + \".srt\"\n subtitle_engine.generate_subtitle(line.audio_path,dto.subtitle_path)\n return dto.subtitle_path\n else:\n return None\n# 字幕矫正 - 拼音匹配\n def correct_subtitle_pinyin(self, text, output_subtitle_path):\n \"\"\"\n 使用拼音匹配算法矫正字幕\n \n text: 原始正确文本\n output_subtitle_path: 字幕文件路径\n \"\"\"\n subtitle_engine.correct_srt_file(text, output_subtitle_path)\n\n# 字幕矫正 - LLM\n def correct_subtitle_llm(self, text, output_subtitle_path, llm_provider_id: int, llm_model: str, batch_size: int = 20):\n \"\"\"\n 使用LLM矫正字幕\n \n text: 原始正确文本\n output_subtitle_path: 字幕文件路径\n llm_provider_id: LLM提供商ID\n llm_model: LLM模型名称\n batch_size: 分批处理时每批的条数\n \"\"\"\n if not self.llm_provider_repository:\n raise Exception(\"LLM Provider Repository 未配置\")\n \n llm_provider = self.llm_provider_repository.get_by_id(llm_provider_id)\n if llm_provider is None:\n raise Exception(f\"LLM服务提供商不存在（ID: {llm_provider_id}）\")\n \n llm_engine = LLMEngine(\n api_key=llm_provider.api_key,\n base_url=llm_provider.api_base_url,\n model_name=llm_model,\n custom_params=llm_provider.custom_params or \"{}\"\n )\n \n subtitle_engine.correct_srt_file_with_llm(\n text, \n output_subtitle_path,\n llm_engine=llm_engine,\n batch_size=batch_size\n )\n\n# 生成字幕\n# def generate_subtitle(self, res_path):\n# subtitle_engine.generate_subtitle(res_path,res_path+\".srt\")\n"
},
{
"path": "SonicVale/app/services/llm_provider_service.py",
"content": "import json\nimport logging\n\nfrom aiohttp.abc import HTTPException\nfrom sqlalchemy import Sequence\n\nfrom app.core.llm_engine import LLMEngine\nfrom app.entity.llm_provider_entity import LLMProviderEntity\nfrom app.models.po import LLMProviderPO\n\nfrom app.repositories.llm_provider_repository import LLMProviderRepository\n\n\nclass LLMProviderService:\n\n def __init__(self, repository: LLMProviderRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_llm_provider(self, entity: LLMProviderEntity):\n \"\"\"创建新LLM供应商\n - 检查同名LLM供应商是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n llm_provider = self.repository.get_by_name(entity.name)\n if llm_provider:\n return None\n # 手动将entity转化为po\n po = LLMProviderPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = LLMProviderEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_llm_provider(self, llm_provider_id: int) -> LLMProviderEntity | None:\n \"\"\"根据 ID 查询LLM供应商\"\"\"\n po = self.repository.get_by_id(llm_provider_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = LLMProviderEntity(**data)\n return res\n\n def get_all_llm_providers(self) -> Sequence[LLMProviderEntity]:\n \"\"\"获取所有LLM供应商列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n LLMProviderEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_llm_provider(self, llm_provider_id: int, data:dict) -> bool:\n \"\"\"更新LLM供应商\n - 可以只更新部分字段\n - 检查同名冲突\n \"\"\"\n name = data[\"name\"]\n if self.repository.get_by_name(name) and self.repository.get_by_name(name).id != llm_provider_id:\n return False\n self.repository.update(llm_provider_id, data)\n return True\n\n def delete_llm_provider(self, llm_provider_id: int) -> bool:\n \"\"\"删除LLM供应商\n - 可以添加业务校验，例如LLM供应商下有章节是否允许删除\n - 后续需要级联删除所有章节内容\n \"\"\"\n res = self.repository.delete(llm_provider_id)\n return res\n\n def test_llm_provider(self, entity: LLMProviderEntity):\n \"\"\"测试LLM供应商\"\"\"\n # 按逗号划分模型名称\n if entity.api_base_url is None or entity.api_key is None or entity.model_list is None:\n return False\n model_lists = entity.model_list.split(\",\")\n custom_params = entity.custom_params\n llm = LLMEngine(entity.api_key, entity.api_base_url, model_lists[0],custom_params)\n try:\n res = llm.generate_text_test(\"请输出一份用户信息，严格使用 JSON 格式，不要包含任何额外文字。字段包括：name, age, city\")\n except Exception as e:\n return False,str(e)\n logging.info(\"测试结果为：%s\", res)\n if res is None:\n return False,\"LLM 未返回任何内容\"\n\n # 7. 校验返回是否为合法 JSON\n try:\n # res = res.replace(\"```json\",'')\n # res = res.replace(\"```\",'')\n json.loads(res)\n except json.JSONDecodeError:\n return False, \"LLM 返回的内容不是合法 JSON，请检查模型 / 提示词\"\n return True,\"测试成功\"\n\n\n\n\n"
