neverlib 0.2.8__py3-none-any.whl → 0.3.0__py3-none-any.whl

neverlib/Docs/audio_aug/test_volume.ipynb

@@ -350,16 +350,16 @@
     "print(\"\\n音量增强方法分析总结:\")\n",
     "print(\"=\" * 80)\n",
     "print(\"1. Linear增强 (volume_aug_linmax):\")\n",
-    "print(\"   - 基于峰值的线性增益，目标为最大峰值为指定的线性值\")\n",
-    "print(\"   - 特点：保持动态范围不变，整体增强或减弱\")\n",
+    "print(\"   - 基于峰值的线性增益, 目标为最大峰值为指定的线性值\")\n",
+    "print(\"   - 特点：保持动态范围不变, 整体增强或减弱\")\n",
     "print(\"   - 适用场景：需要精确控制峰值而不改变音频动态特性的场合\")\n",
     "print(\"\\n2. dBrms增强 (volume_aug_dbrms):\")\n",
-    "print(\"   - 基于RMS电平的增益，目标为指定的dB电平\")\n",
-    "print(\"   - 特点：以能量均值为基准，更接近人耳感知\")\n",
+    "print(\"   - 基于RMS电平的增益, 目标为指定的dB电平\")\n",
+    "print(\"   - 特点：以能量均值为基准, 更接近人耳感知\")\n",
     "print(\"   - 适用场景：需要统一音频能量电平的场合\")\n",
     "print(\"\\n3. LUFS增强 (volume_aug_lufs):\")\n",
-    "print(\"   - 基于国际响度标准的增益，符合广播标准\")\n",
-    "print(\"   - 特点：考虑人耳频率加权，最接近人耳响度感知\")\n",
+    "print(\"   - 基于国际响度标准的增益, 符合广播标准\")\n",
+    "print(\"   - 特点：考虑人耳频率加权, 最接近人耳响度感知\")\n",
     "print(\"   - 适用场景：广播、流媒体、专业音频处理等需要符合响度标准的场合\")\n",
     "print(\"=\" * 80)"
    ]
@@ -386,9 +386,9 @@
     "3. **LUFS增强 (volume_aug_lufs)**\n",
     "   - 基于感知响度单位(LUFS)的增益\n",
     "   - 符合广播标准的响度归一化\n",
-    "   - 考虑人耳感知加权，最接近人耳响度感知\n",
+    "   - 考虑人耳感知加权, 最接近人耳响度感知\n",
     "\n",
-    "各方法适用于不同场景，可根据需求选择合适的音量增强方式。"
+    "各方法适用于不同场景, 可根据需求选择合适的音量增强方式. "
    ]
   }
  ],

neverlib/Docs/filter/biquad.ipynb

@@ -39,7 +39,7 @@
    "source": [
     "# 设计高通滤波器系数\n",
     "fs = 16000  # 采样率\n",
-    "fc = 70  # 截止频率（Hz）\n",
+    "fc = 70  # 截止频率(Hz)\n",
     "# 输入信号\n",
     "input_signal = [0.5, 0.8, 1.0, 0.7, -0.2, -0.6, -0.8, -0.3, -0.3, -0.3, -0.3]\n",
     "\n",

