x

CosyVoice V2/教程.txt

@@ -3,4 +3,9 @@ https://www.bilibili.com/video/BV1pnkFYmEAE/?spm_id_from=333.337.search-card.all
 根目录进入
 cd CosyVoice
 启动虚拟环境
-conda activate CosyVoice_V2
+
+conda activate CosyVoice_V2
+
+
+5. 仅返回最终答案/结果，禁止输出任何分析、推理、解释、步骤说明；
+6. 结果需简洁为纯文本，最短五个字，最多30个字，无多余文字、标点、格式（如无需“答案：”“总结：”等前缀）；

python/.idea/misc.xml

@@ -3,5 +3,5 @@
   <component name="Black">
     <option name="sdkName" value="E:\anaconda3\envs\zghs_shuiwu_api" />
   </component>
-  <component name="ProjectRootManager" version="2" project-jdk-name="python" project-jdk-type="Python SDK" />
+  <component name="ProjectRootManager" version="2" project-jdk-name="AIStreamTest" project-jdk-type="Python SDK" />
 </project>

python/.idea/python.iml

@@ -2,7 +2,7 @@
 <module type="PYTHON_MODULE" version="4">
   <component name="NewModuleRootManager">
     <content url="file://$MODULE_DIR$" />
-    <orderEntry type="jdk" jdkName="python" jdkType="Python SDK" />
+    <orderEntry type="jdk" jdkName="AIStreamTest" jdkType="Python SDK" />
     <orderEntry type="sourceFolder" forTests="false" />
   </component>
 </module>

python/WebSocketFrameHeader.py

@@ -0,0 +1,223 @@
+from enum import IntEnum
+import struct
+import json
+import gzip
+from typing import Optional, Union, Dict, Any, List
+
+
+# -------------------------- 协议常量定义 --------------------------
+class ProtocolConst:
+    PROTOCOL_VERSION = 0b0001  # 协议版本v1
+    HEADER_SIZE = 8  # 头部固定字节数（字节1~字节8）
+    MAX_BODY_SIZE = 1024 * 1024 * 10  # 最大包体大小（10MB，防止内存溢出）
+    STRING_ENCODING = "utf-8"  # 字符串默认编码（统一编码格式，避免乱码）
+
+
+# -------------------------- 枚举定义（业务约定） --------------------------
+class MessageType(IntEnum):
+    """消息类型（4位，字节1高4位）"""
+    PING = 0b0000  # 心跳
+    AUDIO_DATA = 0b0001  # 纯音频数据(pcm)
+    TEXT_MESSAGE = 0b0010  # 纯文本消息
+    AUDIO_TEXT_MIX = 0b0011  # 音频+文本混合数据
+    CONTROL_CMD = 0b0100  # 控制指令（暂停/继续等）
+
+class SerializationType(IntEnum):
+    """序列化方式（4位，字节2高4位）"""
+    RAW = 0b0000  # 原始二进制（音频等）
+    JSON = 0b0001  # JSON格式（文本/指令）
+    STRING = 0b0010  # 直接字符串（纯文本，UTF-8编码，无JSON包装）🔴新增类型
+
+class CompressionType(IntEnum):
+    """压缩方式（4位，字节2低4位）"""
+    NONE = 0b0000  # 无压缩
+    GZIP = 0b0001  # gzip压缩
+
+class ControlCommand(IntEnum):
+    """控制指令类型（配合MessageType.CONTROL_CMD使用）"""
+    HEARTBEAT = 0b0001  # 心跳响应
+    PAUSE = 0b0010  # 暂停
+    RESUME = 0b0011  # 继续
+    STOP = 0b0100  # 停止
+
+# -------------------------- 协议工具类 --------------------------
+class ProtocolCodec:
+    @staticmethod
+    def pack(
+            msg_type: MessageType,
+            body: Union[bytes, str, Dict[str, Any], List[Any]],
+            serialization: Optional[SerializationType] = None,
+            compression: CompressionType = CompressionType.NONE
+    ) -> bytes:
+        """
+        封装协议包（头部 + 包体）
+        :param msg_type: 消息类型
+        :param body: 包体数据（bytes/str/dict/list）
+        :param serialization: 序列化方式（None时自动推导）
+        :param compression: 压缩方式
+        :return: 完整协议包（bytes）
+        """
+        # 1. 自动推导序列化方式（优化：纯文本消息自动用STRING，复杂结构用JSON）🔴修改推导逻辑
+        if serialization is None:
+            if msg_type == MessageType.AUDIO_DATA:
+                serialization = SerializationType.RAW
+            elif msg_type == MessageType.TEXT_MESSAGE:  # 纯文本消息→STRING
+                serialization = SerializationType.STRING
+            elif msg_type in (MessageType.CONTROL_CMD, MessageType.AUDIO_TEXT_MIX):  # 复杂结构→JSON
+                serialization = SerializationType.JSON
+            else:
+                raise ValueError(f"不支持的消息类型：{msg_type}")
+
+        # 2. 序列化包体（新增STRING类型处理）🔴新增分支
+        serialized_body: bytes
+        if serialization == SerializationType.RAW:
+            if not isinstance(body, bytes):
+                raise TypeError("RAW序列化要求body必须是bytes类型")
+            serialized_body = body
+        elif serialization == SerializationType.STRING:
+            if not isinstance(body, str):
+                raise TypeError("STRING序列化要求body必须是str类型")
+            serialized_body = body.encode(ProtocolConst.STRING_ENCODING)  # 直接UTF-8编码，无JSON包装
+        elif serialization == SerializationType.JSON:
+            if isinstance(body, str):
+                serialized_body = body.encode(ProtocolConst.STRING_ENCODING)
+            elif isinstance(body, (dict, list)):
+                serialized_body = json.dumps(body, ensure_ascii=False, separators=(',', ':')).encode(ProtocolConst.STRING_ENCODING)
+            else:
+                raise TypeError("JSON序列化要求body必须是str/dict/list类型")
+        else:
+            raise ValueError(f"不支持的序列化方式：{serialization}")
+
+        # 3. 压缩包体（不变）
+        compressed_body: bytes
+        if compression == CompressionType.GZIP:
+            compressed_body = gzip.compress(serialized_body)
+        elif compression == CompressionType.NONE:
+            compressed_body = serialized_body
+        else:
+            raise ValueError(f"不支持的压缩方式：{compression}")
+
+        # 4. 校验包体大小（不变）
+        body_len = len(compressed_body)
+        if body_len > ProtocolConst.MAX_BODY_SIZE:
+            raise OverflowError(f"包体过大（{body_len}字节），最大支持{ProtocolConst.MAX_BODY_SIZE}字节")
+
+        # 5. 构造头部（不变）
+        byte1 = (msg_type.value << 4) | 0x00  # 消息类型(4位) + 保留位1(4位)
+        byte2 = (serialization.value << 4) | (compression.value & 0x0F)  # 序列化(4位) + 压缩(4位)
+        byte3_4 = struct.pack(">H", ProtocolConst.PROTOCOL_VERSION)  # 协议版本(16位)
+        byte5_8 = struct.pack(">I", body_len)  # 包体长度(32位)
+        header = bytes([byte1, byte2]) + byte3_4 + byte5_8
+        assert len(header) == ProtocolConst.HEADER_SIZE, f"头部长度错误：实际{len(header)}字节，预期{ProtocolConst.HEADER_SIZE}字节"
+
+        return header + compressed_body
+
+    @staticmethod
+    def unpack(packet: bytes) -> tuple[MessageType, SerializationType, CompressionType, Any]:
+        """
+        解析协议包
+        :param packet: 完整协议包（头部 + 包体）
+        :return: (消息类型, 序列化方式, 压缩方式, 原始包体数据)
+        """
+        # 1. 校验包长度（不变）
+        if len(packet) < ProtocolConst.HEADER_SIZE:
+            raise ValueError(f"包长度过短（{len(packet)}字节），至少需要{ProtocolConst.HEADER_SIZE}字节头部")
+
+        # 2. 解析头部（不变）
+        header = packet[:ProtocolConst.HEADER_SIZE]
+        body = packet[ProtocolConst.HEADER_SIZE:]
+        byte1 = header[0]
+        byte2 = header[1]
+
+        msg_type = MessageType((byte1 >> 4) & 0x0F)
+        serialization = SerializationType((byte2 >> 4) & 0x0F)
+        compression = CompressionType(byte2 & 0x0F)
+        version = struct.unpack(">H", header[2:4])[0]
+        body_len = struct.unpack(">I", header[4:8])[0]
+
+        if version != ProtocolConst.PROTOCOL_VERSION:
+            raise ValueError(f"协议版本不匹配：收到v{version}，支持v{ProtocolConst.PROTOCOL_VERSION}")
+        if len(body) != body_len:
+            raise ValueError(f"包体长度不匹配：头部声明{body_len}字节，实际{len(body)}字节")
+
+        # 3. 解压包体（不变）
+        decompressed_body: bytes
+        if compression == CompressionType.GZIP:
+            try:
+                decompressed_body = gzip.decompress(body)
+            except Exception as e:
+                raise ValueError(f"GZIP解压失败：{str(e)}")
+        elif compression == CompressionType.NONE:
+            decompressed_body = body
+        else:
+            raise ValueError(f"不支持的压缩方式：{compression}")
+
+        # 4. 反序列化包体（新增STRING类型处理）🔴新增分支
+        original_body: Any
+        if serialization == SerializationType.RAW:
+            original_body = decompressed_body
+        elif serialization == SerializationType.STRING:
+            try:
+                original_body = decompressed_body.decode(ProtocolConst.STRING_ENCODING)  # 直接UTF-8解码，无JSON解析
+            except UnicodeDecodeError:
+                raise ValueError(f"STRING反序列化失败：{ProtocolConst.STRING_ENCODING}解码错误")
+        elif serialization == SerializationType.JSON:
+            try:
+                original_body = json.loads(decompressed_body.decode(ProtocolConst.STRING_ENCODING))
+            except UnicodeDecodeError:
+                raise ValueError(f"JSON反序列化失败：{ProtocolConst.STRING_ENCODING}解码错误")
+            except json.JSONDecodeError:
+                raise ValueError("JSON反序列化失败：格式错误")
+        else:
+            raise ValueError(f"不支持的序列化方式：{serialization}")
+
+        return msg_type, serialization, compression, original_body
+
+
+# -------------------------- 使用示例（新增纯字符串消息测试） --------------------------
+if __name__ == "__main__":
+    # 示例1：纯文本消息（自动用STRING序列化，无JSON包装）🔴测试新增类型
+    text_body = "你好，这是纯字符串消息（无JSON）"
+    text_packet = ProtocolCodec.pack(
+        msg_type=MessageType.TEXT_MESSAGE,
+        body=text_body,
+        compression=CompressionType.NONE
+    )
+    print(f"纯文本消息包长度：{len(text_packet)}字节")
+    msg_type1, ser1, comp1, body1 = ProtocolCodec.unpack(text_packet)
+    print(f"解析结果：类型={msg_type1.name}，序列化={ser1.name}，压缩={comp1.name}，内容={body1}\n")
+
+    # 示例2：控制指令（JSON序列化，复杂结构）
+    control_body = {"cmd": ControlCommand.HEARTBEAT.value, "timestamp": 1699999999}
+    control_packet = ProtocolCodec.pack(
+        msg_type=MessageType.CONTROL_CMD,
+        body=control_body,
+        compression=CompressionType.NONE
+    )
+    print(f"控制指令包长度：{len(control_packet)}字节")
+    msg_type2, ser2, comp2, body2 = ProtocolCodec.unpack(control_packet)
+    print(f"解析结果：类型={msg_type2.name}，序列化={ser2.name}，压缩={comp2.name}，内容={body2}\n")
+
+    # 示例3：音频数据（RAW序列化）
+    audio_body = b"\x00\x01\x02\x03\x04\x05" * 100  # 模拟PCM数据
+    audio_packet = ProtocolCodec.pack(
+        msg_type=MessageType.AUDIO_DATA,
+        body=audio_body,
+        compression=CompressionType.NONE
+    )
+    print(f"音频数据包长度：{len(audio_packet)}字节")
+    msg_type3, ser3, comp3, body3 = ProtocolCodec.unpack(audio_packet)
+    print(f"解析结果：类型={msg_type3.name}，序列化={ser3.name}，压缩={comp3.name}，数据长度={len(body3)}字节\n")
+
+    # 示例4：混合数据（JSON序列化）
+    mix_body = {
+        "audio_data": list(audio_body[:10]),
+        "text_content": "混合数据仍用JSON"
+    }
+    mix_packet = ProtocolCodec.pack(
+        msg_type=MessageType.AUDIO_TEXT_MIX,
+        body=mix_body
+    )
+    print(f"混合数据包长度：{len(mix_packet)}字节")
+    msg_type4, ser4, comp4, body4 = ProtocolCodec.unpack(mix_packet)
+    print(f"解析结果：类型={msg_type4.name}，序列化={ser4.name}，压缩={comp4.name}，内容={body4}")

python/frontend_ws.py

@@ -66,57 +66,52 @@ async def frontend_websocket_handler(websocket: WebSocket):
     # ====================== 修复 TTS 核心逻辑 ======================
     # 1. 创建 TTS 客户端
     # 1. 创建TTS管理器
-    tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
-
+    # tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
+    tts_manager = TTSManager()
     # 2. 设置结果回调函数（接收完整结果）
     def handle_tts_result(req_id: str, result: Dict[str, Any]):
         """处理TTS结果回调"""
-        print('处理TTS结果回调 ',result)
+        # print('处理TTS结果回调 ',result)
         status = result.get("status")
 
         if status == "completed":
             audio_data = result.get("audio_data")
             sample_rate = result.get("sample_rate")
-
-            if audio_data is not None and len(audio_data) > 0:
-                # 转换为PCM数据
-                pcm_data = (audio_data.astype(np.float32) * 32767).astype(np.int16)
-                pcm_bytes = pcm_data.tobytes()
-                # result_queue.put_nowait({"type": 3, "data": pcm_bytes})
-                result_queue.put_nowait(pcm_bytes)
-                print(f"插入时候队列当前大小xxx: {result_queue.qsize()}")  # 排查队列是否有数据
+            result_queue.put_nowait(audio_data)
+            #
+            # if audio_data is not None and len(audio_data) > 0:
+            #     # 转换为PCM数据
+            #     pcm_data = (audio_data.astype(np.float32) * 32767).astype(np.int16)
+            #     pcm_bytes = pcm_data.tobytes()
+            #     # result_queue.put_nowait({"type": 3, "data": pcm_bytes})
+            #     result_queue.put_nowait(pcm_bytes)
+                # print(f"插入时候队列当前大小xxx: {result_queue.qsize()}")  # 排查队列是否有数据
                 # 保存为PCM文件
 
 
-
-        elif status == "error":
-            error_msg = result.get("message")
-            print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
+        # elif status == "error":
+        #     error_msg = result.get("message")
+        #     print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
 
     tts_manager.set_result_callback(handle_tts_result)
 
     # 3. 设置是否播放（可选，默认True）
-    tts_manager.set_playback_enabled(True)  # 设置为False则不播放
+    tts_manager.set_playback_enabled(False)  # 设置为False则不播放
 
     # 4. 初始化连接
     await tts_manager.initialize()
-    texts = [
-        "你好，这是第一个排队的TTS请求。",
-        "我是第二个请求，会等第一个处理完再执行。",
-        "第三个请求，支持流式播放和队列管理。",
-        "第四个请求，测试队列的自动消费功能。",
-        "最后一个请求，处理完成后会自动结束。"
-    ]
 