},
{
"path": "SonicVale/app/services/multi_emotion_voice_service.py",
"content": "from sqlalchemy import Sequence\n\nfrom app.entity.multi_emotion_voice_entity import MultiEmotionVoiceEntity\nfrom app.models.po import MultiEmotionVoicePO\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\n\n\nclass MultiEmotionVoiceService:\n\n def __init__(self, repository: MultiEmotionVoiceRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_multi_emotion_voice(self, entity: MultiEmotionVoiceEntity):\n \"\"\"创建新多情绪音色变体\n - 检查同名多情绪音色变体是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n if entity.voice_id is None or entity.emotion_id is None or entity.strength_id is None:\n return None\n multi_emotion_voice = self.repository.get_by_voice_id_emotion_id_strength_id(entity.voice_id, entity.emotion_id, entity.strength_id)\n if multi_emotion_voice:\n return None\n po = MultiEmotionVoicePO(**entity.__dict__)\n res = self.repository.create(po)\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = MultiEmotionVoiceEntity(**data)\n\n # 将po转化为entity\n return entity\n # 根据voice_id,emotion_id,strength_id查询多情绪音色变体\n def get_multi_emotion_voice_by_voice_id_emotion_id_strength_id(self, voice_id: int, emotion_id: int, strength_id: int) -> MultiEmotionVoiceEntity | None:\n \"\"\"根据voice_id,emotion_id,strength_id查询多情绪音色变体\"\"\"\n po = self.repository.get_by_voice_id_emotion_id_strength_id(voice_id, emotion_id, strength_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = MultiEmotionVoiceEntity(**data)\n return res\n\n def get_multi_emotion_voice_by_voice_id(self, voice_id: int) -> list[MultiEmotionVoiceEntity] | None:\n \"\"\"根据 voice_id 查询所有的多情绪音色变体\"\"\"\n pos = self.repository.get_by_voice_id(voice_id)\n entities = [\n MultiEmotionVoiceEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def get_multi_emotion_voice_by_id(self, multi_emotion_voice_id: int) -> MultiEmotionVoiceEntity | None:\n \"\"\"根据 ID 获取多情绪音色变体\"\"\"\n po = self.repository.get_by_id(multi_emotion_voice_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = MultiEmotionVoiceEntity(**data)\n return res\n def get_all_multi_emotion_voices(self) -> list[MultiEmotionVoiceEntity]:\n \"\"\"获取所有多情绪音色变体列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n MultiEmotionVoiceEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_multi_emotion_voice(self, multi_emotion_voice_id: int, data:dict) -> bool:\n \"\"\"更新多情绪音色变体\n - 可以只更新部分字段\n - 检查同名冲突\n - 检查project_id不能改变\n \"\"\"\n self.repository.update(multi_emotion_voice_id, data)\n return True\n\n def delete_multi_emotion_voice(self, multi_emotion_voice_id: int) -> bool:\n \"\"\"删除多情绪音色变体\n \"\"\"\n res = self.repository.delete(multi_emotion_voice_id)\n return res\n\n# 删除voice下所有多情绪音色变体\n def delete_multi_emotion_voice_by_voice_id(self, voice_id: int) -> bool:\n \"\"\"删除voice下所有多情绪音色变体\n \"\"\"\n res = self.repository.delete_multi_emotion_voice_by_voice_id(voice_id)\n return res\n"
},
{
"path": "SonicVale/app/services/project_service.py",