neverlib/Docs/filter/filter_family.ipynb

@@ -10,7 +10,7 @@
    "source": [
     "# 音频滤波器示例\n",
     "\n",
-    "这个notebook展示了如何使用 neverlib.filter 中的各种滤波器。我们将展示：\n",
+    "这个notebook展示了如何使用 neverlib.filter 中的各种滤波器. 我们将展示：\n",
     "1. 各种滤波器的频率响应\n",
     "2. 实际音频处理示例\n",
     "3. 与scipy原生函数的对比\n",
@@ -80,7 +80,7 @@
    "source": [
     "# 滤波器示例\n",
     "\n",
-    "下面我们将展示各种滤波器的频率响应。我们使用以下参数：\n",
+    "下面我们将展示各种滤波器的频率响应. 我们使用以下参数：\n",
     "- 采样率 fs = 16000 Hz\n",
     "- 截止频率 fc = 1000 Hz\n",
     "- 品质因数 Q = 1/√2 ≈ 0.707 (Butterworth响应)\n",
@@ -248,7 +248,7 @@
    "source": [
     "# 实际音频处理示例\n",
     "\n",
-    "下面我们将展示如何使用这些滤波器处理实际的音频信号。我们将：\n",
+    "下面我们将展示如何使用这些滤波器处理实际的音频信号. 我们将：\n",
     "1. 加载一个音频文件\n",
     "2. 应用不同的滤波器\n",
     "3. 对比处理前后的频谱\n"
@@ -318,7 +318,7 @@
    "source": [
     "# EQ滤波\n",
     "\n",
-    "通过多个Biquand滤波器, 组成一个EQ滤波器, 实现音频的均衡处理。"
+    "通过多个Biquand滤波器, 组成一个EQ滤波器, 实现音频的均衡处理. "
    ]
   },
   {

neverlib/Docs/vad/VAD_WebRTC.ipynb

@@ -12,12 +12,12 @@
     "mode 0~3\n",
     "\n",
     "0: 最低的语音检测敏感度, \n",
-    "- 认为背景噪声不是语音, 适合环境较安静, 背景噪声少的情况。\n",
-    "- 适合环境较安静, 背景噪声少的情况。\n",
+    "- 认为背景噪声不是语音, 适合环境较安静, 背景噪声少的情况. \n",
+    "- 适合环境较安静, 背景噪声少的情况. \n",
     "\n",
     "3: 最高的语音检测敏感度, \n",
-    "- VAD 会非常积极地尝试将任何噪声过滤掉, 只有明确的语音才会被认为是语音。\n",
-    "- 适合环境较吵, 背景噪声多的情况。"
+    "- VAD 会非常积极地尝试将任何噪声过滤掉, 只有明确的语音才会被认为是语音. \n",
+    "- 适合环境较吵, 背景噪声多的情况. "
    ]
   },
   {

neverlib/Docs/vad/VAD_whisper.ipynb

@@ -7,7 +7,7 @@
    "source": [
     "# whisper base模型\n",
     "\n",
-    "whisper 检测的 词与词之间的VAD 都是连着的。但其实音频不是\n",
+    "whisper 检测的 词与词之间的VAD 都是连着的. 但其实音频不是\n",
     "\n",
     "而且Whisper的VAD并没有直接提供调参接口, 所以无法调整VAD的参数"
    ]
@@ -170,7 +170,7 @@
    "source": [
     "# whisper large-v3 模型\n",
     "\n",
-    "词与词之间的VAD 都是连着的。但其实音频不是\n",
+    "词与词之间的VAD 都是连着的. 但其实音频不是\n",
     "```\n",
     "# Word:  Sometimes, Start: 0.00s, End: 0.70s\n",
     "# Word:  he, Start: 0.70s, End: 0.86s\n",

neverlib/LLM/__init__.py

@@ -0,0 +1,37 @@
+# This file is auto-generated. Do NOT edit manually.
+# Generated by neverlib.QA.gen_init
+
+from typing import TYPE_CHECKING, Any
+
+if TYPE_CHECKING:
+    # 仅在类型检查时导入, 提供IDE补全支持
+    from .bailian import AliyunSession, Aliyun_App
+    from .image import chat_with_ai_image, image_to_base64_uri
+    from .text import chat
+
+# 运行时使用懒加载
+from lazy_loader import attach
+
+__getattr__, __dir__, __all__ = attach(
+    __name__,
+    submodules=[
+        "bailian",
+        "image",
+        "text",
+    ],
+    submod_attrs={
+        "bailian": ['AliyunSession', 'Aliyun_App'],
+        "image": ['chat_with_ai_image', 'image_to_base64_uri'],
+        "text": ['chat'],
+    }
+)
+
+# 显式声明 __all__ 以便 IDE 识别
+if TYPE_CHECKING:
+    __all__ = [
+        'AliyunSession',
+        'Aliyun_App',
+        'chat_with_ai_image',
+        'image_to_base64_uri',
+        'chat',
+    ]