-    req_ids = []
-    for i, text in enumerate(texts):
-        req_id = await tts_manager.synthesize(
-            text,
-            mode="预训练音色",
-            sft_spk="中文女",
-            speed=1.0
-        )
-        req_ids.append(req_id)
+    # texts = [
+    #     "你好，这是第一个排队的TTS请求。你好，这是第一个排队的TTS请求。你好，这是第一个排队的TTS请好",
+    #     "我是第二个请求，会等第一个处理完再执行。",
+    #     "第三个请求，支持流式播放和队列管理。",
+    #     "第四个请求，测试队列的自动消费功能。",
+    #     "最后一个请求，处理完成后会自动结束。"
+    # ]
+    # req_ids = []
+    # for i, text in enumerate(texts):
+    #     req_id = await tts_manager.synthesize(text)
+    #     req_ids.append(req_id)
     # ====================== ASR 结果回调 ======================
     async def asr_result_callback(result: dict):
         """ASR 结果回调：转发前端 + 调用大模型"""
@@ -151,6 +146,13 @@ async def frontend_websocket_handler(websocket: WebSocket):
         """大模型流式回调（纯异步，无阻塞）"""
         if not chunk:
             return
+        # req_id = await tts_manager.synthesize(
+        #     chunk,
+        #     mode="预训练音色",
+        #     sft_spk="中文女",
+        #     speed=1.0
+        # )
+        req_id = await tts_manager.synthesize(chunk)
         # 1. 异步插入队列（替代put_nowait，避免队列满时抛异常）
         # 1. 非阻塞插入（队列满则丢弃，优先保证实时性）
         # result_queue.put_nowait(chunk)

python/llm_client.py

@@ -79,6 +79,7 @@ class LLMClient:
                     response.raise_for_status()
                     # 实时迭代流式响应
                     async for line in response.content.iter_chunked(LLM_CONFIG["stream_chunk_size"]):
+
                         if not line:
                             continue
                         line_data = line.decode("utf-8")
@@ -95,13 +96,14 @@ class LLMClient:
                                     res_conversation_id = data["conversation_id"]
                                 # 提取内容
                                 chunk = data.get("content", data.get("answer", data.get("message", "")))
+                                # print('大模型返回的', chunk)
                                 if chunk:
                                     full_response += chunk
                                     if stream_callback:
                                         await stream_callback(chunk, res_conversation_id, False)
                                         await asyncio.sleep(0)  # 让出调度权
                             except json.JSONDecodeError as e:
-                                print(f"解析流式数据失败: {e}")
+                                # print(f"大模型解析流式数据失败: {e}")
                                 continue
             else:
                 # 异步非流式请求

python/protocols/__init__.py

@@ -0,0 +1,41 @@
+from .protocols import (
+    CompressionBits,
+    EventType,
+    HeaderSizeBits,
+    Message,
+    MsgType,
+    MsgTypeFlagBits,
+    SerializationBits,
+    VersionBits,
+    audio_only_client,
+    cancel_session,
+    finish_connection,
+    finish_session,
+    full_client_request,
+    receive_message,
+    start_connection,
+    start_session,
+    task_request,
+    wait_for_event,
+)
+
+__all__ = [
+    "CompressionBits",
+    "EventType",
+    "HeaderSizeBits",
+    "Message",
+    "MsgType",
+    "MsgTypeFlagBits",
+    "SerializationBits",
+    "VersionBits",
+    "audio_only_client",
+    "cancel_session",
+    "finish_connection",
+    "finish_session",
+    "full_client_request",
+    "receive_message",
+    "start_connection",
+    "start_session",
+    "task_request",
+    "wait_for_event",
+]

python/protocols/protocols.py

@@ -0,0 +1,543 @@
+import io
+import logging
+import struct
+from dataclasses import dataclass
+from enum import IntEnum
+from typing import Callable, List
+
+import websockets
+
+logger = logging.getLogger(__name__)
+
+
+class MsgType(IntEnum):
+    """Message type enumeration"""
+
+    Invalid = 0
+    FullClientRequest = 0b1
+    AudioOnlyClient = 0b10
+    FullServerResponse = 0b1001
+    AudioOnlyServer = 0b1011
+    FrontEndResultServer = 0b1100
+    Error = 0b1111
+
+    # Alias
+    ServerACK = AudioOnlyServer
+
+    def __str__(self) -> str:
+        return self.name if self.name else f"MsgType({self.value})"
+
+
+class MsgTypeFlagBits(IntEnum):
+    """Message type flag bits"""
+
+    NoSeq = 0  # Non-terminal packet with no sequence
+    PositiveSeq = 0b1  # Non-terminal packet with sequence > 0
+    LastNoSeq = 0b10  # Last packet with no sequence
+    NegativeSeq = 0b11  # Last packet with sequence < 0
+    WithEvent = 0b100  # Payload contains event number (int32)
+
+
+class VersionBits(IntEnum):
+    """Version bits"""
+
+    Version1 = 1
+    Version2 = 2
+    Version3 = 3
+    Version4 = 4
+
+
+class HeaderSizeBits(IntEnum):
+    """Header size bits"""
+
+    HeaderSize4 = 1
+    HeaderSize8 = 2
+    HeaderSize12 = 3
+    HeaderSize16 = 4
+
+
+class SerializationBits(IntEnum):
+    """Serialization method bits"""
+
+    Raw = 0
+    JSON = 0b1
+    Thrift = 0b11
+    Custom = 0b1111
+
+
+class CompressionBits(IntEnum):
+    """Compression method bits"""
+
+    None_ = 0
+    Gzip = 0b1
+    Custom = 0b1111
+
+
+class EventType(IntEnum):
+    """Event type enumeration"""
+
+    None_ = 0  # Default event
+
+    # 1 ~ 49 Upstream Connection events
+    StartConnection = 1
+    StartTask = 1  # Alias of StartConnection
+    FinishConnection = 2
+    FinishTask = 2  # Alias of FinishConnection
+
+    # 50 ~ 99 Downstream Connection events
+    ConnectionStarted = 50  # Connection established successfully
+    TaskStarted = 50  # Alias of ConnectionStarted
+    ConnectionFailed = 51  # Connection failed (possibly due to authentication failure)
+    TaskFailed = 51  # Alias of ConnectionFailed
+    ConnectionFinished = 52  # Connection ended
+    TaskFinished = 52  # Alias of ConnectionFinished
+
+    # 100 ~ 149 Upstream Session events
+    StartSession = 100
+    CancelSession = 101
+    FinishSession = 102
+
+    # 150 ~ 199 Downstream Session events
+    SessionStarted = 150
+    SessionCanceled = 151
+    SessionFinished = 152
+    SessionFailed = 153
+    UsageResponse = 154  # Usage response
+    ChargeData = 154  # Alias of UsageResponse
+
+    # 200 ~ 249 Upstream general events
+    TaskRequest = 200
+    UpdateConfig = 201
+
+    # 250 ~ 299 Downstream general events
+    AudioMuted = 250
+
+    # 300 ~ 349 Upstream TTS events
+    SayHello = 300
+
+    # 350 ~ 399 Downstream TTS events
+    TTSSentenceStart = 350
+    TTSSentenceEnd = 351
+    TTSResponse = 352
+    TTSEnded = 359
+    PodcastRoundStart = 360
+    PodcastRoundResponse = 361
+    PodcastRoundEnd = 362
+
+    # 450 ~ 499 Downstream ASR events
+    ASRInfo = 450
+    ASRResponse = 451
+    ASREnded = 459
+
+    # 500 ~ 549 Upstream dialogue events
+    ChatTTSText = 500  # (Ground-Truth-Alignment) text for speech synthesis
+
+    # 550 ~ 599 Downstream dialogue events
+    ChatResponse = 550
+    ChatEnded = 559
+
+    # 650 ~ 699 Downstream dialogue events
+    # Events for source (original) language subtitle
+    SourceSubtitleStart = 650
+    SourceSubtitleResponse = 651
+    SourceSubtitleEnd = 652
+    # Events for target (translation) language subtitle
+    TranslationSubtitleStart = 653
+    TranslationSubtitleResponse = 654
+    TranslationSubtitleEnd = 655
+
+    def __str__(self) -> str:
+        return self.name if self.name else f"EventType({self.value})"
+
+
+@dataclass
+class Message:
+    """Message object
+
+    Message format:
+    0                 1                 2                 3
+    | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |    Version      |   Header Size   |     Msg Type    |      Flags      |
+    |   (4 bits)      |    (4 bits)     |     (4 bits)    |     (4 bits)    |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    | Serialization   |   Compression   |           Reserved                |
+    |   (4 bits)      |    (4 bits)     |           (8 bits)                |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                                       |
+    |                   Optional Header Extensions                          |
+    |                     (if Header Size > 1)                              |
+    |                                                                       |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                                       |
+    |                           Payload                                     |
+    |                      (variable length)                                |
+    |                                                                       |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    """
+
+    version: VersionBits = VersionBits.Version1
+    header_size: HeaderSizeBits = HeaderSizeBits.HeaderSize4
+    type: MsgType = MsgType.Invalid
+    flag: MsgTypeFlagBits = MsgTypeFlagBits.NoSeq
+    serialization: SerializationBits = SerializationBits.JSON
+    compression: CompressionBits = CompressionBits.None_
+
+    event: EventType = EventType.None_
+    session_id: str = ""
+    connect_id: str = ""
+    sequence: int = 0
+    error_code: int = 0
+
+    payload: bytes = b""
+
+    @classmethod
+    def from_bytes(cls, data: bytes) -> "Message":
+        """Create message object from bytes"""
+        if len(data) < 3:
+            raise ValueError(
+                f"Data too short: expected at least 3 bytes, got {len(data)}"
+            )
+
+        type_and_flag = data[1]
+        msg_type = MsgType(type_and_flag >> 4)
+        flag = MsgTypeFlagBits(type_and_flag & 0b00001111)
+
+        msg = cls(type=msg_type, flag=flag)
+        msg.unmarshal(data)
+        return msg
+
+    def marshal(self) -> bytes:
+        """Serialize message to bytes"""
+        buffer = io.BytesIO()
+
+        # Write header
+        header = [
+            (self.version << 4) | self.header_size,
+            (self.type << 4) | self.flag,
+            (self.serialization << 4) | self.compression,
+        ]
+
+        header_size = 4 * self.header_size
+        if padding := header_size - len(header):
+            header.extend([0] * padding)
+
+        buffer.write(bytes(header))
+
+        # Write other fields
+        writers = self._get_writers()
+        for writer in writers:
+            writer(buffer)
+
+        return buffer.getvalue()
+
+    def unmarshal(self, data: bytes) -> None:
+        """Deserialize message from bytes"""
+        buffer = io.BytesIO(data)
+
+        # Read version and header size
+        version_and_header_size = buffer.read(1)[0]
+        self.version = VersionBits(version_and_header_size >> 4)
+        self.header_size = HeaderSizeBits(version_and_header_size & 0b00001111)
+
+        # Skip second byte
+        buffer.read(1)
+
+        # Read serialization and compression methods
+        serialization_compression = buffer.read(1)[0]
+        self.serialization = SerializationBits(serialization_compression >> 4)
+        self.compression = CompressionBits(serialization_compression & 0b00001111)
+
+        # Skip header padding
+        header_size = 4 * self.header_size
+        read_size = 3
+        if padding_size := header_size - read_size:
+            buffer.read(padding_size)
+
+        # Read other fields
+        readers = self._get_readers()
+        for reader in readers:
+            reader(buffer)
+
+        # Check for remaining data
+        remaining = buffer.read()
+        if remaining:
+            raise ValueError(f"Unexpected data after message: {remaining}")
+
+    def _get_writers(self) -> List[Callable[[io.BytesIO], None]]:
+        """Get list of writer functions"""
+        writers = []
+
+        if self.flag == MsgTypeFlagBits.WithEvent:
+            writers.extend([self._write_event, self._write_session_id])
+
+        if self.type in [
+            MsgType.FullClientRequest,
+            MsgType.FullServerResponse,
+            MsgType.FrontEndResultServer,
+            MsgType.AudioOnlyClient,
+            MsgType.AudioOnlyServer,
+        ]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                writers.append(self._write_sequence)
+        elif self.type == MsgType.Error:
+            writers.append(self._write_error_code)
+        else:
+            raise ValueError(f"Unsupported message type: {self.type}")
+
+        writers.append(self._write_payload)
+        return writers
+
+    def _get_readers(self) -> List[Callable[[io.BytesIO], None]]:
+        """Get list of reader functions"""
+        readers = []
+
+        if self.type in [
+            MsgType.FullClientRequest,
+            MsgType.FullServerResponse,
+            MsgType.FrontEndResultServer,
+            MsgType.AudioOnlyClient,
+            MsgType.AudioOnlyServer,
+        ]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                readers.append(self._read_sequence)
+        elif self.type == MsgType.Error:
+            readers.append(self._read_error_code)
+        else:
+            raise ValueError(f"Unsupported message type: {self.type}")
+
+        if self.flag == MsgTypeFlagBits.WithEvent:
+            readers.extend(
+                [self._read_event, self._read_session_id, self._read_connect_id]
+            )
+
+        readers.append(self._read_payload)
+        return readers
+
+    def _write_event(self, buffer: io.BytesIO) -> None:
+        """Write event"""
+        buffer.write(struct.pack(">i", self.event))
+
+    def _write_session_id(self, buffer: io.BytesIO) -> None:
+        """Write session ID"""
+        if self.event in [
+            EventType.StartConnection,
+            EventType.FinishConnection,
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+        ]:
+            return
+
+        session_id_bytes = self.session_id.encode("utf-8")
+        size = len(session_id_bytes)
+        if size > 0xFFFFFFFF:
+            raise ValueError(f"Session ID size ({size}) exceeds max(uint32)")
+
+        buffer.write(struct.pack(">I", size))
+        if size > 0:
+            buffer.write(session_id_bytes)
+
+    def _write_sequence(self, buffer: io.BytesIO) -> None:
+        """Write sequence number"""
+        buffer.write(struct.pack(">i", self.sequence))
+
+    def _write_error_code(self, buffer: io.BytesIO) -> None:
+        """Write error code"""
+        buffer.write(struct.pack(">I", self.error_code))
+
+    def _write_payload(self, buffer: io.BytesIO) -> None:
+        """Write payload"""
+        size = len(self.payload)
+        if size > 0xFFFFFFFF:
+            raise ValueError(f"Payload size ({size}) exceeds max(uint32)")
+
+        buffer.write(struct.pack(">I", size))
+        buffer.write(self.payload)
+
+    def _read_event(self, buffer: io.BytesIO) -> None:
+        """Read event"""
+        event_bytes = buffer.read(4)
+        if event_bytes:
+            self.event = EventType(struct.unpack(">i", event_bytes)[0])
+
+    def _read_session_id(self, buffer: io.BytesIO) -> None:
+        """Read session ID"""
+        if self.event in [
+            EventType.StartConnection,
+            EventType.FinishConnection,
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+            EventType.ConnectionFinished,
+        ]:
+            return
+
+        size_bytes = buffer.read(4)
+        if size_bytes:
+            size = struct.unpack(">I", size_bytes)[0]
+            if size > 0:
+                session_id_bytes = buffer.read(size)
+                if len(session_id_bytes) == size:
+                    self.session_id = session_id_bytes.decode("utf-8")
+
+    def _read_connect_id(self, buffer: io.BytesIO) -> None:
+        """Read connection ID"""
+        if self.event in [
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+            EventType.ConnectionFinished,
+        ]:
+            size_bytes = buffer.read(4)
+            if size_bytes:
+                size = struct.unpack(">I", size_bytes)[0]
+                if size > 0:
+                    self.connect_id = buffer.read(size).decode("utf-8")
+
+    def _read_sequence(self, buffer: io.BytesIO) -> None:
+        """Read sequence number"""
+        sequence_bytes = buffer.read(4)
+        if sequence_bytes:
+            self.sequence = struct.unpack(">i", sequence_bytes)[0]
+
+    def _read_error_code(self, buffer: io.BytesIO) -> None:
+        """Read error code"""
+        error_code_bytes = buffer.read(4)
+        if error_code_bytes:
+            self.error_code = struct.unpack(">I", error_code_bytes)[0]
+
+    def _read_payload(self, buffer: io.BytesIO) -> None:
+        """Read payload"""
+        size_bytes = buffer.read(4)
+        if size_bytes:
+            size = struct.unpack(">I", size_bytes)[0]
+            if size > 0:
+                self.payload = buffer.read(size)
+
+    def __str__(self) -> str:
+        """String representation"""
+        if self.type in [MsgType.AudioOnlyServer, MsgType.AudioOnlyClient]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                return f"MsgType: {self.type}, EventType:{self.event}, Sequence: {self.sequence}, PayloadSize: {len(self.payload)}"
+            return f"MsgType: {self.type}, EventType:{self.event}, PayloadSize: {len(self.payload)}"
+        elif self.type == MsgType.Error:
+            return f"MsgType: {self.type}, EventType:{self.event}, ErrorCode: {self.error_code}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+        else:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                return f"MsgType: {self.type}, EventType:{self.event}, Sequence: {self.sequence}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+            return f"MsgType: {self.type}, EventType:{self.event}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+
+
+async def receive_message(websocket: websockets.WebSocketClientProtocol) -> Message:
+    """Receive message from websocket"""
+    try:
+        data = await websocket.recv()
+        if isinstance(data, str):
+            raise ValueError(f"Unexpected text message: {data}")
+        elif isinstance(data, bytes):
+            msg = Message.from_bytes(data)
+            logger.info(f"Received: {msg}")
+            return msg
+        else:
+            raise ValueError(f"Unexpected message type: {type(data)}")
+    except Exception as e:
+        logger.error(f"Failed to receive message: {e}")
+        raise
+
+
+async def wait_for_event(
+    websocket: websockets.WebSocketClientProtocol,
+    msg_type: MsgType,
+    event_type: EventType,
+) -> Message:
+    """Wait for specific event"""
+    while True:
+        msg = await receive_message(websocket)
+        if msg.type != msg_type or msg.event != event_type:
+            raise ValueError(f"Unexpected message: {msg}")
+        if msg.type == msg_type and msg.event == event_type:
+            return msg
+
+
+async def full_client_request(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes
+) -> None:
+    """Send full client message"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.NoSeq)
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def audio_only_client(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, flag: MsgTypeFlagBits
+) -> None:
+    """Send audio-only client message"""
+    msg = Message(type=MsgType.AudioOnlyClient, flag=flag)
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def start_connection(websocket: websockets.WebSocketClientProtocol) -> None:
+    """Start connection"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.StartConnection
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def finish_connection(websocket: websockets.WebSocketClientProtocol) -> None:
+    """Finish connection"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.FinishConnection
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def start_session(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, session_id: str
+) -> None:
+    """Start session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.StartSession
+    msg.session_id = session_id
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def finish_session(
+    websocket: websockets.WebSocketClientProtocol, session_id: str
+) -> None:
+    """Finish session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.FinishSession
+    msg.session_id = session_id
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def cancel_session(
+    websocket: websockets.WebSocketClientProtocol, session_id: str
+) -> None:
+    """Cancel session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.CancelSession
+    msg.session_id = session_id
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def task_request(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, session_id: str
+) -> None:
+    """Send task request"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.TaskRequest
+    msg.session_id = session_id
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())