"content": "import os\nimport re\nimport logging\n\nfrom sqlalchemy import Sequence\n\nfrom app.entity.project_entity import ProjectEntity\nfrom app.models.po import ProjectPO\n\nfrom app.repositories.project_repository import ProjectRepository\n\n\nclass ProjectService:\n\n def __init__(self, repository: ProjectRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_project(self, entity: ProjectEntity):\n \"\"\"创建新项目\n - 检查同名项目是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n project = self.repository.get_by_name(entity.name)\n if project:\n return None, \"项目已存在\"\n # 判断项目根路径是否存在\n if not os.path.exists(entity.project_root_path):\n logging.info(\"项目根路径不存在\")\n return None, \"项目根路径不存在\"\n # 手动将entity转化为po\n po = ProjectPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = ProjectEntity(**data)\n\n # 将po转化为entity\n return entity, \"创建成功\"\n\n\n def get_project(self, project_id: int) -> ProjectEntity | None:\n \"\"\"根据 ID 查询项目\"\"\"\n po = self.repository.get_by_id(project_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = ProjectEntity(**data)\n return res\n\n def get_all_projects(self) -> Sequence[ProjectEntity]:\n \"\"\"获取所有项目列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n ProjectEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_project(self, project_id: int, data:dict) -> bool:\n \"\"\"更新项目\n - 可以只更新部分字段\n - 检查同名冲突\n \"\"\"\n name = data[\"name\"]\n if self.repository.get_by_name(name) and self.repository.get_by_name(name).id != project_id:\n return False\n self.repository.update(project_id, data)\n return True\n\n def delete_project(self, project_id: int) -> bool:\n \"\"\"删除项目\n - 可以添加业务校验，例如项目下有章节是否允许删除\n - 后续需要级联删除所有章节内容\n \"\"\"\n res = self.repository.delete(project_id)\n return res\n\n\n def search_projects(self, keyword: str) -> Sequence[ProjectEntity]:\n \"\"\"模糊搜索项目\"\"\"\n\n # 解析content，按照章节\n def parse_content(self, content):\n \"\"\"解析内容，按照章节\"\"\"\n # 正则匹配常见章节格式（支持中英文数字）\n chapter_pattern = re.compile(\n r'(第[\\d一二三四五六七八九十百千]+[章回节部卷].*?)(?=\\n|$)'\n )\n # 找到所有章节标题位置\n matches = list(chapter_pattern.finditer(content))\n chapters = []\n # 如果没找到章节，直接返回整个文本\n if not matches:\n return chapters\n\n for i, match in enumerate(matches):\n start = match.end()\n end = matches[i + 1].start() if i + 1 < len(matches) else len(content)\n\n chapter_name = match.group(1).strip()\n chapter_content = content[start:end].strip()\n chapters.append({\n \"chapter_name\": chapter_name,\n \"content\": chapter_content\n })\n # 排序\n # chapters.sort(key=lambda x: x[\"chapter_name\"])\n # 不需要排序了，因为是顺序解析得到的\n return chapters\n"
},
{
"path": "SonicVale/app/services/prompt_service.py",
"content": "from numba.scripts.generate_lower_listing import description\nfrom sqlalchemy import Sequence\n\nfrom app.core.enums import TaskEnum\nfrom app.core.llm_engine import LLMEngine\nfrom app.core.prompts import get_prompt_str\nfrom app.entity.prompt_entity import PromptEntity\nfrom app.models.po import PromptPO\n\nfrom app.repositories.prompt_repository import PromptRepository\n\n\nclass PromptService:\n\n def __init__(self, repository: PromptRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n # 拆分台词prompt验证\n def validate_prompt_with_DUBBING(self, content: str):\n REQUIRED_BLOCKS = [\n # (\"\", \"\", \"{possible_characters}\"),\n # (\"\", \"\", \"{possible_emotions}\"),\n # (\"\", \"\", \"{possible_strengths}\"),\n (\"\", \"\", \"{novel_content}\"),\n ]\n for start, end, placeholder in REQUIRED_BLOCKS:\n if start not in content or end not in content or placeholder not in content:\n return False\n return True\n\n # 创建默认提示词\n def create_default_prompt(self):\n task = TaskEnum.DUBBING\n name = \"默认拆分台词提示词\"\n description = \"默认拆分台词提示词\"\n content = get_prompt_str()\n self.create_prompt(PromptEntity(name=name, description=description, content=content, task=task))\n