neverlib/LLM/bailian.py

@@ -0,0 +1,342 @@
+'''
+Author: 凌逆战 | Never
+Date: 2026-01-07 00:47:30
+Description: 阿里云百炼应用调用封装
+'''
+import os
+from typing import Optional, Dict, Any, List
+
+
+def Aliyun_App(inputs: str, prompt: Optional[str] = None,
+               api_key: Optional[str] = None,
+               app_id: Optional[str] = None,
+               stream: bool = False,
+               messages: Optional[List[Dict[str, str]]] = None,
+               workspace: Optional[str] = None,
+               temperature: Optional[float] = None,
+               top_p: Optional[float] = None,
+               top_k: Optional[int] = None,
+               return_full_response: bool = False,
+               **kwargs
+               ) -> str | Dict[str, Any]:
+    """
+    调用阿里云百炼应用
+
+    Args:
+        inputs: 输入文本
+        prompt: 提示词(可选, 会与inputs拼接)
+        api_key: API密钥
+        app_id: 应用ID(必填)
+        stream: 是否启用流式输出
+        messages: 对话消息列表, 格式: [{'role': 'user', 'content': '问题'}, {'role': 'assistant', 'content': '回答'}]
+        workspace: 工作空间
+        temperature: 温度参数 (0-2), 控制随机性和多样性
+            - 值越高：输出越随机、越有创造性
+            - 值越低：输出越确定、越保守
+            - 建议: 财经分析用0.1-0.3, 创意内容用0.7-1.0
+        top_p: 核采样参数(Nucleus Sampling), 范围 (0-1)
+            - 作用: 从累积概率达到p的候选词中采样
+            - 0.1: 只考虑累积概率前10%的词(更确定)
+            - 0.9: 考虑累积概率前90%的词(更多样)
+            - 建议: 专业分析用0.1-0.3, 常规对话用0.7, 创意内容用0.9
+        top_k: Top-K采样候选集大小
+            - 作用: 每次生成只从概率最高的k个词中选择
+            - 值越小：确定性越高(如10-30)
+            - 值越大：随机性越高(如50-100)
+            - 默认0: 禁用top_k策略, 只使用top_p
+        return_full_response: 是否返回完整响应(包含thoughts、session_id等)
+        **kwargs: 其他API支持的参数
+            - enable_thinking: 启用深度思考模式
+            - has_thoughts: 包含思考内容
+            - seed: 随机种子(用于复现结果)
+
+    Returns:
+        str: 默认返回响应文本字符串
+        dict: 当 return_full_response=True 时返回完整响应字典：
+            {
+                "status": "success" | "error",
+                "text": "响应文本",
+                "request_id": "请求ID",
+                "session_id": "会话ID(可选)",
+                "finish_reason": "结束原因(可选)",
+                "thoughts": [...]  # 思考过程(可选),
+                "doc_references": [...]  # 文档引用(可选),
+            }
+
+    Raises:
+        ValueError: 参数验证失败
+
+    来源: https://help.aliyun.com/zh/model-studio/call-single-agent-application/
+    """
+    try:
+        from http import HTTPStatus
+        from dashscope import Application
+    except ImportError as e:
+        if 'dashscope' in str(e):
+            raise ImportError("请安装dashscope库: pip install dashscope")
+        else:
+            raise ImportError(f"导入失败: {e}. 如果是因为HTTPStatus, 请确保使用Python 3.5或更高版本")
+
+    if not app_id:
+        raise ValueError("app_id参数不能为空")
+
+    if not inputs and not prompt:
+        raise ValueError("inputs和prompt至少需要提供一个")
+
+    # 构建输入
+    if prompt and inputs:
+        llm_input = f"{prompt}\n{inputs}"
+    elif prompt:
+        llm_input = prompt
+    else:
+        llm_input = inputs
+
+    # 构建API调用参数
+    call_params = {
+        'app_id': app_id,
+        'prompt': llm_input,
+        'stream': stream
+    }
+
+    # 添加可选参数
+    if api_key:
+        call_params['api_key'] = api_key
+    if messages:
+        call_params['messages'] = messages
+    if workspace:
+        call_params['workspace'] = workspace
+    if temperature is not None:
+        call_params['temperature'] = temperature