python/tts_client-本地tts.py

@@ -0,0 +1,551 @@
+import asyncio
+import json
+import websockets
+import numpy as np
+import sounddevice as sd
+from typing import Optional, Callable, Dict, Any, Coroutine, List
+from dataclasses import dataclass, field
+import uuid
+
+
+@dataclass
+class TTSRequest:
+    """TTS请求对象（带唯一标识）"""
+    tts_text: str
+    mode: str = "预训练音色"
+    sft_spk: str = ""
+    seed: int = field(default_factory=lambda: np.random.randint(1, 100000000))
+    stream: bool = True
+    speed: float = 1.0
+    prompt_wav: str = ""
+    instruct_text: str = ""
+    request_id: str = field(default_factory=lambda: str(uuid.uuid4()))  # 唯一请求ID
+
+
+class CosyVoiceTTSSocketClient:
+    """CosyVoice TTS WebSocket 客户端（异步/流式/带任务队列）"""
+
+    def __init__(self, ws_url: str = "ws://localhost:50000/ws/tts", max_queue_size: int = 100):
+        """
+        初始化客户端
+        :param ws_url: WebSocket 服务端地址
+        :param max_queue_size: 最大队列长度（防止内存溢出）
+        """
+        self.ws_url = ws_url
+        self.websocket: Optional[websockets.WebSocketClientProtocol] = None
+        self.is_connected = False
+        self.is_processing = False  # 是否正在处理请求
+
+        # 异步任务队列（FIFO）
+        self.request_queue: asyncio.Queue[TTSRequest] = asyncio.Queue(maxsize=max_queue_size)
+
+        # 回调函数定义（所有回调都带request_id，方便关联请求）
+        self.on_task_enqueue: Callable[[str], None] = lambda req_id: None  # 任务入队回调
+        self.on_start: Callable[[str, Dict[str, Any]], None] = lambda req_id, data: None  # 合成开始回调
+        self.on_audio_chunk: Callable[[str, np.ndarray], None] = lambda req_id, chunk: None  # 音频块回调
+        self.on_end: Callable[[str, Dict[str, Any]], None] = lambda req_id, data: None  # 合成结束回调
+        self.on_error: Callable[[str, str], None] = lambda req_id, msg: None  # 错误回调
+        self.on_queue_full: Callable[[str], None] = lambda req_id: None  # 队列满回调
+
+        # 音频播放流
+        self.play_stream: Optional[sd.OutputStream] = None
+        self.current_req_id: Optional[str] = None
+
+        # 是否启用播放功能
+        self.enable_playback: bool = True
+
+        # 外部回调函数，用于返回完整结果
+        self.external_callback: Optional[Callable[[str, Dict[str, Any]], None]] = None
+
+        # 存储每个请求的音频数据
+        self.audio_buffers: Dict[str, List[np.ndarray]] = {}
+
+    async def connect(self):
+        """建立 WebSocket 连接（初始化一次）"""
+        if not self.is_connected:
+            try:
+                self.websocket = await websockets.connect(self.ws_url)
+                self.is_connected = True
+                print(f"成功连接到 TTS 服务端: {self.ws_url}")
+
+                # 启动队列消费协程（后台运行）
+                asyncio.create_task(self._consume_queue())
+            except Exception as e:
+                raise ConnectionError(f"连接失败: {str(e)}")
+
+    async def disconnect(self):
+        """关闭 WebSocket 连接"""
+        if self.is_connected and self.websocket:
+            await self.websocket.close()
+            self.is_connected = False
+            self.websocket = None
+            print("已断开与 TTS 服务端的连接")
+
+        if self.play_stream:
+            self.play_stream.stop()
+            self.play_stream.close()
+            self.play_stream = None
+
+    def set_external_callback(self, callback: Callable[[str, Dict[str, Any]], None]):
+        """设置外部回调函数，用于返回完整结果"""
+        self.external_callback = callback
+
+    def set_playback_enabled(self, enabled: bool):
+        """设置是否启用音频播放"""
+        self.enable_playback = enabled
+
+    async def add_tts_request(self, tts_text: str, **kwargs) -> str:
+        """
+        添加TTS请求到队列（异步非阻塞）
+        :param tts_text: 要合成的文本
+        :param kwargs: 其他TTS参数
+        :return: 唯一请求ID
+        """
+        # 创建请求对象
+        request = TTSRequest(tts_text=tts_text, **kwargs)
+        req_id = request.request_id
+
+        # 初始化音频缓冲区
+        self.audio_buffers[req_id] = []
+
+        # 异步入队
+        try:
+            await self.request_queue.put(request)
+            self.on_task_enqueue(req_id)
+            print(f"请求 {req_id} 已加入队列，当前队列长度: {self.request_queue.qsize()}")
+            return req_id
+        except asyncio.QueueFull:
+            self.on_queue_full(req_id)
+            raise Exception(f"队列已满，请求 {req_id} 入队失败")
+
+    async def _consume_queue(self):
+        """消费队列（后台协程，自动处理排队请求）"""
+        print("队列消费协程已启动")
+        while True:
+            try:
+                # 等待队列中有请求（阻塞）
+                request = await self.request_queue.get()
+                req_id = request.request_id
+
+                # 标记为处理中
+                self.is_processing = True
+                print(f"开始处理请求 {req_id}，剩余队列长度: {self.request_queue.qsize()}")
+
+                # 处理当前请求
+                await self._process_single_request(request)
+
+                # 标记任务完成
+                self.request_queue.task_done()
+                self.is_processing = False
+
+            except Exception as e:
+                print(f"队列消费异常: {str(e)}")
+                self.is_processing = False
+                # 短暂等待后继续消费，避免死循环
+                await asyncio.sleep(0.1)
+
+    async def _process_single_request(self, request: TTSRequest):
+        """处理单个TTS请求"""
+        req_id = request.request_id
+
+        # 参数校验
+        if not request.tts_text:
+            self.on_error(req_id, "合成文本不能为空")
+            return
+
+        # 确保已连接
+        if not self.is_connected:
+            try:
+                await self.connect()
+            except Exception as e:
+                self.on_error(req_id, f"连接失败: {str(e)}")
+                return
+
+        # 构造请求数据
+        request_data = {
+            "tts_text": request.tts_text,
+            "mode": request.mode,
+            "sft_spk": request.sft_spk,
+            "seed": request.seed,
+            "stream": request.stream,
+            "speed": request.speed,
+            "prompt_wav": request.prompt_wav,
+            "instruct_text": request.instruct_text
+        }
+
+        try:
+            # 发送请求
+            await self.websocket.send(json.dumps(request_data))
+
+            # 处理返回的流式数据
+            await self._handle_response(req_id)
+
+        except websockets.exceptions.ConnectionClosed:
+            self.is_connected = False
+            self.on_error(req_id, "连接已关闭")
+        except Exception as e:
+            self.on_error(req_id, f"处理请求失败: {str(e)}")
+
+    async def _handle_response(self, req_id: str):
+        """处理单个请求的服务端响应"""
+        if not self.websocket:
+            return
+
+        try:
+            sample_rate = None
+            while True:
+                # 接收服务端消息（异步）
+                response = await self.websocket.recv()
+                data = json.loads(response)
+
+                # 根据状态分发到不同回调（都带req_id）
+                status = data.get("status")
+                if status == "start":
+                    # 合成开始 - 返回采样率等信息
+                    sample_rate = data.get("sample_rate")
+                    self.on_start(req_id, data)
+
+                elif status == "stream":
+                    # 流式音频块 - 转为 numpy 数组
+                    audio_chunk = np.array(data["audio_chunk"], dtype=np.float32)
+
+                    # 保存到缓冲区
+                    self.audio_buffers[req_id].append(audio_chunk)
+
+                    # 音频块回调
+                    self.on_audio_chunk(req_id, audio_chunk)
+
+                elif status == "end":
+                    # 合成结束
+                    self.on_end(req_id, data)
+
+                    # 组装完整音频数据
+                    full_audio = np.concatenate(self.audio_buffers[req_id]) if self.audio_buffers[req_id] else np.array(
+                        [], dtype=np.float32)
+
+                    # 准备返回给外部回调的数据
+                    result_data = {
+                        "status": "completed",
+                        "request_id": req_id,
+                        "sample_rate": sample_rate,
+                        "audio_data": full_audio,
+                        "audio_length": len(full_audio),
+                        "message": data.get("msg", "合成完成")
+                    }
+
+                    # 调用外部回调（非阻塞）
+                    if self.external_callback:
+                        # 使用create_task确保回调不会阻塞主流程
+                        asyncio.create_task(self._call_external_callback(req_id, result_data))
+
+                    # 清理缓冲区
+                    if req_id in self.audio_buffers:
+                        del self.audio_buffers[req_id]
+
+                    break
+
+                elif status == "error":
+                    # 错误处理
+                    error_msg = data.get("msg", "未知错误")
+                    self.on_error(req_id, error_msg)
+
+                    # 准备错误结果数据
+                    error_data = {
+                        "status": "error",
+                        "request_id": req_id,
+                        "message": error_msg
+                    }
+
+                    # 调用外部回调
+                    if self.external_callback:
+                        asyncio.create_task(self._call_external_callback(req_id, error_data))
+
+                    # 清理缓冲区
+                    if req_id in self.audio_buffers:
+                        del self.audio_buffers[req_id]
+
+                    break
+
+        except Exception as e:
+            error_msg = f"接收数据异常: {str(e)}"
+            self.on_error(req_id, error_msg)
+
+            # 准备错误结果数据
+            error_data = {
+                "status": "error",
+                "request_id": req_id,
+                "message": error_msg
+            }
+
+            # 调用外部回调
+            if self.external_callback:
+                asyncio.create_task(self._call_external_callback(req_id, error_data))
+
+            # 清理缓冲区
+            if req_id in self.audio_buffers:
+                del self.audio_buffers[req_id]
+
+    async def _call_external_callback(self, req_id: str, result_data: Dict[str, Any]):
+        """异步调用外部回调函数"""
+        try:
+            if self.external_callback:
+                # 确保回调是异步的，不会阻塞
+                if asyncio.iscoroutinefunction(self.external_callback):
+                    await self.external_callback(req_id, result_data)
+                else:
+                    # 如果是同步函数，在线程池中执行
+                    await asyncio.get_event_loop().run_in_executor(
+                        None, self.external_callback, req_id, result_data
+                    )
+        except Exception as e:
+            print(f"外部回调执行异常: {str(e)}")
+
+    def _play_audio_chunk(self, chunk: np.ndarray):
+        """播放音频块（非阻塞）"""
+        if not self.enable_playback or not self.play_stream:
+            return
+
+        try:
+            if chunk.size > 0:
+                self.play_stream.write(chunk)
+        except Exception as e:
+            print(f"音频播放异常: {str(e)}")
+
+    async def wait_queue_empty(self):
+        """等待队列所有任务处理完成（阻塞）"""
+        await self.request_queue.join()
+        print("所有队列任务已处理完成")
+
+
+# ------------------------------
+# 使用示例
+# ------------------------------
+class TTSManager:
+    """TTS管理器 - 供外部代码调用"""
+
+    def __init__(self, ws_url: str = "ws://localhost:50000/ws/tts"):
+        self.client = CosyVoiceTTSSocketClient(ws_url)
+        self._setup_callbacks()
+
+    def _setup_callbacks(self):
+        """设置内部回调"""
+
+        # 合成开始回调
+        def on_tts_start(req_id, data):
+            print(f"\n🎤 开始合成 [{req_id[:8]}] - 采样率: {data.get('sample_rate')}")
+
+            # 初始化播放流（如果需要播放）
+            if self.client.enable_playback:
+                sample_rate = data.get("sample_rate", 24000)
+                if self.client.play_stream:
+                    self.client.play_stream.stop()
+                    self.client.play_stream.close()
+
+                self.client.play_stream = sd.OutputStream(
+                    samplerate=sample_rate,
+                    channels=1,
+                    dtype=np.float32
+                )
+                self.client.play_stream.start()
+                self.client.current_req_id = req_id
+
+        # 音频块回调
+        def on_audio_chunk(req_id, chunk):
+            if req_id == self.client.current_req_id:
+                self.client._play_audio_chunk(chunk)
+
+        # 合成结束回调
+        def on_tts_end(req_id, data):
+            print(f"🏁 合成完成 [{req_id[:8]}] - {data.get('msg', '完成')}")
+
+        # 错误回调
+        def on_tts_error(req_id, msg):
+            print(f"❌ 合成失败 [{req_id[:8]}] - {msg}")
+
+        self.client.on_start = on_tts_start
+        self.client.on_audio_chunk = on_audio_chunk
+        self.client.on_end = on_tts_end
+        self.client.on_error = on_tts_error
+
+    async def initialize(self):
+        """初始化连接"""
+        await self.client.connect()
+
+    async def shutdown(self):
+        """关闭连接"""
+        await self.client.disconnect()
+
+    def set_result_callback(self, callback: Callable[[str, Dict[str, Any]], None]):
+        """设置结果回调函数"""
+        self.client.set_external_callback(callback)
+
+    def set_playback_enabled(self, enabled: bool):
+        """设置是否启用播放"""
+        self.client.set_playback_enabled(enabled)
+
+    async def synthesize(self, text: str, **kwargs) -> str:
+        """
+        异步合成文本（非阻塞）
+        :param text: 要合成的文本
+        :param kwargs: 其他参数（mode, sft_spk, speed等）
+        :return: 请求ID
+        """
+        return await self.client.add_tts_request(text, **kwargs)
+
+    async def wait_all_completed(self):
+        """等待所有任务完成"""
+        await self.client.wait_queue_empty()
+
+
+# ------------------------------
+# 外部调用示例
+# ------------------------------
+async def external_usage_example():
+    """外部代码使用示例"""
+
+    # 1. 创建TTS管理器
+    tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
+
+    # 2. 设置结果回调函数（接收完整结果）
+    def handle_tts_result1(req_id: str, result: Dict[str, Any]):
+        """处理TTS结果回调"""
+        status = result.get("status")
+
+        if status == "completed":
+            audio_data = result.get("audio_data")
+            sample_rate = result.get("sample_rate")
+            audio_length = result.get("audio_length")
+
+            print(f"✅ 收到TTS结果 [{req_id[:8]}]: 长度{audio_length}采样点, 采样率{sample_rate}Hz")
+
+            # 这里可以保存音频文件或进行其他处理
+            # 注意：audio_data是完整的numpy数组
+
+        elif status == "error":
+            error_msg = result.get("message")
+            print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
+
+    def handle_tts_result(req_id: str, result: Dict[str, Any]):
+        """处理TTS结果回调 - 保存为PCM文件"""
+        print('处理TTS结果回调', result)
+        status = result.get("status")
+        print('status')
+        if status == "completed":
+            audio_data = result.get("audio_data")
+            sample_rate = result.get("sample_rate")
+
+            if audio_data is not None and len(audio_data) > 0:
+                # 转换为PCM数据
+                pcm_data = (audio_data.astype(np.float32) * 32767).astype(np.int16)
+                pcm_bytes = pcm_data.tobytes()
+
+                # 保存为PCM文件
+                filename = f"tts_output_{req_id[:8]}.pcm"
+                with open(filename, "wb") as f:
+                    f.write(pcm_bytes)
+
+                print(f"💾 PCM文件已保存: {filename}, 大小: {len(pcm_bytes)}字节")
+
+                return filename
+
+        elif status == "error":
+            error_msg = result.get("message")
+            print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
+
+        return None
+    tts_manager.set_result_callback(handle_tts_result)
+
+    # 3. 设置是否播放（可选，默认True）
+    tts_manager.set_playback_enabled(True)  # 设置为False则不播放
+
+    # 4. 初始化连接
+    await tts_manager.initialize()
+
+    # 5. 异步合成多个文本（非阻塞）
+    texts = [
+        "你好，这是第一个排队的TTS请求。",
+        "我是第二个请求，会等第一个处理完再执行。",
+        "第三个请求，支持流式播放和队列管理。",
+        "第四个请求，测试队列的自动消费功能。",
+        "最后一个请求，处理完成后会自动结束。"
+    ]
+
+    req_ids = []
+    for i, text in enumerate(texts):
+        req_id = await tts_manager.synthesize(
+            text,
+            mode="预训练音色",
+            sft_spk="中文女",
+            speed=1.0
+        )
+        req_ids.append(req_id)
+        print(f"已提交请求 {i + 1}: ID={req_id[:8]}")
+
+        # 可以立即继续其他操作，不需要等待
+        await asyncio.sleep(0.5)  # 模拟其他操作
+
+    # 6. 可以在这里做其他事情，TTS会在后台处理
+
+    # 7. 等待所有TTS任务完成（可选）
+    await tts_manager.wait_all_completed()
+
+    # 8. 关闭连接
+    await tts_manager.shutdown()
+
+
+async def advanced_usage_example():
+    """高级使用示例：动态控制播放和结果处理"""
+
+    tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
+
+    # 自定义结果处理器
+    class ResultProcessor:
+        def __init__(self):
+            self.results = {}
+
+        async def process_result(self, req_id: str, result: Dict[str, Any]):
+            """异步处理结果"""
+            if result.get("status") == "completed":
+                # 保存音频数据等处理
+                self.results[req_id] = result
+                print(f"处理完成: {req_id[:8]}, 音频长度: {result.get('audio_length')}")
+
+                # 这里可以添加自定义逻辑，如保存到文件、发送到网络等
+
+    processor = ResultProcessor()
+    tts_manager.set_result_callback(processor.process_result)
+
+    # 禁用播放，只获取数据
+    tts_manager.set_playback_enabled(False)
+
+    await tts_manager.initialize()
+
+    # 批量合成
+    tasks = []
+    for i in range(3):
+        text = f"这是第{i+1}个测试文本，用于演示高级功能。"
+        task = tts_manager.synthesize(text)
+        tasks.append(task)
+
+        # 并行提交所有请求
+        req_ids = await asyncio.gather(*tasks)
+
+        # 等待完成
+        await tts_manager.wait_all_completed()
+        await tts_manager.shutdown()
+
+        print(f"所有任务完成，共处理 {len(processor.results)} 个结果")
+
+if __name__ == "__main__":
+    # 运行示例
+    try:
+        # 基础使用示例
+        asyncio.run(external_usage_example())
+
+        # 高级使用示例（取消注释运行）
+        # asyncio.run(advanced_usage_example())
+
+    except KeyboardInterrupt:
+        print("\n程序被用户中断")
+    except Exception as e:
+        print(f"程序异常: {str(e)}")