return True\n\n def create_prompt(self, entity: PromptEntity):\n \"\"\"创建新提示词\n - 检查同名提示词是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n prompt = self.repository.get_by_name(entity.name)\n if prompt:\n return None\n # 判断task是否存在于task_enum中\n if entity.task not in TaskEnum:\n return None\n\n # 验证拆分台词的提示词\n if entity.task == TaskEnum.DUBBING:\n isValid = self.validate_prompt_with_DUBBING(entity.content)\n if not isValid:\n return None\n\n # 手动将entity转化为po\n po = PromptPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = PromptEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_prompt(self, prompt_id: int) -> PromptEntity | None:\n \"\"\"根据 ID 查询提示词\"\"\"\n po = self.repository.get_by_id(prompt_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = PromptEntity(**data)\n return res\n\n def get_all_prompts(self) -> Sequence[PromptEntity]:\n \"\"\"获取所有提示词列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n PromptEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_prompt(self, prompt_id: int, data:dict) -> bool:\n \"\"\"更新提示词\n - 可以只更新部分字段\n - 检查同名冲突\n \"\"\"\n name = data[\"name\"]\n task = data.get(\"task\")\n if self.repository.get_by_name(name) and self.repository.get_by_name(name).id != prompt_id:\n return False\n # 如果改的是content\n\n if TaskEnum(task) == TaskEnum.DUBBING:\n if not self.validate_prompt_with_DUBBING(content=data['content']):\n return False\n\n self.repository.update(prompt_id, data)\n return True\n\n def delete_prompt(self, prompt_id: int) -> bool:\n \"\"\"删除提示词\n - 可以添加业务校验，例如提示词下有章节是否允许删除\n - 后续需要级联删除所有章节内容\n \"\"\"\n res = self.repository.delete(prompt_id)\n return res\n # 根据task 获取提示词列表\n def get_prompt_by_task(self, task: str) -> Sequence[PromptEntity]:\n pos = self.repository.get_by_task(task)\n entities = [\n PromptEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n # 获取所有的stak\n def get_all_tasks(self) -> Sequence[str]:\n # 这些写死了，后面也在这添加，不改成数据库\n return list(TaskEnum)\n\n\n # def test_prompt(self, entity: PromptEntity):\n # \"\"\"测试提示词\"\"\"\n # # 按逗号划分模型名称\n # if entity.api_base_url is None or entity.api_key is None or entity.model_list is None:\n # return False\n # model_lists = entity.model_list.split(\",\")\n # llm = LLMEngine(entity.api_key, entity.api_base_url, model_lists[0])\n # res = llm.generate_text_test(\"你好\")\n # if res is not None:\n # return True\n # return False\n\n# 根据名字获取提示词\n def get_prompt_by_name(self, name: str) -> PromptEntity | None:\n \"\"\"根据名字获取提示词\"\"\"\n po = self.repository.get_by_name(name)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = PromptEntity(**data)\n return res\n\n\n\n\n"
},
{
"path": "SonicVale/app/services/role_service.py",
"content": "from sqlalchemy import Sequence\n\nfrom app.entity.role_entity import RoleEntity\nfrom app.models.po import RolePO\nfrom app.repositories.role_repository import RoleRepository\n\n\nclass RoleService:\n\n def __init__(self, repository: RoleRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_role(self, entity: RoleEntity):\n \"\"\"创建新角色\n - 检查同名角色是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n\n role = self.repository.get_by_name(entity.name, entity.project_id)\n if role:\n return None\n # 手动将entity转化为po\n po = RolePO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = RoleEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_role(self, role_id: int) -> RoleEntity | None:\n \"\"\"根据 ID 查询角色\"\"\"\n po = self.repository.get_by_id(role_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = RoleEntity(**data)\n return res\n\n def get_all_roles(self,project_id: int) -> Sequence[RoleEntity]:\n \"\"\"获取所有角色列表\"\"\"\n