+    if top_p is not None:
+        call_params['top_p'] = top_p
+    if top_k is not None:
+        call_params['top_k'] = top_k
+
+    # 添加额外参数
+    call_params.update(kwargs)
+
+    # 调用API
+    response = Application.call(**call_params)
+
+    # 检查响应状态
+    if response.status_code != HTTPStatus.OK:
+        if return_full_response:
+            return {
+                "status": "error",
+                "text": "",
+                "message": f"API调用失败: [{response.code}] {response.message}",
+                "code": response.code
+            }
+        return "API调用失败"
+
+    # 提取文本
+    output_text = response.output.text if hasattr(response.output, 'text') else str(response.output)
+
+    # 如果需要完整响应
+    if return_full_response:
+        result = {
+            "status": "success",
+            "text": output_text,
+            "request_id": response.request_id
+        }
+
+        # 添加 session_id
+        if hasattr(response.output, 'session_id') and response.output.session_id:
+            result['session_id'] = response.output.session_id
+
+        # 添加 finish_reason
+        if hasattr(response.output, 'finish_reason') and response.output.finish_reason:
+            result['finish_reason'] = response.output.finish_reason
+
+        # 添加思考过程(深度思考)
+        if hasattr(response.output, 'thoughts') and response.output.thoughts:
+            result['thoughts'] = [
+                {
+                    'thought': t.thought,
+                    'action_type': t.action_type,
+                    'action_name': t.action_name,
+                    'observation': t.observation
+                }
+                for t in response.output.thoughts
+            ]
+
+        # 添加文档引用(RAG场景)
+        if hasattr(response.output, 'doc_references') and response.output.doc_references:
+            result['doc_references'] = [
+                {
+                    'title': ref.title,
+                    'text': ref.text,
+                    'doc_name': ref.doc_name
+                }
+                for ref in response.output.doc_references
+            ]
+
+        return result
+
+    return output_text
+
+
+class AliyunSession:
+    """
+    阿里云百炼应用会话管理类
+    自动维护对话历史, 支持多轮对话
+
+    使用示例:
+        session = AliyunSession(app_id="xxx", api_key="xxx", temperature=0.3)
+
+        # 第一轮对话
+        result1 = session.chat("介绍一下股票市场")
+
+        # 第二轮对话(自动带上历史)
+        result2 = session.chat("那A股和港股有什么区别?")
+
+        # 查看历史
+        print(session.get_history_text())
+    """
+
+    def __init__(self, app_id: str, api_key: Optional[str] = None,
+                 max_history: int = 10, **default_params):
+        """
+        初始化会话
+
+        Args:
+            app_id: 应用ID
+            api_key: API密钥
+            max_history: 最多保留的历史轮数(默认10轮, 超过会自动删除最早的)
+            **default_params: 默认参数(如temperature, top_p等)
+        """
+        self.app_id = app_id
+        self.api_key = api_key
+        self.max_history = max_history
+        self.default_params = default_params
+        self.messages: List[Dict[str, str]] = []
+        self.session_id: Optional[str] = None
+
+    def chat(self, inputs: str, prompt: Optional[str] = None, **kwargs) -> str | Dict[str, Any]:
+        """
+        发送消息并自动维护历史
+
+        Args:
+            inputs: 用户输入
+            prompt: 系统提示词(可选)
+            **kwargs: 覆盖默认参数(如 return_full_response=True)
+
+        Returns:
+            str: 默认返回响应文本