python/tts_client.py

@@ -3,38 +3,73 @@ import json
 import websockets
 import numpy as np
 import sounddevice as sd
-from typing import Optional, Callable, Dict, Any, Coroutine, List
+from typing import Optional, Callable, Dict, Any, List
 from dataclasses import dataclass, field
 import uuid
+import copy
+from protocols import (
+    EventType,
+    MsgType,
+    finish_connection,
+    finish_session,
+    receive_message,
+    start_connection,
+    start_session,
+    task_request,
+    wait_for_event,
+)
+
+# ------------------------------
+# 配置常量（可根据需求调整）
+# ------------------------------
+DEFAULT_APPID = "7069844318"
+DEFAULT_ACCESS_TOKEN = "osFMEJr20SSTWRql43cJlZkAOg7iwvxu"
+DEFAULT_ENDPOINT = "wss://openspeech.bytedance.com/api/v3/tts/bidirection"
+DEFAULT_VOICE_TYPE = "zh_female_gaolengyujie_emo_v2_mars_bigtts"
+DEFAULT_ENCODING = "pcm"
+DEFAULT_SAMPLE_RATE = 16000
 
 
 @dataclass
 class TTSRequest:
     """TTS请求对象（带唯一标识）"""
     tts_text: str
-    mode: str = "预训练音色"
-    sft_spk: str = ""
-    seed: int = field(default_factory=lambda: np.random.randint(1, 100000000))
-    stream: bool = True
-    speed: float = 1.0
-    prompt_wav: str = ""
-    instruct_text: str = ""
+    voice_type: str = DEFAULT_VOICE_TYPE
+    encoding: str = DEFAULT_ENCODING
+    speed: float = 1.0  # 语速（字节跳动TTS支持，需服务端兼容）
+    stream: bool = True  # 是否流式合成
     request_id: str = field(default_factory=lambda: str(uuid.uuid4()))  # 唯一请求ID
+    session_id: str = field(default_factory=lambda: str(uuid.uuid4()))  # 会话ID（每个请求一个会话）
 
 
-class CosyVoiceTTSSocketClient:
-    """CosyVoice TTS WebSocket 客户端（异步/流式/带任务队列）"""
+class ByteDanceTTSSocketClient:
+    """字节跳动 TTS WebSocket 客户端（异步/流式/带任务队列）"""
 
-    def __init__(self, ws_url: str = "ws://localhost:50000/ws/tts", max_queue_size: int = 100):
+    def __init__(
+        self,
+        appid: str = DEFAULT_APPID,
+        access_token: str = DEFAULT_ACCESS_TOKEN,
+        endpoint: str = DEFAULT_ENDPOINT,
+        max_queue_size: int = 100
+    ):
         """
         初始化客户端
-        :param ws_url: WebSocket 服务端地址
+        :param appid: 字节跳动APP ID
+        :param access_token: 访问令牌
+        :param endpoint: WebSocket 服务端地址
         :param max_queue_size: 最大队列长度（防止内存溢出）
         """
-        self.ws_url = ws_url
+        # 基础配置
+        self.appid = appid
+        self.access_token = access_token
+        self.endpoint = endpoint
+        self.max_queue_size = max_queue_size
+
+        # WebSocket 连接状态
         self.websocket: Optional[websockets.WebSocketClientProtocol] = None
         self.is_connected = False
         self.is_processing = False  # 是否正在处理请求
+        self.logid: Optional[str] = None  # 服务端返回的日志ID
 
         # 异步任务队列（FIFO）
         self.request_queue: asyncio.Queue[TTSRequest] = asyncio.Queue(maxsize=max_queue_size)
@@ -42,45 +77,84 @@ class CosyVoiceTTSSocketClient:
         # 回调函数定义（所有回调都带request_id，方便关联请求）
         self.on_task_enqueue: Callable[[str], None] = lambda req_id: None  # 任务入队回调
         self.on_start: Callable[[str, Dict[str, Any]], None] = lambda req_id, data: None  # 合成开始回调
-        self.on_audio_chunk: Callable[[str, np.ndarray], None] = lambda req_id, chunk: None  # 音频块回调
+        self.on_audio_chunk: Callable[[str, bytes], None] = lambda req_id, chunk: None  # 音频块回调（原始字节）
         self.on_end: Callable[[str, Dict[str, Any]], None] = lambda req_id, data: None  # 合成结束回调
         self.on_error: Callable[[str, str], None] = lambda req_id, msg: None  # 错误回调
         self.on_queue_full: Callable[[str], None] = lambda req_id: None  # 队列满回调
 
-        # 音频播放流
+        # 音频播放相关（支持MP3格式直接播放）
         self.play_stream: Optional[sd.OutputStream] = None
         self.current_req_id: Optional[str] = None
+        self.enable_playback: bool = True  # 是否启用实时播放
 
-        # 是否启用播放功能
-        self.enable_playback: bool = True
-
-        # 外部回调函数，用于返回完整结果
+        # 外部回调函数（返回完整结果）
         self.external_callback: Optional[Callable[[str, Dict[str, Any]], None]] = None
 
-        # 存储每个请求的音频数据
-        self.audio_buffers: Dict[str, List[np.ndarray]] = {}
+        # 存储每个请求的完整音频数据（原始字节）
+        self.audio_buffers: Dict[str, List[bytes]] = {}
+
+    def _get_resource_id(self, voice_type: str) -> str:
+        """根据音色类型获取资源ID（字节跳动TTS协议要求）"""
+        if voice_type.startswith("S_"):
+            return "volc.megatts.default"
+        return "volc.service_type.10029"
+
+    async def _create_websocket_connection(self):
+        """创建WebSocket连接（内部使用）"""
+        headers = {
+            "X-Api-App-Key": self.appid,
+            "X-Api-Access-Key": self.access_token,
+            "X-Api-Resource-Id": self._get_resource_id(DEFAULT_VOICE_TYPE),  # 用默认音色获取资源ID
+            "X-Api-Connect-Id": str(uuid.uuid4()),
+        }
+
+        print(f"连接到 TTS 服务端: {self.endpoint}")
+        self.websocket = await websockets.connect(
+            self.endpoint,
+            additional_headers=headers,
+            max_size=10 * 1024 * 1024  # 10MB缓冲区
+        )
+        self.is_connected = True
+        self.logid = self.websocket.response.headers.get("x-tt-logid")
+        print(f"连接成功，LogID: {self.logid}")
+
+        # 发送连接启动指令
+        await start_connection(self.websocket)
+        await wait_for_event(
+            self.websocket, MsgType.FullServerResponse, EventType.ConnectionStarted
+        )
+        print("TTS连接已初始化完成")
 
     async def connect(self):
-        """建立 WebSocket 连接（初始化一次）"""
+        """建立WebSocket连接（外部调用，初始化一次）"""
         if not self.is_connected:
             try:
-                self.websocket = await websockets.connect(self.ws_url)
-                self.is_connected = True
-                print(f"成功连接到 TTS 服务端: {self.ws_url}")
-
+                await self._create_websocket_connection()
                 # 启动队列消费协程（后台运行）
                 asyncio.create_task(self._consume_queue())
             except Exception as e:
-                raise ConnectionError(f"连接失败: {str(e)}")
+                error_msg = f"连接失败: {str(e)}"
+                print(error_msg)
+                raise ConnectionError(error_msg)
 
     async def disconnect(self):
-        """关闭 WebSocket 连接"""
+        """关闭WebSocket连接"""
         if self.is_connected and self.websocket:
-            await self.websocket.close()
-            self.is_connected = False
-            self.websocket = None
-            print("已断开与 TTS 服务端的连接")
+            try:
+                # 发送连接结束指令
+                await finish_connection(self.websocket)
+                await wait_for_event(
+                    self.websocket, MsgType.FullServerResponse, EventType.ConnectionFinished
+                )
+            except Exception as e:
+                print(f"关闭连接时异常: {str(e)}")
+            finally:
+                await self.websocket.close()
+                self.is_connected = False
+                self.websocket = None
+                print("已断开与TTS服务端的连接")
 
+        # 清理音频播放流
         if self.play_stream:
             self.play_stream.stop()
             self.play_stream.close()
@@ -91,281 +165,377 @@ class CosyVoiceTTSSocketClient:
         self.external_callback = callback
 
     def set_playback_enabled(self, enabled: bool):
-        """设置是否启用音频播放"""
+        """设置是否启用音频实时播放"""
         self.enable_playback = enabled
+        print(f"音频实时播放已{'启用' if enabled else '禁用'}")
 
-    async def add_tts_request(self, tts_text: str, **kwargs) -> str:
+    async def synthesize(self, tts_text: str, **kwargs) -> str:
         """
-        添加TTS请求到队列（异步非阻塞）
+        异步非阻塞添加TTS请求到队列
         :param tts_text: 要合成的文本
-        :param kwargs: 其他TTS参数
+        :param kwargs: 其他TTS参数（voice_type, encoding, speed等）
         :return: 唯一请求ID
         """
-        # 创建请求对象
+        # 创建请求对象（支持覆盖默认参数）
         request = TTSRequest(tts_text=tts_text, **kwargs)
         req_id = request.request_id
 