pos = self.repository.get_all(project_id)\n # pos -> entities\n\n entities = [\n RoleEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_role(self, role_id: int, data:dict) -> bool:\n \"\"\"更新角色\n - 可以只更新部分字段\n - 检查同名冲突\n - 检查project_id不能改变\n \"\"\"\n name = data[\"name\"]\n project_id = data[\"project_id\"]\n if self.repository.get_by_name(name, project_id) and self.repository.get_by_name(name,project_id).id != role_id:\n return False\n po = self.repository.get_by_id(role_id)\n # 防止改变project_id\n if po.project_id != project_id:\n return False\n self.repository.update(role_id, data)\n return True\n\n def delete_role(self, role_id: int) -> bool:\n \"\"\"删除角色\n \"\"\"\n res = self.repository.delete(role_id)\n return res\n"
},
{
"path": "SonicVale/app/services/strength_service.py",
"content": "from sqlalchemy import Sequence\n\nfrom app.entity.strength_entity import StrengthEntity\nfrom app.models.po import StrengthPO\nfrom app.repositories.strength_repository import StrengthRepository\n\n\nclass StrengthService:\n\n def __init__(self, repository: StrengthRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def create_strength(self, entity: StrengthEntity):\n \"\"\"创建新情绪强弱枚举\n - 检查同名情绪强弱枚举是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n\n strength = self.repository.get_by_name(entity.name)\n if strength:\n return None\n # 手动将entity转化为po\n po = StrengthPO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = StrengthEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_strength(self, strength_id: int) -> StrengthEntity | None:\n \"\"\"根据 ID 查询情绪强弱枚举\"\"\"\n po = self.repository.get_by_id(strength_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = StrengthEntity(**data)\n return res\n\n def get_all_strengths(self) -> Sequence[StrengthEntity]:\n \"\"\"获取所有情绪强弱枚举列表\"\"\"\n pos = self.repository.get_all()\n # pos -> entities\n\n entities = [\n StrengthEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_strength(self, strength_id: int, data:dict) -> bool:\n \"\"\"更新情绪强弱枚举\n - 可以只更新部分字段\n \"\"\"\n\n name = data.get(\"name\")\n if name and self.repository.get_by_name(name):\n return False\n self.repository.update(strength_id, data)\n return True\n\n def delete_strength(self, strength_id: int) -> bool:\n \"\"\"删除情绪强弱枚举\n \"\"\"\n res = self.repository.delete(strength_id)\n return res\n\n def get_strength_by_name(self, name: str) -> StrengthEntity | None:\n \"\"\"根据名称查询情绪强弱枚举\"\"\"\n po = self.repository.get_by_name(name)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = StrengthEntity(**data)\n return res\n"
},
{
"path": "SonicVale/app/services/tts_provider_service.py",
"content": "import requests\nimport logging\nfrom sqlalchemy import Sequence\n\nfrom app.entity.tts_provider_entity import TTSProviderEntity\nfrom app.models.po import TTSProviderPO\nfrom app.repositories.tts_provider_repository import TTSProviderRepository\n\n\nclass TTSProviderService:\n\n def __init__(self, repository: TTSProviderRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n\n def get_all_tts_providers(self) -> list[TTSProviderEntity]:\n \"\"\"查询所有tts供应商\"\"\"\n pos = self.repository.get_all()\n res = [TTSProviderEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}) for po in pos]\n return res\n\n def get_tts_provider(self, tts_provider_id: int) -> TTSProviderEntity | None:\n \"\"\"根据 ID 查询tts供应商\"\"\"\n po = self.repository.get_by_id(tts_provider_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = TTSProviderEntity(**data)\n return res\n\n\n def update_tts_provider(self, tts_provider_id: int, data:dict) -> bool:\n \"\"\"更新tts供应商\n - 可以只更新部分字段\n - 检查同名冲突\n - 