+            dict: 当 kwargs 中包含 return_full_response=True 时返回完整响应
+        """
+        # 合并默认参数和传入参数
+        call_params = {**self.default_params, **kwargs}
+
+        # 调用API
+        result = Aliyun_App(
+            inputs=inputs,
+            prompt=prompt,
+            app_id=self.app_id,
+            api_key=self.api_key,
+            messages=self.messages.copy() if self.messages else None,
+            **call_params
+        )
+
+        # 如果调用成功, 添加到历史
+        # 添加用户消息
+        user_content = f"{prompt}\n{inputs}" if prompt else inputs
+        self.messages.append({"role": "user", "content": user_content})
+
+        # 添加助手回复
+        self.messages.append({"role": "assistant", "content": result})
+
+        # 限制历史长度(保留最近的 max_history 轮对话)
+        if len(self.messages) > self.max_history * 2:
+            removed = len(self.messages) - self.max_history * 2
+            self.messages = self.messages[removed:]
+
+        return result
+
+    def clear_history(self):
+        """清空对话历史"""
+        self.messages.clear()
+        self.session_id = None
+
+    def get_history(self) -> List[Dict[str, str]]:
+        """获取当前对话历史"""
+        return self.messages.copy()
+
+    def get_history_text(self) -> str:
+        """获取格式化的对话历史文本"""
+        history_lines = []
+        for msg in self.messages:
+            role = "用户" if msg["role"] == "user" else "助手"
+            history_lines.append(f"{role}: {msg['content'][:100]}...")
+        return "\n\n".join(history_lines)
+
+    def set_history(self, messages: List[Dict[str, str]]):
+        """设置对话历史(用于恢复会话)"""
+        self.messages = messages.copy()
+
+    def __repr__(self):
+        return f"AliyunSession(app_id={self.app_id}, history={len(self.messages) // 2}轮)"
+
+
+if __name__ == "__main__":
+    import dotenv
+    dotenv.load_dotenv()
+
+    # 使用示例1: 基本调用
+    print("=" * 50)
+    print("示例1: 基本调用")
+    print("=" * 50)
+    result = Aliyun_App(
+        inputs="你好, 请介绍一下最近的A股市场情况",
+        app_id=os.getenv("ALIYUN_APP_ID"),
+        api_key=os.getenv("DASHSCOPE_API_KEY")
+    )
+    print(f"响应: {result[:200]}...")
+
+    # 使用示例2: 使用会话管理类(自动维护历史)
+    print("\n" + "=" * 50)
+    print("示例2: 使用AliyunSession - 多轮对话")
+    print("=" * 50)
+
+    session = AliyunSession(
+        app_id=os.getenv("ALIYUN_APP_ID"),
+        api_key=os.getenv("DASHSCOPE_API_KEY"),
+        temperature=0.5,
+        max_history=5
+    )
+
+    # 第一轮
+    print("\n[第1轮]")
+    r1 = session.chat("介绍一下A股市场")
+    print(f"用户: 介绍一下A股市场")
+    print(f"助手: {r1[:150]}...")
+
+    # 第二轮(自动带上历史)
+    print("\n[第2轮]")
+    r2 = session.chat("那主要有哪些板块?")
+    print(f"用户: 那主要有哪些板块?")
+    print(f"助手: {r2[:150]}...")
+
+    # 查看会话信息
+    print(f"\n{session}")
+
+    # 使用示例3: 启用深度思考模式(返回完整响应)
+    print("\n" + "=" * 50)
+    print("示例3: 深度思考模式")
+    print("=" * 50)
+    result_thinking = Aliyun_App(
+        inputs="分析今日A股大盘走势",
+        app_id=os.getenv("ALIYUN_APP_ID"),
+        api_key=os.getenv("DASHSCOPE_API_KEY"),
+        return_full_response=True,
+        enable_thinking=True
+    )
+
+    if isinstance(result_thinking, dict):
+        print(f"状态: {result_thinking.get('status')}")
+        print(f"响应: {result_thinking.get('text', '')[:200]}...")
+
+        if result_thinking.get('thoughts'):
+            print(f"\n发现 {len(result_thinking['thoughts'])} 条思考过程")
+        if result_thinking.get('finish_reason'):
+            print(f"结束原因: {result_thinking['finish_reason']}")