         # 初始化音频缓冲区
         self.audio_buffers[req_id] = []
 
-        # 异步入队
+        # 异步入队（非阻塞）
         try:
             await self.request_queue.put(request)
             self.on_task_enqueue(req_id)
-            print(f"请求 {req_id} 已加入队列，当前队列长度: {self.request_queue.qsize()}")
+            print(f"请求 [{req_id[:8]}] 已加入队列，当前队列长度: {self.request_queue.qsize()}")
             return req_id
         except asyncio.QueueFull:
             self.on_queue_full(req_id)
-            raise Exception(f"队列已满，请求 {req_id} 入队失败")
+            error_msg = f"队列已满（最大长度{self.max_queue_size}），请求 [{req_id[:8]}] 入队失败"
+            print(error_msg)
+            raise Exception(error_msg)
 
     async def _consume_queue(self):
         """消费队列（后台协程，自动处理排队请求）"""
         print("队列消费协程已启动")
         while True:
             try:
-                # 等待队列中有请求（阻塞）
+                # 等待队列中有请求（阻塞，直到有任务）
                 request = await self.request_queue.get()
                 req_id = request.request_id
 
                 # 标记为处理中
                 self.is_processing = True
-                print(f"开始处理请求 {req_id}，剩余队列长度: {self.request_queue.qsize()}")
+                # print(f"\n开始处理请求 [{req_id[:8]}]，剩余队列长度: {self.request_queue.qsize()}")
 
-                # 处理当前请求
+                # 处理单个请求
                 await self._process_single_request(request)
 
-                # 标记任务完成
+                # 标记任务完成（让Queue知道可以继续）
                 self.request_queue.task_done()
                 self.is_processing = False
 
             except Exception as e:
-                print(f"队列消费异常: {str(e)}")
+                error_msg = f"队列消费异常: {str(e)}"
+                print(error_msg)
                 self.is_processing = False
-                # 短暂等待后继续消费，避免死循环
+                # 短暂等待，避免死循环占用CPU
                 await asyncio.sleep(0.1)
 
-    async def _process_single_request(self, request: TTSRequest):
-        """处理单个TTS请求"""
-        req_id = request.request_id
-
-        # 参数校验
-        if not request.tts_text:
-            self.on_error(req_id, "合成文本不能为空")
-            return
-
-        # 确保已连接
-        if not self.is_connected:
-            try:
-                await self.connect()
-            except Exception as e:
-                self.on_error(req_id, f"连接失败: {str(e)}")
-                return
-
-        # 构造请求数据
-        request_data = {
-            "tts_text": request.tts_text,
-            "mode": request.mode,
-            "sft_spk": request.sft_spk,
-            "seed": request.seed,
-            "stream": request.stream,
-            "speed": request.speed,
-            "prompt_wav": request.prompt_wav,
-            "instruct_text": request.instruct_text
+    def _build_base_request(self, request: TTSRequest) -> Dict[str, Any]:
+        """构建字节跳动TTS基础请求参数"""
+        aaa = {
+            "user": {"uid": str(uuid.uuid4())},
+            "namespace": "BidirectionalTTS",
+            "req_params": {
+                "speaker": request.voice_type,
+                "audio_params": {
+                    "format": request.encoding,
+                    "sample_rate": DEFAULT_SAMPLE_RATE,
+                    "enable_timestamp": True,
+                },
+                "additions": json.dumps({"disable_markdown_filter": False}),
+                "speed": request.speed,  # 语速参数（需服务端支持）
+            },
         }
+        print('aaa', aaa)
+        return aaa
+    async def _send_text_stream(self, request: TTSRequest, session_id: str):
+        """流式发送文本（逐字符发送，字节跳动TTS流式协议要求）"""
+        base_request = self._build_base_request(request)
+        text = request.tts_text.strip()
 
-        try:
-            # 发送请求
-            await self.websocket.send(json.dumps(request_data))
-
-            # 处理返回的流式数据
-            await self._handle_response(req_id)
-
-        except websockets.exceptions.ConnectionClosed:
-            self.is_connected = False
-            self.on_error(req_id, "连接已关闭")
-        except Exception as e:
-            self.on_error(req_id, f"处理请求失败: {str(e)}")
-
-    async def _handle_response(self, req_id: str):
-        """处理单个请求的服务端响应"""
-        if not self.websocket:
+        if not text:
+            print(f"请求 [{request.request_id[:8]}] 文本为空，跳过发送")
             return
 
+        # 逐字符发送（控制发送速率，避免拥塞）
+        for char in text:
+            if not self.is_connected or not self.websocket:
+                raise ConnectionError("连接已断开，无法继续发送文本")
+
+            # 构建单个字符的任务请求
+            task_req = copy.deepcopy(base_request)
+            task_req["event"] = EventType.TaskRequest
+            task_req["req_params"]["text"] = char
+
+            # 发送任务请求
+            await task_request(
+                self.websocket,
+                json.dumps(task_req).encode("utf-8"),
+                session_id
+            )
+            # 控制发送速率（5ms/字符，可调整）
+            await asyncio.sleep(0.005)
+
+        # 发送会话结束指令
+        await finish_session(self.websocket, session_id)
+        print(f"请求 [{request.request_id[:8]}] 文本发送完成")
+
+    async def _handle_audio_response(self, req_id: str, session_id: str, request: TTSRequest) -> Dict[str, Any]:
+        """处理服务端的流式音频响应"""
+        if not self.websocket:
+            raise ConnectionError("WebSocket连接未建立")
+
+        sample_rate = DEFAULT_SAMPLE_RATE
+        audio_received = False
+
         try:
-            sample_rate = None
             while True:
-                # 接收服务端消息（异步）
-                response = await self.websocket.recv()
-                data = json.loads(response)
+                # 接收服务端消息（异步阻塞）
+                msg = await receive_message(self.websocket)
 
-                # 根据状态分发到不同回调（都带req_id）
-                status = data.get("status")
-                if status == "start":
-                    # 合成开始 - 返回采样率等信息
-                    sample_rate = data.get("sample_rate")
-                    self.on_start(req_id, data)
+                if msg.type == MsgType.FullServerResponse:
+                    # 完整响应（开始/结束/错误）
+                    if msg.event == EventType.SessionStarted:
+                        # 会话开始回调
+                        start_data = {
+                            "session_id": session_id,
+                            "sample_rate": sample_rate,
+                            "encoding": request.encoding,
+                            "voice_type": request.voice_type,
+                            "logid": self.logid
+                        }
+                        self.on_start(req_id, start_data)
+                        print(f"请求 [{req_id[:8]}] 合成开始")
 
-                elif status == "stream":
-                    # 流式音频块 - 转为 numpy 数组
-                    audio_chunk = np.array(data["audio_chunk"], dtype=np.float32)
+                    elif msg.event == EventType.SessionFinished:
+                        # 会话结束，退出循环
+                        end_data = {"session_id": session_id, "message": "合成完成"}
+                        self.on_end(req_id, end_data)
+                        print(f"请求 [{req_id[:8]}] 合成结束")
+                        break
 
-                    # 保存到缓冲区
-                    self.audio_buffers[req_id].append(audio_chunk)
+                elif msg.type == MsgType.AudioOnlyServer:
+                    # 流式音频数据（原始字节）
+                    audio_chunk = msg.payload
+                    if audio_chunk:
+                        audio_received = True
+                        # 保存到缓冲区
+                        self.audio_buffers[req_id].append(audio_chunk)
+                        # 音频块回调
+                        self.on_audio_chunk(req_id, audio_chunk)
+                        # 实时播放（如果启用）
+                        await self._play_audio_chunk(req_id, audio_chunk, sample_rate)
 
-                    # 音频块回调
-                    self.on_audio_chunk(req_id, audio_chunk)
+                else:
+                    # 未知消息类型
+                    raise RuntimeError(f"收到未知消息类型: {msg.type}, 内容: {msg}")
 
-                elif status == "end":
-                    # 合成结束
-                    self.on_end(req_id, data)
-
-                    # 组装完整音频数据
-                    full_audio = np.concatenate(self.audio_buffers[req_id]) if self.audio_buffers[req_id] else np.array(
-                        [], dtype=np.float32)
-
-                    # 准备返回给外部回调的数据
-                    result_data = {
-                        "status": "completed",
-                        "request_id": req_id,
-                        "sample_rate": sample_rate,
-                        "audio_data": full_audio,
-                        "audio_length": len(full_audio),
-                        "message": data.get("msg", "合成完成")
-                    }
-
-                    # 调用外部回调（非阻塞）
-                    if self.external_callback:
-                        # 使用create_task确保回调不会阻塞主流程
-                        asyncio.create_task(self._call_external_callback(req_id, result_data))
-
-                    # 清理缓冲区
-                    if req_id in self.audio_buffers:
-                        del self.audio_buffers[req_id]
-
-                    break
-
-                elif status == "error":
-                    # 错误处理
-                    error_msg = data.get("msg", "未知错误")
-                    self.on_error(req_id, error_msg)
-
-                    # 准备错误结果数据
-                    error_data = {
-                        "status": "error",
-                        "request_id": req_id,
-                        "message": error_msg
-                    }
-
-                    # 调用外部回调
-                    if self.external_callback:
-                        asyncio.create_task(self._call_external_callback(req_id, error_data))
-
-                    # 清理缓冲区
-                    if req_id in self.audio_buffers:
-                        del self.audio_buffers[req_id]
-
-                    break
+            # 组装完整结果
+            full_audio = b"".join(self.audio_buffers[req_id]) if self.audio_buffers[req_id] else b""
+            return {
+                "status": "completed",
+                "request_id": req_id,
+                "session_id": session_id,
+                "sample_rate": sample_rate,
+                "encoding": request.encoding,
+                "audio_data": full_audio,  # 完整音频字节数据
+                "audio_length": len(full_audio),
+                "message": "合成成功" if audio_received else "合成完成但未收到音频数据"
+            }
 
         except Exception as e:
-            error_msg = f"接收数据异常: {str(e)}"
+            error_msg = f"处理音频响应异常: {str(e)}"
             self.on_error(req_id, error_msg)
-
-            # 准备错误结果数据
-            error_data = {
+            return {
                 "status": "error",
                 "request_id": req_id,
+                "session_id": session_id,
                 "message": error_msg
             }
 
-            # 调用外部回调
-            if self.external_callback:
-                asyncio.create_task(self._call_external_callback(req_id, error_data))
+    async def _play_audio_chunk(self, req_id: str, chunk: bytes, sample_rate: int):
+        """实时播放音频块（支持MP3格式）"""
+        if not self.enable_playback:
+            return
 
+        # 确保当前请求是正在播放的请求
+        if self.current_req_id is None:
+            self.current_req_id = req_id
+
+        if req_id != self.current_req_id:
+            # 切换请求时，重置播放流
+            if self.play_stream:
+                self.play_stream.stop()
+                self.play_stream.close()
+            self.current_req_id = req_id
+
+        try:
+            # 初始化播放流（如果未初始化）
+            if not self.play_stream:
+                self.play_stream = sd.OutputStream(
+                    samplerate=sample_rate,
+                    channels=1,  # 单声道
+                    dtype=np.float32
+                )
+                self.play_stream.start()
+
+            # MP3字节 → 音频数组（直接播放）
+            # 注意：sounddevice默认支持PCM格式，如果是MP3需要解码，这里简化处理（实际使用建议用pydub解码）
+            # 如需支持MP3播放，请安装 pydub: pip install pydub ffmpeg
+            try:
+                # 简化处理：假设服务端返回PCM（如果是MP3，需替换为解码逻辑）
+                audio_array = np.frombuffer(chunk, dtype=np.float32)
+                if audio_array.size > 0:
+                    self.play_stream.write(audio_array)
+            except Exception as e:
+                print(f"音频播放异常: {str(e)}，请确保音频格式正确")
+
+        except Exception as e:
+            print(f"播放流初始化失败: {str(e)}")
+
+    async def _process_single_request(self, request: TTSRequest):
+        """处理单个TTS请求（完整流程：连接→启动会话→流式发送文本→接收音频→回调结果）"""
+        req_id = request.request_id
+        session_id = request.session_id
+
+        # 参数校验
+        if not request.tts_text.strip():
+            error_msg = "合成文本不能为空"
+            self.on_error(req_id, error_msg)
+            self._send_external_callback(req_id, {
+                "status": "error",
+                "request_id": req_id,
+                "session_id": session_id,
+                "message": error_msg
+            })
+            return
+
+        # 确保连接已建立（断开时自动重连）
+        if not self.is_connected:
+            print(f"请求 [{req_id[:8]}] 处理时连接已断开，尝试重连...")
+            try:
+                await self._create_websocket_connection()
+            except Exception as e:
+                error_msg = f"重连失败: {str(e)}"
+                self.on_error(req_id, error_msg)
+                self._send_external_callback(req_id, {
+                    "status": "error",
+                    "request_id": req_id,
+                    "session_id": session_id,
+                    "message": error_msg
+                })
+                return
+
+        try:
+            # 1. 启动会话
+            base_request = self._build_base_request(request)
+            start_session_req = copy.deepcopy(base_request)
+            start_session_req["event"] = EventType.StartSession
+
+            await start_session(
+                self.websocket,
+                json.dumps(start_session_req).encode("utf-8"),
+                session_id
+            )
+            # 等待会话启动成功
+            await wait_for_event(
+                self.websocket, MsgType.FullServerResponse, EventType.SessionStarted
+            )
+
+            # 2. 异步流式发送文本（后台任务，不阻塞接收音频）
+            send_task = asyncio.create_task(self._send_text_stream(request, session_id))
+
+            # 3. 接收并处理音频响应
+            result_data = await self._handle_audio_response(req_id, session_id, request)
+
+            # 4. 等待文本发送任务完成
+            await send_task
+
+            # 5. 发送外部回调
+            self._send_external_callback(req_id, result_data)
+
+        except Exception as e:
+            error_msg = f"处理请求 [{req_id[:8]}] 异常: {str(e)}"
+            self.on_error(req_id, error_msg)
+            self._send_external_callback(req_id, {
+                "status": "error",
+                "request_id": req_id,
+                "session_id": session_id,
+                "message": error_msg
+            })
+        finally:
             # 清理缓冲区
             if req_id in self.audio_buffers:
                 del self.audio_buffers[req_id]
 
-    async def _call_external_callback(self, req_id: str, result_data: Dict[str, Any]):
-        """异步调用外部回调函数"""
-        try:
-            if self.external_callback:
-                # 确保回调是异步的，不会阻塞
-                if asyncio.iscoroutinefunction(self.external_callback):
-                    await self.external_callback(req_id, result_data)
-                else:
-                    # 如果是同步函数，在线程池中执行
-                    await asyncio.get_event_loop().run_in_executor(
-                        None, self.external_callback, req_id, result_data
-                    )
-        except Exception as e:
-            print(f"外部回调执行异常: {str(e)}")
-
-    def _play_audio_chunk(self, chunk: np.ndarray):
-        """播放音频块（非阻塞）"""
-        if not self.enable_playback or not self.play_stream:
+    def _send_external_callback(self, req_id: str, result_data: Dict[str, Any]):
+        """发送外部回调（支持同步/异步回调函数）"""
+        if not self.external_callback:
             return
 
         try:
-            if chunk.size > 0:
-                self.play_stream.write(chunk)
+            # 异步回调：直接await
+            if asyncio.iscoroutinefunction(self.external_callback):
+                asyncio.create_task(self.external_callback(req_id, result_data))
+            # 同步回调：在线程池中执行（避免阻塞事件循环）
+            else:
+                asyncio.get_event_loop().run_in_executor(
+                    None, self.external_callback, req_id, result_data
+                )
         except Exception as e:
-            print(f"音频播放异常: {str(e)}")
+            print(f"外部回调执行异常: {str(e)}")
 
-    async def wait_queue_empty(self):
-        """等待队列所有任务处理完成（阻塞）"""
+    async def wait_all_completed(self):
+        """等待队列中所有任务处理完成（阻塞）"""
         await self.request_queue.join()
-        print("所有队列任务已处理完成")
+        print("\n所有队列任务已处理完成")
 