检查project_id不能改变\n \"\"\"\n name = data[\"name\"]\n if self.repository.get_by_name(name) and self.repository.get_by_name(name).id != tts_provider_id:\n return False\n self.repository.update(tts_provider_id, data)\n return True\n\n def delete_tts_provider(self, tts_provider_id: int) -> bool:\n \"\"\"删除tts供应商\n \"\"\"\n res = self.repository.delete(tts_provider_id)\n return res\n\n def create_default_tts_provider(self):\n \"\"\"创建默认的tts供应商\"\"\"\n if self.repository.get_by_name(\"index_tts\") :\n return\n if self.repository.get_by_id(1) :\n return\n po = TTSProviderPO(name=\"index_tts\", id=1,status=1, api_base_url=\"\", api_key=\"\")\n self.repository.create(po)\n\n def test_tts_provider(self, entity: TTSProviderEntity):\n # 拿到url\n api_base_url = entity.api_base_url\n if not api_base_url:\n return False\n # ping api\n # 调用\n try:\n resp = requests.get(api_base_url, timeout=5)\n\n # 如果返回 200-399 都认为是通的（有些服务会 302 重定向）\n if 200 <= resp.status_code < 400:\n try:\n data = resp.json()\n if \"endpoints\" in data:\n return True\n else:\n logging.error(\"TTS provider test failed: 'endpoints' missing in response\")\n return False\n except ValueError:\n logging.error(\"TTS provider test failed: response is not valid JSON\")\n return False\n else:\n logging.error(\"TTS provider test failed: status %s\", resp.status_code)\n return False\n\n except Exception as e:\n logging.exception(\"TTS provider test failed: %s\", e)\n return False\n\n\n\n"
},
{
"path": "SonicVale/app/services/voice_service.py",
"content": "import json\nimport os\nimport shutil\nimport tempfile\nimport zipfile\nfrom typing import List, Tuple\n\nfrom sqlalchemy import Sequence\n\nfrom app.core.audio_engin import AudioProcessor\nfrom app.dto.voice_dto import VoiceAudioProcessDTO\nfrom app.entity.voice_entity import VoiceEntity\nfrom app.models.po import VoicePO\nfrom app.repositories.multi_emotion_voice_repository import MultiEmotionVoiceRepository\nfrom app.repositories.voice_repository import VoiceRepository\n\n\nclass VoiceService:\n\n def __init__(self, repository: VoiceRepository,multi_emotion_voice_repository: MultiEmotionVoiceRepository):\n \"\"\"注入 repository\"\"\"\n self.repository = repository\n self.multi_emotion_voice_repository = multi_emotion_voice_repository\n\n def create_voice(self, entity: VoiceEntity):\n \"\"\"创建新音色\n - 检查同名音色是否存在\n - 如果存在，抛出异常或返回错误\n - 调用 repository.create 插入数据库\n \"\"\"\n\n voice = self.repository.get_by_name(entity.name, entity.tts_provider_id)\n if voice:\n return None\n # 手动将entity转化为po\n po = VoicePO(**entity.__dict__)\n res = self.repository.create(po)\n\n # res(po) --> entity\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n entity = VoiceEntity(**data)\n\n # 将po转化为entity\n return entity\n\n\n def get_voice(self, voice_id: int) -> VoiceEntity | None:\n \"\"\"根据 ID 查询音色\"\"\"\n po = self.repository.get_by_id(voice_id)\n if not po:\n return None\n data = {k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")}\n res = VoiceEntity(**data)\n return res\n\n def get_all_voices(self,tts_provider_id: int) -> Sequence[VoiceEntity]:\n \"\"\"获取所有音色列表\"\"\"\n pos = self.repository.get_all(tts_provider_id)\n # pos -> entities\n\n entities = [\n VoiceEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n return entities\n\n def update_voice(self, voice_id: int, data:dict) -> bool:\n \"\"\"更新音色\n - 可以只更新部分字段\n - 检查同名冲突\n - 检查project_id不能改变\n \"\"\"\n name = data[\"name\"]\n tts_provider_id = data[\"tts_provider_id\"]\n if self.repository.get_by_name(name, tts_provider_id) and self.repository.get_by_name(name,tts_provider_id).id != voice_id:\n return False\n po = self.repository.get_by_id(voice_id)\n # 防止改变project_id\n if po.tts_provider_id != tts_provider_id:\n return False\n self.repository.update(voice_id, data)\n return True\n\n def delete_voice(self, voice_id: int) -> bool:\n \"\"\"删除音色,需要保证事务\n \"\"\"\n\n res = self.repository.delete(voice_id)\n self.multi_emotion_voice_repository.delete_multi_emotion_voice_by_voice_id(voice_id)\n return res\n\n def export_voices(self, tts_provider_id: int, export_path: str, ids: List[int] | None = None) -> str:\n \"\"\"导出音色库到zip文件\n - 获取所有音色\n - 将音色信息和对应的音频文件打包到zip\n - 返回zip文件路径\n \"\"\"\n if ids is None:\n voices = self.get_all_voices(tts_provider_id)\n else:\n pos = self.repository.get_by_ids(tts_provider_id, ids)\n voices = [\n VoiceEntity(**{k: v for k, v in po.__dict__.items() if not k.startswith(\"_\")})\n for po in pos\n ]\n if not voices:\n return None\n\n # 确保导出目录存在\n os.makedirs(os.path.dirname(export_path) if os.path.dirname(export_path) else \".\", exist_ok=True)\n\n # 创建zip文件\n with zipfile.ZipFile(export_path, 'w', zipfile.ZIP_DEFLATED) as zipf:\n # 准备音色元数据\n voices_metadata = []\n \n for voice in voices:\n voice_data = {\n \"name\": voice.name,\n \"description\": voice.description,\n \"is_multi_emotion\": voice.is_multi_emotion,\n \"reference_file\": None\n }\n \n # 如果有参考音频文件，添加到zip\n if voice.reference_path and os.path.exists(voice.reference_path):\n # 保持原文件名\n file_name = os.path.basename(voice.reference_path)\n # 使用音色名称作为子目录，避免文件名冲突\n archive_path = f\"voices/{voice.name}/{file_name}\"\n zipf.write(voice.reference_path, archive_path)\n voice_data[\"reference_file\"] = archive_path\n \n voices_metadata.append(voice_data)\n \n # 写入元数据文件\n metadata_json = json.dumps(voices_metadata, ensure_ascii=False, indent=2)\n zipf.writestr(\"voices_metadata.json\", metadata_json)\n \n return export_path\n\n def import_voices(self, tts_provider_id: int, zip_path: str, target_dir: str) -> Tuple[int, int, List[str]]:\n \"\"\"从zip文件导入音色库\n - 解压zip文件\n - 将音频文件复制到指定目录\n - 添加音色到数据库（跳过重名的）\n - 返回: (成功数量, 跳过数量, 跳过的音色名称列表)\n \"\"\"\n if not os.path.exists(zip_path):\n raise FileNotFoundError(f\"zip文件不存在: {zip_path}\")\n \n # 确保目标目录存在\n os.makedirs(target_dir, exist_ok=True)\n \n success_count = 0\n skipped_count = 0\n skipped_names = []\n \n # 创建临时目录解压\n with tempfile.TemporaryDirectory() as temp_dir:\n # 解压zip文件\n with zipfile.ZipFile(zip_path, 'r') as zipf:\n zipf.extractall(temp_dir)\n \n # 读取元数据\n metadata_path = os.path.join(temp_dir, \"voices_metadata.json\")\n if not os.path.exists(metadata_path):\n raise ValueError(\"无效的音色库文件：缺少voices_metadata.json\")\n \n with open(metadata_path, 'r', encoding='utf-8') as f:\n voices_metadata = json.load(f)\n \n for voice_data in voices_metadata:\n voice_name = voice_data[\"name\"]\n \n # 检查是否已存在同名音色\n existing = self.repository.get_by_name(voice_name, tts_provider_id)\n if existing:\n skipped_count += 1\n skipped_names.append(voice_name)\n continue\n \n reference_path = None\n \n # 如果有参考音频文件，复制到目标目录\n if voice_data.get(\"reference_file\"):\n source_file = os.path.join(temp_dir, voice_data[\"reference_file\"])\n if os.path.exists(source_file):\n # 使用音色名称作为文件名，保留原扩展名\n file_ext = os.path.splitext(source_file)[1]\n file_name = f\"{voice_name}{file_ext}\"\n dest_file = os.path.join(target_dir, file_name)\n shutil.copy2(source_file, dest_file)\n reference_path = dest_file\n \n # 创建音色实体\n entity = VoiceEntity(\n name=voice_name,\n tts_provider_id=tts_provider_id,\n reference_path=reference_path,\n description=voice_data.get(\"description\"),\n is_multi_emotion=voice_data.get(\"is_multi_emotion\", 0)\n )\n \n # 保存到数据库\n po = VoicePO(**entity.