neverlib/LLM/image.py

@@ -0,0 +1,73 @@
+'''
+Author: 凌逆战 | Never
+Date: 2026-01-07 00:44:24
+Description: 
+'''
+import os
+import base64
+import mimetypes
+from openai import OpenAI
+
+
+def image_to_base64_uri(file_path):
+    """
+    将本地图片文件转换为Base64编码的Data URI. 
+    """
+    # 自动猜测MIME类型
+    mime_type, _ = mimetypes.guess_type(file_path)
+    if mime_type is None:
+        # 如果无法猜测, 可以根据文件扩展名设置一个默认值
+        if file_path.lower().endswith(('.jpg', '.jpeg')):
+            mime_type = 'image/jpeg'
+        elif file_path.lower().endswith('.png'):
+            mime_type = 'image/png'
+        elif file_path.lower().endswith('.gif'):
+            mime_type = 'image/gif'
+        elif file_path.lower().endswith('.webp'):
+            mime_type = 'image/webp'
+        else:
+            print(
+                f"Warning: Could not determine MIME type for {file_path}. Skipping.")
+            return None  # 或者抛出错误
+
+    with open(file_path, "rb") as image_file:
+        binary_data = image_file.read()
+        base64_encoded_data = base64.b64encode(binary_data)
+        base64_string = base64_encoded_data.decode('utf-8')
+        return f"data:{mime_type};base64,{base64_string}"
+
+
+def chat_with_ai_image(inputs, image_path_list, prompt=None,
+                       model="doubao-1-5-vision-pro-32k-250115"):
+    """
+    使用豆包大模型分析图片和文本内容. 
+    支持多个图片, 图片可以是本地路径或网络URL. 
+    """
+    image_contents, messages = [], []
+    for image_path in image_path_list:
+        if "http" in image_path:    # 是网络图片
+            image_contents.append({
+                "type": "image_url",
+                "image_url": {"url": image_path}
+            })
+        else:                       # 是本地图片路径
+            base64_uri = image_to_base64_uri(image_path)
+            image_contents.append({
+                "type": "image_url",
+                "image_url": {"url": base64_uri}
+            })
+
+    image_contents.append({"type": "text", "text": inputs})
+    if prompt:
+        messages.append({"role": "system", "content": prompt})
+    messages.append({"role": "user", "content": image_contents})
+
+    client = OpenAI(
+        base_url="https://ark.cn-beijing.volces.com/api/v3",
+        api_key=os.environ.get("DOUBAO_API_KEY"),
+    )
+    response = client.chat.completions.create(
+        model=model,
+        messages=messages,
+    )
+    return response.choices[0].message.content

neverlib/LLM/text.py

@@ -0,0 +1,32 @@
+'''
+Author: 凌逆战 | Never
+Date: 2026-01-07 00:35:35
+Description: 
+'''
+
+
+def chat(inputs, prompt=None, model=None, timeout=180, base_url=None, api_key=None):
+    """
+    使用OpenAI API进行对话(非流式)
+    """
+    try:
+        from openai import OpenAI
+    except ImportError:
+        raise ImportError("请安装openai库: pip install openai")
+    assert model, "model不能为空"
+    assert base_url and api_key, "base_url和api_key不能为空"
+    client = OpenAI(base_url=base_url, api_key=api_key, timeout=timeout)
+
+    messages = []
+    if prompt:
+        messages.append({"role": "system", "content": prompt})
+    messages.append({"role": "user", "content": inputs})
+
+    completion = client.chat.completions.create(
+        model=model, messages=messages, timeout=timeout)
+
+    output = completion.choices[0].message.content
+    # 如果有引用, 则添加到输出中
+    if hasattr(completion, "references"):
+        output += "\n" + str(completion.references)
+    return output