 
 # ------------------------------
-# 使用示例
+# 使用示例（与你提供的风格完全一致）
 # ------------------------------
 class TTSManager:
-    """TTS管理器 - 供外部代码调用"""
+    """TTS管理器 - 供外部代码调用（封装客户端，简化使用）"""
 
-    def __init__(self, ws_url: str = "ws://localhost:50000/ws/tts"):
-        self.client = CosyVoiceTTSSocketClient(ws_url)
-        self._setup_callbacks()
+    def __init__(
+        self,
+        appid: str = DEFAULT_APPID,
+        access_token: str = DEFAULT_ACCESS_TOKEN,
+        endpoint: str = DEFAULT_ENDPOINT
+    ):
+        self.client = ByteDanceTTSSocketClient(
+            appid=appid,
+            access_token=access_token,
+            endpoint=endpoint
+        )
+        self._setup_internal_callbacks()
 
-    def _setup_callbacks(self):
-        """设置内部回调"""
+    def _setup_internal_callbacks(self):
+        """设置内部回调（日志/状态提示）"""
 
-        # 合成开始回调
-        def on_tts_start(req_id, data):
-            print(f"\n🎤 开始合成 [{req_id[:8]}] - 采样率: {data.get('sample_rate')}")
+        def on_task_enqueue(req_id: str):
+            """任务入队回调"""
+            print(f"📥 任务 [{req_id[:8]}] 已入队")
 
-            # 初始化播放流（如果需要播放）
-            if self.client.enable_playback:
-                sample_rate = data.get("sample_rate", 24000)
-                if self.client.play_stream:
-                    self.client.play_stream.stop()
-                    self.client.play_stream.close()
+        def on_tts_start(req_id: str, data: Dict[str, Any]):
+            """合成开始回调"""
+            print(f"🎤 合成开始 [{req_id[:8]}] - 采样率: {data['sample_rate']}, 编码: {data['encoding']}")
 
-                self.client.play_stream = sd.OutputStream(
-                    samplerate=sample_rate,
-                    channels=1,
-                    dtype=np.float32
-                )
-                self.client.play_stream.start()
-                self.client.current_req_id = req_id
+        def on_audio_chunk(req_id: str, chunk: bytes):
+            """音频块回调（内部仅打印日志，外部通过external_callback获取）"""
+            print(f"🔊 收到音频块 [{req_id[:8]}] - 大小: {len(chunk)}字节", end="\r")
 
-        # 音频块回调
-        def on_audio_chunk(req_id, chunk):
-            if req_id == self.client.current_req_id:
-                self.client._play_audio_chunk(chunk)
+        def on_tts_end(req_id: str, data: Dict[str, Any]):
+            """合成结束回调"""
+            print(f"\n🏁 合成结束 [{req_id[:8]}] - 会话ID: {data['session_id']}")
 
-        # 合成结束回调
-        def on_tts_end(req_id, data):
-            print(f"🏁 合成完成 [{req_id[:8]}] - {data.get('msg', '完成')}")
+        def on_tts_error(req_id: str, msg: str):
+            """错误回调"""
+            print(f"\n❌ 合成失败 [{req_id[:8]}] - 错误: {msg}")
 
-        # 错误回调
-        def on_tts_error(req_id, msg):
-            print(f"❌ 合成失败 [{req_id[:8]}] - {msg}")
+        def on_queue_full(req_id: str):
+            """队列满回调"""
+            print(f"⚠️  队列已满，请求 [{req_id[:8]}] 入队失败")
 
+        # 绑定内部回调
+        self.client.on_task_enqueue = on_task_enqueue
         self.client.on_start = on_tts_start
         self.client.on_audio_chunk = on_audio_chunk
         self.client.on_end = on_tts_end
         self.client.on_error = on_tts_error
+        self.client.on_queue_full = on_queue_full
 
     async def initialize(self):
         """初始化连接"""
@@ -376,25 +546,25 @@ class TTSManager:
         await self.client.disconnect()
 
     def set_result_callback(self, callback: Callable[[str, Dict[str, Any]], None]):
-        """设置结果回调函数"""
+        """设置外部结果回调（获取完整音频数据）"""
         self.client.set_external_callback(callback)
 
     def set_playback_enabled(self, enabled: bool):
-        """设置是否启用播放"""
+        """设置是否启用实时播放"""
         self.client.set_playback_enabled(enabled)
 
     async def synthesize(self, text: str, **kwargs) -> str:
         """
-        异步合成文本（非阻塞）
+        异步非阻塞合成文本
         :param text: 要合成的文本
-        :param kwargs: 其他参数（mode, sft_spk, speed等）
+        :param kwargs: 其他参数（voice_type, encoding, speed等）
         :return: 请求ID
         """
-        return await self.client.add_tts_request(text, **kwargs)
+        return await self.client.synthesize(text, **kwargs)
 
     async def wait_all_completed(self):
         """等待所有任务完成"""
-        await self.client.wait_queue_empty()
+        await self.client.wait_all_completed()
 
 
 # ------------------------------
@@ -402,147 +572,120 @@ class TTSManager:
 # ------------------------------
 async def external_usage_example():
     """外部代码使用示例"""
+    # 1. 创建TTS管理器（可替换为自己的appid和access_token）
+    tts_manager = TTSManager(
+        appid=DEFAULT_APPID,
+        access_token=DEFAULT_ACCESS_TOKEN,
+        endpoint=DEFAULT_ENDPOINT
+    )
 
-    # 1. 创建TTS管理器
-    tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
-
-    # 2. 设置结果回调函数（接收完整结果）
-    def handle_tts_result1(req_id: str, result: Dict[str, Any]):
-        """处理TTS结果回调"""
+    # 2. 设置外部结果回调（获取完整音频数据）
+    def handle_tts_result(req_id: str, result: Dict[str, Any]):
+        """处理TTS完整结果（同步回调）"""
         status = result.get("status")
-
         if status == "completed":
             audio_data = result.get("audio_data")
-            sample_rate = result.get("sample_rate")
+            encoding = result.get("encoding")
             audio_length = result.get("audio_length")
 
-            print(f"✅ 收到TTS结果 [{req_id[:8]}]: 长度{audio_length}采样点, 采样率{sample_rate}Hz")
+            print(f"\n✅ 收到完整结果 [{req_id[:8]}] - 长度: {audio_length}字节, 编码: {encoding}")
 
-            # 这里可以保存音频文件或进行其他处理
-            # 注意：audio_data是完整的numpy数组
+            # 保存音频文件
+            filename = f"tts_output_{req_id[:8]}.{encoding}"
+            with open(filename, "wb") as f:
+                f.write(audio_data)
+            print(f"💾 音频文件已保存: {filename}")
 
         elif status == "error":
             error_msg = result.get("message")
-            print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
+            print(f"\n❌ 请求 [{req_id[:8]}] 处理失败: {error_msg}")
 
-    def handle_tts_result(req_id: str, result: Dict[str, Any]):
-        """处理TTS结果回调 - 保存为PCM文件"""
-        print('处理TTS结果回调', result)
-        status = result.get("status")
-        print('status')
-        if status == "completed":
-            audio_data = result.get("audio_data")
-            sample_rate = result.get("sample_rate")
-
-            if audio_data is not None and len(audio_data) > 0:
-                # 转换为PCM数据
-                pcm_data = (audio_data.astype(np.float32) * 32767).astype(np.int16)
-                pcm_bytes = pcm_data.tobytes()
-
-                # 保存为PCM文件
-                filename = f"tts_output_{req_id[:8]}.pcm"
-                with open(filename, "wb") as f:
-                    f.write(pcm_bytes)
-
-                print(f"💾 PCM文件已保存: {filename}, 大小: {len(pcm_bytes)}字节")
-
-                return filename
-
-        elif status == "error":
-            error_msg = result.get("message")
-            print(f"❌ TTS处理失败 [{req_id[:8]}]: {error_msg}")
-
-        return None
+    # 绑定外部回调
     tts_manager.set_result_callback(handle_tts_result)
 
-    # 3. 设置是否播放（可选，默认True）
-    tts_manager.set_playback_enabled(True)  # 设置为False则不播放
+    # 3. 设置是否启用实时播放（默认True）
+    tts_manager.set_playback_enabled(True)
 
     # 4. 初始化连接
     await tts_manager.initialize()
 
-    # 5. 异步合成多个文本（非阻塞）
+    # 5. 异步提交多个TTS请求（非阻塞）
     texts = [
-        "你好，这是第一个排队的TTS请求。",
-        "我是第二个请求，会等第一个处理完再执行。",
-        "第三个请求，支持流式播放和队列管理。",
-        "第四个请求，测试队列的自动消费功能。",
-        "最后一个请求，处理完成后会自动结束。"
+        "你好，这是字节跳动TTS的流式合成测试。",
+        "我支持异步非阻塞调用，多个请求可以排队处理。",
+        "每个请求都会返回唯一的ID，方便你跟踪结果。",
+        "音频数据会通过回调函数返回，支持实时播放和保存文件。",
+        "最后一个测试句子，演示队列的自动消费功能。"
     ]
 
     req_ids = []
     for i, text in enumerate(texts):
+        # 提交请求（非阻塞，立即返回）
         req_id = await tts_manager.synthesize(
             text,
-            mode="预训练音色",
-            sft_spk="中文女",
+            voice_type=DEFAULT_VOICE_TYPE,
+            encoding=DEFAULT_ENCODING,
             speed=1.0
         )
         req_ids.append(req_id)
-        print(f"已提交请求 {i + 1}: ID={req_id[:8]}")
+        print(f"📤 已提交请求 {i+1}: ID={req_id[:8]}")
 
-        # 可以立即继续其他操作，不需要等待
-        await asyncio.sleep(0.5)  # 模拟其他操作
+        # 模拟其他业务逻辑（无需等待TTS完成）
+        await asyncio.sleep(0.3)
 
-    # 6. 可以在这里做其他事情，TTS会在后台处理
-
-    # 7. 等待所有TTS任务完成（可选）
+    # 6. 等待所有TTS任务完成（可选，根据业务需求决定是否等待）
     await tts_manager.wait_all_completed()
 
-    # 8. 关闭连接
+    # 7. 关闭连接（程序退出前调用）
     await tts_manager.shutdown()
 
 
+# ------------------------------
+# 异步结果回调示例（高级用法）
+# ------------------------------
+async def async_result_callback(req_id: str, result: Dict[str, Any]):
+    """异步结果回调（支持异步操作，如上传音频到服务器）"""
+    if result["status"] == "completed":
+        print(f"\n⚡ 异步处理结果 [{req_id[:8]}] - 开始上传音频...")
+        # 模拟异步上传操作
+        await asyncio.sleep(0.5)
+        print(f"⚡ 异步处理结果 [{req_id[:8]}] - 音频上传完成")
+
+
 async def advanced_usage_example():
-    """高级使用示例：动态控制播放和结果处理"""
+    """高级使用示例：异步回调 + 禁用播放 + 批量请求"""
+    tts_manager = TTSManager()
 
-    tts_manager = TTSManager(ws_url="ws://10.10.10.202:50000/ws/tts")
+    # 设置异步结果回调
+    tts_manager.set_result_callback(async_result_callback)
 
-    # 自定义结果处理器
-    class ResultProcessor:
-        def __init__(self):
-            self.results = {}
-
-        async def process_result(self, req_id: str, result: Dict[str, Any]):
-            """异步处理结果"""
-            if result.get("status") == "completed":
-                # 保存音频数据等处理
-                self.results[req_id] = result
-                print(f"处理完成: {req_id[:8]}, 音频长度: {result.get('audio_length')}")
-
-                # 这里可以添加自定义逻辑，如保存到文件、发送到网络等
-
-    processor = ResultProcessor()
-    tts_manager.set_result_callback(processor.process_result)
-
-    # 禁用播放，只获取数据
+    # 禁用实时播放（只获取音频数据）
     tts_manager.set_playback_enabled(False)
 
     await tts_manager.initialize()
 
-    # 批量合成
+    # 批量提交请求（并行提交）
     tasks = []
     for i in range(3):
-        text = f"这是第{i+1}个测试文本，用于演示高级功能。"
-        task = tts_manager.synthesize(text)
+        text = f"这是第{i+1}个高级测试文本，使用异步回调处理结果。"
+        task = tts_manager.synthesize(text, speed=0.9)
         tasks.append(task)
 
-        # 并行提交所有请求
-        req_ids = await asyncio.gather(*tasks)
+    # 并行提交所有请求
+    req_ids = await asyncio.gather(*tasks)
+    print(f"\n已并行提交 {len(req_ids)} 个请求")
 
-        # 等待完成
-        await tts_manager.wait_all_completed()
-        await tts_manager.shutdown()
+    # 等待所有任务完成
+    await tts_manager.wait_all_completed()
+    await tts_manager.shutdown()
 
-        print(f"所有任务完成，共处理 {len(processor.results)} 个结果")
 
 if __name__ == "__main__":
-    # 运行示例
     try:
-        # 基础使用示例
+        # 运行基础使用示例
         asyncio.run(external_usage_example())
 
-        # 高级使用示例（取消注释运行）
+        # 运行高级使用示例（取消注释）
         # asyncio.run(advanced_usage_example())
 
     except KeyboardInterrupt:

volcengine_bidirection_demo/.idea/.gitignore

@@ -0,0 +1,5 @@
+# 默认忽略的文件
+/shelf/
+/workspace.xml
+# 基于编辑器的 HTTP 客户端请求
+/httpRequests/

volcengine_bidirection_demo/.idea/inspectionProfiles/Project_Default.xml

@@ -0,0 +1,12 @@
+<component name="InspectionProjectProfileManager">
+  <profile version="1.0">
+    <option name="myName" value="Project Default" />
+    <inspection_tool class="PyUnresolvedReferencesInspection" enabled="true" level="WARNING" enabled_by_default="true">
+      <option name="ignoredIdentifiers">
+        <list>
+          <option value="javalang.tree" />
+        </list>
+      </option>
+    </inspection_tool>
+  </profile>
+</component>

volcengine_bidirection_demo/.idea/inspectionProfiles/profiles_settings.xml

@@ -0,0 +1,6 @@
+<component name="InspectionProjectProfileManager">
+  <settings>
+    <option name="USE_PROJECT_PROFILE" value="false" />
+    <version value="1.0" />
+  </settings>
+</component>

volcengine_bidirection_demo/.idea/misc.xml

@@ -0,0 +1,7 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="Black">
+    <option name="sdkName" value="Python 3.13 (volcengine_bidirection_demo)" />
+  </component>
+  <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.13 (volcengine_bidirection_demo)" project-jdk-type="Python SDK" />
+</project>

volcengine_bidirection_demo/.idea/modules.xml

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectModuleManager">
+    <modules>
+      <module fileurl="file://$PROJECT_DIR$/.idea/volcengine_bidirection_demo.iml" filepath="$PROJECT_DIR$/.idea/volcengine_bidirection_demo.iml" />
+    </modules>
+  </component>
+</project>

volcengine_bidirection_demo/.idea/vcs.xml

@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="VcsDirectoryMappings">
+    <mapping directory="$PROJECT_DIR$/.." vcs="Git" />
+  </component>
+</project>

volcengine_bidirection_demo/.idea/volcengine_bidirection_demo.iml

@@ -0,0 +1,14 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<module type="PYTHON_MODULE" version="4">
+  <component name="NewModuleRootManager">
+    <content url="file://$MODULE_DIR$">
+      <excludeFolder url="file://$MODULE_DIR$/.venv" />
+    </content>
+    <orderEntry type="jdk" jdkName="Python 3.13 (volcengine_bidirection_demo)" jdkType="Python SDK" />
+    <orderEntry type="sourceFolder" forTests="false" />
+  </component>
+  <component name="PyDocumentationSettings">
+    <option name="format" value="PLAIN" />
+    <option name="myDocStringFormat" value="Plain" />
+  </component>
+</module>