__dict__)\n self.repository.create(po)\n success_count += 1\n \n return success_count, skipped_count, skipped_names\n\n def process_audio(self, dto: VoiceAudioProcessDTO) -> bool:\n \"\"\"处理音色参考音频\n - 变速、音量调整\n - 裁剪/删除区间\n - 添加/裁剪末尾静音\n - 指定位置插入静音\n \"\"\"\n audio_path = dto.audio_path\n if not os.path.exists(audio_path):\n raise FileNotFoundError(audio_path)\n \n processor = AudioProcessor(audio_path)\n \n start_ms = dto.start_ms\n end_ms = dto.end_ms\n speed = dto.speed\n volume = dto.volume\n current_ms = dto.current_ms\n silence_sec = dto.silence_sec\n \n # ---------- (1) 优先裁剪 ----------\n if start_ms is not None and end_ms is not None and end_ms > start_ms:\n processor.cut(start_ms, end_ms)\n \n # ---------- (2) 插入静音 ----------\n elif current_ms is not None and silence_sec is not None and silence_sec != 0:\n processor.insert_silence(current_ms, silence_sec)\n \n # ---------- (3) 末尾静音/裁剪 ----------\n elif current_ms is None and silence_sec is not None and silence_sec != 0:\n processor.append_silence(silence_sec)\n \n # ---------- (4) 音量 + 变速 ----------\n if speed != 1.0:\n processor.change_speed(speed)\n if volume != 1.0:\n processor.change_volume(volume)\n \n return True\n\n def copy_voice(self, source_voice_id: int, new_name: str, target_dir: str = None) -> VoiceEntity:\n \"\"\"复制音色\n - 获取源音色信息\n - 复制音频文件到目标目录\n - 创建新音色记录\n - 返回新音色实体\n \"\"\"\n # 获取源音色\n source_voice = self.get_voice(source_voice_id)\n if not source_voice:\n raise ValueError(\"源音色不存在\")\n \n # 检查新名称是否已存在\n existing = self.repository.get_by_name(new_name, source_voice.tts_provider_id)\n if existing:\n raise ValueError(f\"音色名称 '{new_name}' 已存在\")\n \n new_reference_path = None\n \n # 处理音频文件复制\n if source_voice.reference_path and os.path.exists(source_voice.reference_path):\n # 确定目标目录\n if target_dir and target_dir.strip():\n dest_dir = target_dir.strip()\n else:\n # 使用源音频所在目录\n dest_dir = os.path.dirname(source_voice.reference_path)\n \n # 确保目标目录存在\n os.makedirs(dest_dir, exist_ok=True)\n \n # 获取源文件扩展名\n file_ext = os.path.splitext(source_voice.reference_path)[1]\n # 使用新音色名作为文件名\n new_file_name = f\"{new_name}{file_ext}\"\n new_reference_path = os.path.join(dest_dir, new_file_name)\n \n # 复制文件\n shutil.copy2(source_voice.reference_path, new_reference_path)\n \n # 创建新音色实体\n new_entity = VoiceEntity(\n name=new_name,\n tts_provider_id=source_voice.tts_provider_id,\n reference_path=new_reference_path,\n description=source_voice.description,\n is_multi_emotion=source_voice.is_multi_emotion\n )\n \n # 保存到数据库\n po = VoicePO(**new_entity.__dict__)\n res = self.repository.create(po)\n \n # 返回新建的音色实体\n data = {k: v for k, v in res.__dict__.items() if not k.startswith(\"_\")}\n return VoiceEntity(**data)\n"
},
{
"path": "SonicVale/requirements.txt",
"content": "fastapi==0.119.0\nnumba==0.61.2\nnumpy==2.3.3\nopenai==2.8.0\nopenpyxl==3.1.5\npydantic==2.12.2\npypinyin==0.55.0\nRequests==2.32.5\nsoundfile==0.13.1\nSQLAlchemy==2.0.44\nstarlette==0.48.0\nuvicorn==0.37.0\n"
},
{
"path": "sonicvale-front/.gitignore",
"content": "# Logs\nlogs\n*.log\nnpm-debug.log*\nyarn-debug.log*\nyarn-error.log*\npnpm-debug.log*\nlerna-debug.log*\n\nnode_modules\ndist\nrelease\ndist-ssr\n*.local\n\n*.exe\n\n# Editor directories and files\n.vscode/*\n!.vscode/extensions.json\n.idea\n.DS_Store\n*.suo\n*.ntvs*\n*.njsproj\n*.sln\n*.sw?\n"
},
{
"path": "sonicvale-front/.vscode/extensions.json",
"content": "{\n \"recommendations\": [\"Vue.volar\"]\n}\n"
},
{
"path": "sonicvale-front/README.md",
"content": "# Vue 3 + Vite\n\nThis template should help get you started developing with Vue 3 in Vite. The template uses Vue 3 `\n