volcengine_bidirection_demo/examples/volcengine/bidirection.py

@@ -0,0 +1,169 @@
+#!/usr/bin/env python3
+import argparse
+import asyncio
+import copy
+import json
+import logging
+import uuid
+
+import websockets
+
+from protocols import (
+    EventType,
+    MsgType,
+    finish_connection,
+    finish_session,
+    receive_message,
+    start_connection,
+    start_session,
+    task_request,
+    wait_for_event,
+)
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+
+def get_resource_id(voice: str) -> str:
+    if voice.startswith("S_"):
+        return "volc.megatts.default"
+    return "volc.service_type.10029"
+
+
+async def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--appid", required=False, help="APP ID", default='7069844318')
+    parser.add_argument("--access_token", required=False, help="Access Token", default='osFMEJr20SSTWRql43cJlZkAOg7iwvxu')
+    parser.add_argument("--resource_id", default="", help="seed-tts-2.0")
+    parser.add_argument("--text", required=False, help="Text to convert", default='测试一下豆包的tts')
+    parser.add_argument("--voice_type", default="zh_female_xueayi_saturn_bigtts", required=False, help="Voice type")
+    parser.add_argument("--encoding", default="mp3", help="Output file encoding")
+    parser.add_argument(
+        "--endpoint",
+        default="wss://openspeech.bytedance.com/api/v3/tts/bidirection",
+        help="WebSocket endpoint URL",
+    )
+
+    args = parser.parse_args()
+
+    # Connect to server
+    headers = {
+        "X-Api-App-Key": args.appid,
+        "X-Api-Access-Key": args.access_token,
+        "X-Api-Resource-Id": (
+            args.resource_id if args.resource_id else get_resource_id(args.voice_type)
+        ),
+        "X-Api-Connect-Id": str(uuid.uuid4()),
+    }
+
+    logger.info(f"Connecting to {args.endpoint} with headers: {headers}")
+    websocket = await websockets.connect(
+        args.endpoint, additional_headers=headers, max_size=10 * 1024 * 1024
+    )
+    logger.info(
+        f"Connected to WebSocket server, Logid: {websocket.response.headers['x-tt-logid']}",
+    )
+
+    try:
+        # Start connection
+        await start_connection(websocket)
+        await wait_for_event(
+            websocket, MsgType.FullServerResponse, EventType.ConnectionStarted
+        )
+
+        # Process each sentence
+        sentences = args.text.split("。")
+        audio_received = False
+
+        for i, sentence in enumerate(sentences):
+            if not sentence:
+                continue
+
+            # every session can have different parameters
+            base_request = {
+                "user": {
+                    "uid": str(uuid.uuid4()),
+                },
+                "namespace": "BidirectionalTTS",
+                "req_params": {
+                    "speaker": args.voice_type,
+                    "audio_params": {
+                        "format": args.encoding,
+                        "sample_rate": 24000,
+                        "enable_timestamp": True,
+                    },
+                    "additions": json.dumps(
+                        {
+                            "disable_markdown_filter": False,
+                        }
+                    ),
+                },
+            }
+
+            # Start session
+            start_session_request = copy.deepcopy(base_request)
+            start_session_request["event"] = EventType.StartSession
+            session_id = str(uuid.uuid4())
+            await start_session(
+                websocket, json.dumps(start_session_request).encode(), session_id
+            )
+            await wait_for_event(
+                websocket, MsgType.FullServerResponse, EventType.SessionStarted
+            )
+
+            # Send characters one by one
+            async def send_chars():
+                for char in sentence:
+                    synthesis_request = copy.deepcopy(base_request)
+                    synthesis_request["event"] = EventType.TaskRequest
+                    synthesis_request["req_params"]["text"] = char
+                    await task_request(
+                        websocket, json.dumps(synthesis_request).encode(), session_id
+                    )
+                    await asyncio.sleep(0.005)  # 5ms delay between characters
+
+                await finish_session(websocket, session_id)
+
+            # Start sending characters in background
+            send_task = asyncio.create_task(send_chars())
+
+            # Receive audio data
+            audio_data = bytearray()
+            while True:
+                msg = await receive_message(websocket)
+
+                if msg.type == MsgType.FullServerResponse:
+                    if msg.event == EventType.SessionFinished:
+                        break
+                elif msg.type == MsgType.AudioOnlyServer:
+                    if not audio_received and len(audio_data) > 0:
+                        audio_received = True
+                    audio_data.extend(msg.payload)
+                else:
+                    raise RuntimeError(f"TTS conversion failed: {msg}")
+
+            # Wait for send_chars to complete
+            await send_task
+
+            # Save audio file if we received any data
+            if audio_data:
+                filename = f"{args.voice_type}_session_{i}.{args.encoding}"
+                with open(filename, "wb") as f:
+                    f.write(audio_data)
+                logger.info(f"Audio received: {len(audio_data)}, saved to {filename}")
+
+        if not audio_received:
+            raise RuntimeError("No audio data received")
+
+    finally:
+        # Finish connection
+        await finish_connection(websocket)
+        msg = await wait_for_event(
+            websocket, MsgType.FullServerResponse, EventType.ConnectionFinished
+        )
+        await websocket.close()
+        logger.info("Connection closed")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

volcengine_bidirection_demo/protocols/__init__.py

@@ -0,0 +1,41 @@
+from .protocols import (
+    CompressionBits,
+    EventType,
+    HeaderSizeBits,
+    Message,
+    MsgType,
+    MsgTypeFlagBits,
+    SerializationBits,
+    VersionBits,
+    audio_only_client,
+    cancel_session,
+    finish_connection,
+    finish_session,
+    full_client_request,
+    receive_message,
+    start_connection,
+    start_session,
+    task_request,
+    wait_for_event,
+)
+
+__all__ = [
+    "CompressionBits",
+    "EventType",
+    "HeaderSizeBits",
+    "Message",
+    "MsgType",
+    "MsgTypeFlagBits",
+    "SerializationBits",
+    "VersionBits",
+    "audio_only_client",
+    "cancel_session",
+    "finish_connection",
+    "finish_session",
+    "full_client_request",
+    "receive_message",
+    "start_connection",
+    "start_session",
+    "task_request",
+    "wait_for_event",
+]

volcengine_bidirection_demo/protocols/protocols.py

@@ -0,0 +1,543 @@
+import io
+import logging
+import struct
+from dataclasses import dataclass
+from enum import IntEnum
+from typing import Callable, List
+
+import websockets
+
+logger = logging.getLogger(__name__)
+
+
+class MsgType(IntEnum):
+    """Message type enumeration"""
+
+    Invalid = 0
+    FullClientRequest = 0b1
+    AudioOnlyClient = 0b10
+    FullServerResponse = 0b1001
+    AudioOnlyServer = 0b1011
+    FrontEndResultServer = 0b1100
+    Error = 0b1111
+
+    # Alias
+    ServerACK = AudioOnlyServer
+
+    def __str__(self) -> str:
+        return self.name if self.name else f"MsgType({self.value})"
+
+
+class MsgTypeFlagBits(IntEnum):
+    """Message type flag bits"""
+
+    NoSeq = 0  # Non-terminal packet with no sequence
+    PositiveSeq = 0b1  # Non-terminal packet with sequence > 0
+    LastNoSeq = 0b10  # Last packet with no sequence
+    NegativeSeq = 0b11  # Last packet with sequence < 0
+    WithEvent = 0b100  # Payload contains event number (int32)
+
+
+class VersionBits(IntEnum):
+    """Version bits"""
+
+    Version1 = 1
+    Version2 = 2
+    Version3 = 3
+    Version4 = 4
+
+
+class HeaderSizeBits(IntEnum):
+    """Header size bits"""
+
+    HeaderSize4 = 1
+    HeaderSize8 = 2
+    HeaderSize12 = 3
+    HeaderSize16 = 4
+
+
+class SerializationBits(IntEnum):
+    """Serialization method bits"""
+
+    Raw = 0
+    JSON = 0b1
+    Thrift = 0b11
+    Custom = 0b1111
+
+
+class CompressionBits(IntEnum):
+    """Compression method bits"""
+
+    None_ = 0
+    Gzip = 0b1
+    Custom = 0b1111
+
+
+class EventType(IntEnum):
+    """Event type enumeration"""
+
+    None_ = 0  # Default event
+
+    # 1 ~ 49 Upstream Connection events
+    StartConnection = 1
+    StartTask = 1  # Alias of StartConnection
+    FinishConnection = 2
+    FinishTask = 2  # Alias of FinishConnection
+
+    # 50 ~ 99 Downstream Connection events
+    ConnectionStarted = 50  # Connection established successfully
+    TaskStarted = 50  # Alias of ConnectionStarted
+    ConnectionFailed = 51  # Connection failed (possibly due to authentication failure)
+    TaskFailed = 51  # Alias of ConnectionFailed
+    ConnectionFinished = 52  # Connection ended
+    TaskFinished = 52  # Alias of ConnectionFinished
+
+    # 100 ~ 149 Upstream Session events
+    StartSession = 100
+    CancelSession = 101
+    FinishSession = 102
+
+    # 150 ~ 199 Downstream Session events
+    SessionStarted = 150
+    SessionCanceled = 151
+    SessionFinished = 152
+    SessionFailed = 153
+    UsageResponse = 154  # Usage response
+    ChargeData = 154  # Alias of UsageResponse
+
+    # 200 ~ 249 Upstream general events
+    TaskRequest = 200
+    UpdateConfig = 201
+
+    # 250 ~ 299 Downstream general events
+    AudioMuted = 250
+
+    # 300 ~ 349 Upstream TTS events
+    SayHello = 300
+
+    # 350 ~ 399 Downstream TTS events
+    TTSSentenceStart = 350
+    TTSSentenceEnd = 351
+    TTSResponse = 352
+    TTSEnded = 359
+    PodcastRoundStart = 360
+    PodcastRoundResponse = 361
+    PodcastRoundEnd = 362
+
+    # 450 ~ 499 Downstream ASR events
+    ASRInfo = 450
+    ASRResponse = 451
+    ASREnded = 459
+
+    # 500 ~ 549 Upstream dialogue events
+    ChatTTSText = 500  # (Ground-Truth-Alignment) text for speech synthesis
+
+    # 550 ~ 599 Downstream dialogue events
+    ChatResponse = 550
+    ChatEnded = 559
+
+    # 650 ~ 699 Downstream dialogue events
+    # Events for source (original) language subtitle
+    SourceSubtitleStart = 650
+    SourceSubtitleResponse = 651
+    SourceSubtitleEnd = 652
+    # Events for target (translation) language subtitle
+    TranslationSubtitleStart = 653
+    TranslationSubtitleResponse = 654
+    TranslationSubtitleEnd = 655
+
+    def __str__(self) -> str:
+        return self.name if self.name else f"EventType({self.value})"
+
+
+@dataclass
+class Message:
+    """Message object
+
+    Message format:
+    0                 1                 2                 3
+    | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 | 0 1 2 3 4 5 6 7 |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |    Version      |   Header Size   |     Msg Type    |      Flags      |
+    |   (4 bits)      |    (4 bits)     |     (4 bits)    |     (4 bits)    |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    | Serialization   |   Compression   |           Reserved                |
+    |   (4 bits)      |    (4 bits)     |           (8 bits)                |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                                       |
+    |                   Optional Header Extensions                          |
+    |                     (if Header Size > 1)                              |
+    |                                                                       |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                                       |
+    |                           Payload                                     |
+    |                      (variable length)                                |
+    |                                                                       |
+    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    """
+
+    version: VersionBits = VersionBits.Version1
+    header_size: HeaderSizeBits = HeaderSizeBits.HeaderSize4
+    type: MsgType = MsgType.Invalid
+    flag: MsgTypeFlagBits = MsgTypeFlagBits.NoSeq
+    serialization: SerializationBits = SerializationBits.JSON
+    compression: CompressionBits = CompressionBits.None_
+
+    event: EventType = EventType.None_
+    session_id: str = ""
+    connect_id: str = ""
+    sequence: int = 0
+    error_code: int = 0
+
+    payload: bytes = b""
+
+    @classmethod
+    def from_bytes(cls, data: bytes) -> "Message":
+        """Create message object from bytes"""
+        if len(data) < 3:
+            raise ValueError(
+                f"Data too short: expected at least 3 bytes, got {len(data)}"
+            )
+
+        type_and_flag = data[1]
+        msg_type = MsgType(type_and_flag >> 4)
+        flag = MsgTypeFlagBits(type_and_flag & 0b00001111)
+
+        msg = cls(type=msg_type, flag=flag)
+        msg.unmarshal(data)
+        return msg
+
+    def marshal(self) -> bytes:
+        """Serialize message to bytes"""
+        buffer = io.BytesIO()
+
+        # Write header
+        header = [
+            (self.version << 4) | self.header_size,
+            (self.type << 4) | self.flag,
+            (self.serialization << 4) | self.compression,
+        ]
+
+        header_size = 4 * self.header_size
+        if padding := header_size - len(header):
+            header.extend([0] * padding)
+
+        buffer.write(bytes(header))
+
+        # Write other fields
+        writers = self._get_writers()
+        for writer in writers:
+            writer(buffer)
+
+        return buffer.getvalue()
+
+    def unmarshal(self, data: bytes) -> None:
+        """Deserialize message from bytes"""
+        buffer = io.BytesIO(data)
+
+        # Read version and header size
+        version_and_header_size = buffer.read(1)[0]
+        self.version = VersionBits(version_and_header_size >> 4)
+        self.header_size = HeaderSizeBits(version_and_header_size & 0b00001111)
+
+        # Skip second byte
+        buffer.read(1)
+
+        # Read serialization and compression methods
+        serialization_compression = buffer.read(1)[0]
+        self.serialization = SerializationBits(serialization_compression >> 4)
+        self.compression = CompressionBits(serialization_compression & 0b00001111)
+
+        # Skip header padding
+        header_size = 4 * self.header_size
+        read_size = 3
+        if padding_size := header_size - read_size:
+            buffer.read(padding_size)
+
+        # Read other fields
+        readers = self._get_readers()
+        for reader in readers:
+            reader(buffer)
+
+        # Check for remaining data
+        remaining = buffer.read()
+        if remaining:
+            raise ValueError(f"Unexpected data after message: {remaining}")
+
+    def _get_writers(self) -> List[Callable[[io.BytesIO], None]]:
+        """Get list of writer functions"""
+        writers = []
+
+        if self.flag == MsgTypeFlagBits.WithEvent:
+            writers.extend([self._write_event, self._write_session_id])
+
+        if self.type in [
+            MsgType.FullClientRequest,
+            MsgType.FullServerResponse,
+            MsgType.FrontEndResultServer,
+            MsgType.AudioOnlyClient,
+            MsgType.AudioOnlyServer,
+        ]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                writers.append(self._write_sequence)
+        elif self.type == MsgType.Error:
+            writers.append(self._write_error_code)
+        else:
+            raise ValueError(f"Unsupported message type: {self.type}")
+
+        writers.append(self._write_payload)
+        return writers
+
+    def _get_readers(self) -> List[Callable[[io.BytesIO], None]]:
+        """Get list of reader functions"""
+        readers = []
+
+        if self.type in [
+            MsgType.FullClientRequest,
+            MsgType.FullServerResponse,
+            MsgType.FrontEndResultServer,
+            MsgType.AudioOnlyClient,
+            MsgType.AudioOnlyServer,
+        ]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                readers.append(self._read_sequence)
+        elif self.type == MsgType.Error:
+            readers.append(self._read_error_code)
+        else:
+            raise ValueError(f"Unsupported message type: {self.type}")
+
+        if self.flag == MsgTypeFlagBits.WithEvent:
+            readers.extend(
+                [self._read_event, self._read_session_id, self._read_connect_id]
+            )
+
+        readers.append(self._read_payload)
+        return readers
+
+    def _write_event(self, buffer: io.BytesIO) -> None:
+        """Write event"""
+        buffer.write(struct.pack(">i", self.event))
+
+    def _write_session_id(self, buffer: io.BytesIO) -> None:
+        """Write session ID"""
+        if self.event in [
+            EventType.StartConnection,
+            EventType.FinishConnection,
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+        ]:
+            return
+
+        session_id_bytes = self.session_id.encode("utf-8")
+        size = len(session_id_bytes)
+        if size > 0xFFFFFFFF:
+            raise ValueError(f"Session ID size ({size}) exceeds max(uint32)")
+
+        buffer.write(struct.pack(">I", size))
+        if size > 0:
+            buffer.write(session_id_bytes)
+
+    def _write_sequence(self, buffer: io.BytesIO) -> None:
+        """Write sequence number"""
+        buffer.write(struct.pack(">i", self.sequence))
+
+    def _write_error_code(self, buffer: io.BytesIO) -> None:
+        """Write error code"""
+        buffer.write(struct.pack(">I", self.error_code))
+
+    def _write_payload(self, buffer: io.BytesIO) -> None:
+        """Write payload"""
+        size = len(self.payload)
+        if size > 0xFFFFFFFF:
+            raise ValueError(f"Payload size ({size}) exceeds max(uint32)")
+
+        buffer.write(struct.pack(">I", size))
+        buffer.write(self.payload)
+
+    def _read_event(self, buffer: io.BytesIO) -> None:
+        """Read event"""
+        event_bytes = buffer.read(4)
+        if event_bytes:
+            self.event = EventType(struct.unpack(">i", event_bytes)[0])
+
+    def _read_session_id(self, buffer: io.BytesIO) -> None:
+        """Read session ID"""
+        if self.event in [
+            EventType.StartConnection,
+            EventType.FinishConnection,
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+            EventType.ConnectionFinished,
+        ]:
+            return
+
+        size_bytes = buffer.read(4)
+        if size_bytes:
+            size = struct.unpack(">I", size_bytes)[0]
+            if size > 0:
+                session_id_bytes = buffer.read(size)
+                if len(session_id_bytes) == size:
+                    self.session_id = session_id_bytes.decode("utf-8")
+
+    def _read_connect_id(self, buffer: io.BytesIO) -> None:
+        """Read connection ID"""
+        if self.event in [
+            EventType.ConnectionStarted,
+            EventType.ConnectionFailed,
+            EventType.ConnectionFinished,
+        ]:
+            size_bytes = buffer.read(4)
+            if size_bytes:
+                size = struct.unpack(">I", size_bytes)[0]
+                if size > 0:
+                    self.connect_id = buffer.read(size).decode("utf-8")
+
+    def _read_sequence(self, buffer: io.BytesIO) -> None:
+        """Read sequence number"""
+        sequence_bytes = buffer.read(4)
+        if sequence_bytes:
+            self.sequence = struct.unpack(">i", sequence_bytes)[0]
+
+    def _read_error_code(self, buffer: io.BytesIO) -> None:
+        """Read error code"""
+        error_code_bytes = buffer.read(4)
+        if error_code_bytes:
+            self.error_code = struct.unpack(">I", error_code_bytes)[0]
+
+    def _read_payload(self, buffer: io.BytesIO) -> None:
+        """Read payload"""
+        size_bytes = buffer.read(4)
+        if size_bytes:
+            size = struct.unpack(">I", size_bytes)[0]
+            if size > 0:
+                self.payload = buffer.read(size)
+
+    def __str__(self) -> str:
+        """String representation"""
+        if self.type in [MsgType.AudioOnlyServer, MsgType.AudioOnlyClient]:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                return f"MsgType: {self.type}, EventType:{self.event}, Sequence: {self.sequence}, PayloadSize: {len(self.payload)}"
+            return f"MsgType: {self.type}, EventType:{self.event}, PayloadSize: {len(self.payload)}"
+        elif self.type == MsgType.Error:
+            return f"MsgType: {self.type}, EventType:{self.event}, ErrorCode: {self.error_code}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+        else:
+            if self.flag in [MsgTypeFlagBits.PositiveSeq, MsgTypeFlagBits.NegativeSeq]:
+                return f"MsgType: {self.type}, EventType:{self.event}, Sequence: {self.sequence}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+            return f"MsgType: {self.type}, EventType:{self.event}, Payload: {self.payload.decode('utf-8', 'ignore')}"
+
+
+async def receive_message(websocket: websockets.WebSocketClientProtocol) -> Message:
+    """Receive message from websocket"""
+    try:
+        data = await websocket.recv()
+        if isinstance(data, str):
+            raise ValueError(f"Unexpected text message: {data}")
+        elif isinstance(data, bytes):
+            msg = Message.from_bytes(data)
+            logger.info(f"Received: {msg}")
+            return msg
+        else:
+            raise ValueError(f"Unexpected message type: {type(data)}")
+    except Exception as e:
+        logger.error(f"Failed to receive message: {e}")
+        raise
+
+
+async def wait_for_event(
+    websocket: websockets.WebSocketClientProtocol,
+    msg_type: MsgType,
+    event_type: EventType,
+) -> Message:
+    """Wait for specific event"""
+    while True:
+        msg = await receive_message(websocket)
+        if msg.type != msg_type or msg.event != event_type:
+            raise ValueError(f"Unexpected message: {msg}")
+        if msg.type == msg_type and msg.event == event_type:
+            return msg
+
+
+async def full_client_request(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes
+) -> None:
+    """Send full client message"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.NoSeq)
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def audio_only_client(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, flag: MsgTypeFlagBits
+) -> None:
+    """Send audio-only client message"""
+    msg = Message(type=MsgType.AudioOnlyClient, flag=flag)
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def start_connection(websocket: websockets.WebSocketClientProtocol) -> None:
+    """Start connection"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.StartConnection
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def finish_connection(websocket: websockets.WebSocketClientProtocol) -> None:
+    """Finish connection"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.FinishConnection
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def start_session(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, session_id: str
+) -> None:
+    """Start session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.StartSession
+    msg.session_id = session_id
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def finish_session(
+    websocket: websockets.WebSocketClientProtocol, session_id: str
+) -> None:
+    """Finish session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.FinishSession
+    msg.session_id = session_id
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def cancel_session(
+    websocket: websockets.WebSocketClientProtocol, session_id: str
+) -> None:
+    """Cancel session"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.CancelSession
+    msg.session_id = session_id
+    msg.payload = b"{}"
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())
+
+
+async def task_request(
+    websocket: websockets.WebSocketClientProtocol, payload: bytes, session_id: str
+) -> None:
+    """Send task request"""
+    msg = Message(type=MsgType.FullClientRequest, flag=MsgTypeFlagBits.WithEvent)
+    msg.event = EventType.TaskRequest
+    msg.session_id = session_id
+    msg.payload = payload
+    logger.info(f"Sending: {msg}")
+    await websocket.send(msg.marshal())

volcengine_bidirection_demo/pyproject.toml

@@ -0,0 +1,11 @@
+[build-system]
+requires = ["setuptools>=42", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "volc-speech-python-sdk"
+version = "0.1.0"
+requires-python = ">=3.9"
+dependencies = [
+    "websockets>=14.0",
+]

volcengine_bidirection_demo/setup.py

@@ -0,0 +1,13 @@
+from setuptools import find_packages, setup
+
+setup(
+    name="volc-speech-python-sdk",
+    version="0.1.0",
+    packages=find_packages(include=["protocols"]),
+    install_requires=[
+        "websockets>=14.0",
+    ],
+    python_requires=">=3.9",
+)
+
+python examples/volcengine/bidirection.py --appid 7069844318 --access_token osFMEJr20SSTWRql43cJlZkAOg7iwvxu --voice_type zh_female_tianxinxiaomei_emo_v2_mars_bigtts --text "你好，我是火山引擎的语音合成服务。这是一个美好的旅程。"

volcengine_bidirection_demo/volc_speech_python_sdk.egg-info/PKG-INFO

@@ -0,0 +1,6 @@
+Metadata-Version: 2.4
+Name: volc-speech-python-sdk
+Version: 0.1.0
+Requires-Python: >=3.9
+Requires-Dist: websockets>=14.0
+Dynamic: requires-python

volcengine_bidirection_demo/volc_speech_python_sdk.egg-info/SOURCES.txt

@@ -0,0 +1,9 @@
+pyproject.toml
+setup.py
+protocols/__init__.py
+protocols/protocols.py
+volc_speech_python_sdk.egg-info/PKG-INFO
+volc_speech_python_sdk.egg-info/SOURCES.txt
+volc_speech_python_sdk.egg-info/dependency_links.txt
+volc_speech_python_sdk.egg-info/requires.txt
+volc_speech_python_sdk.egg-info/top_level.txt

volcengine_bidirection_demo/volc_speech_python_sdk.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

volcengine_bidirection_demo/volc_speech_python_sdk.egg-info/requires.txt

@@ -0,0 +1 @@
+websockets>=14.0

volcengine_bidirection_demo/volc_speech_python_sdk.egg-info/top_level.txt

@@ -0,0 +1 @@
+protocols

测试流式传输uniapp/node_modules/.package-lock.json

@@ -0,0 +1,20 @@
+{
+  "name": "测试流式传输uniapp",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "node_modules/@types/pako": {
+      "version": "2.0.4",
+      "resolved": "https://registry.npmmirror.com/@types/pako/-/pako-2.0.4.tgz",
+      "integrity": "sha512-VWDCbrLeVXJM9fihYodcLiIv0ku+AlOa/TQ1SvYOaBuyrSKgEcro95LJyIsJ4vSo6BXIxOKxiJAat04CmST9Fw==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/pako": {
+      "version": "2.1.0",
+      "resolved": "https://registry.npmmirror.com/pako/-/pako-2.1.0.tgz",
+      "integrity": "sha512-w+eufiZ1WuJYgPXbV/PO3NCMEc3xqylkKHzp8bxp1uW4qaSNQUkwmLLEc3kKsfz8lpV1F8Ht3U1Cm+9Srog2ug==",
+      "license": "(MIT AND Zlib)"
+    }
+  }
+}

测试流式传输uniapp/node_modules/@types/pako/LICENSE

@@ -0,0 +1,21 @@
+    MIT License
+
+    Copyright (c) Microsoft Corporation.
+
+    Permission is hereby granted, free of charge, to any person obtaining a copy
+    of this software and associated documentation files (the "Software"), to deal
+    in the Software without restriction, including without limitation the rights
+    to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+    copies of the Software, and to permit persons to whom the Software is
+    furnished to do so, subject to the following conditions:
+
+    The above copyright notice and this permission notice shall be included in all
+    copies or substantial portions of the Software.
+
+    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+    SOFTWARE

测试流式传输uniapp/node_modules/@types/pako/README.md

@@ -0,0 +1,15 @@
+# Installation
+> `npm install --save @types/pako`
+
+# Summary
+This package contains type definitions for pako (https://github.com/nodeca/pako).
+
+# Details
+Files were exported from https://github.com/DefinitelyTyped/DefinitelyTyped/tree/master/types/pako.
+
+### Additional Details
+ * Last updated: Fri, 15 Aug 2025 08:39:32 GMT
+ * Dependencies: none
+
+# Credits
+These definitions were written by [Caleb Eggensperger](https://github.com/calebegg), [Muhammet Öztürk](https://github.com/hlthi), and [Thibault Poisson](https://github.com/OrIOg).

测试流式传输uniapp/node_modules/@types/pako/index.d.ts

@@ -0,0 +1,161 @@
+export = Pako;
+export as namespace pako;
+
+declare namespace Pako {
+    enum constants {
+        // FlushValues
+        Z_NO_FLUSH = 0,
+        Z_PARTIAL_FLUSH = 1,
+        Z_SYNC_FLUSH = 2,
+        Z_FULL_FLUSH = 3,
+        Z_FINISH = 4,
+        Z_BLOCK = 5,
+        Z_TREES = 6,
+        // StrategyValues
+        Z_FILTERED = 1,
+        Z_HUFFMAN_ONLY = 2,
+        Z_RLE = 3,
+        Z_FIXED = 4,
+        Z_DEFAULT_STRATEGY = 0,
+        // ReturnCodes
+        Z_OK = 0,
+        Z_STREAM_END = 1,
+        Z_NEED_DICT = 2,
+        Z_ERRNO = -1,
+        Z_STREAM_ERROR = -2,
+        Z_DATA_ERROR = -3,
+        Z_BUF_ERROR = -5,
+    }
+
+    type FlushValues =
+        | constants.Z_NO_FLUSH
+        | constants.Z_PARTIAL_FLUSH
+        | constants.Z_SYNC_FLUSH
+        | constants.Z_FINISH
+        | constants.Z_BLOCK
+        | constants.Z_TREES;
+
+    type StrategyValues =
+        | constants.Z_FILTERED
+        | constants.Z_HUFFMAN_ONLY
+        | constants.Z_RLE
+        | constants.Z_FIXED
+        | constants.Z_DEFAULT_STRATEGY;
+
+    type ReturnCodes =
+        | constants.Z_OK
+        | constants.Z_STREAM_END
+        | constants.Z_NEED_DICT
+        | constants.Z_ERRNO
+        | constants.Z_STREAM_ERROR
+        | constants.Z_DATA_ERROR
+        | constants.Z_BUF_ERROR
+        | constants.Z_DEFAULT_STRATEGY;
+
+    interface DeflateOptions {
+        level?: -1 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | undefined;
+        windowBits?: number | undefined;
+        memLevel?: number | undefined;
+        strategy?: StrategyValues | undefined;
+        dictionary?: any;
+        raw?: boolean | undefined;
+        chunkSize?: number | undefined;
+        gzip?: boolean | undefined;
+        header?: Header | undefined;
+    }
+
+    interface DeflateFunctionOptions {
+        level?: -1 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | undefined;
+        windowBits?: number | undefined;
+        memLevel?: number | undefined;
+        strategy?: StrategyValues | undefined;
+        dictionary?: any;
+        raw?: boolean | undefined;
+    }
+
+    interface InflateOptions {
+        windowBits?: number | undefined;
+        dictionary?: any;
+        raw?: boolean | undefined;
+        to?: "string" | undefined;
+        chunkSize?: number | undefined;
+    }
+
+    interface InflateFunctionOptions {
+        windowBits?: number | undefined;
+        raw?: boolean | undefined;
+        to?: "string" | undefined;
+    }
+
+    interface Header {
+        text?: boolean | undefined;
+        time?: number | undefined;
+        os?: number | undefined;
+        extra?: number[] | undefined;
+        name?: string | undefined;
+        comment?: string | undefined;
+        hcrc?: boolean | undefined;
+    }
+
+    type Data = Uint8Array | ArrayBuffer;
+
+    // For TS <=5.6 compatibility: Uint8Array<ArrayBuffer> in TS >=5.7, Uint8Array in TS <=5.6
+    type Uint8ArrayReturnType = InstanceType<typeof Uint8Array>;
+
+    /**
+     * Compress data with deflate algorithm and options.
+     */
+    function deflate(data: Data | string, options?: DeflateFunctionOptions): Uint8ArrayReturnType;
+
+    /**
+     * The same as deflate, but creates raw data, without wrapper (header and adler32 crc).
+     */
+    function deflateRaw(data: Data | string, options?: DeflateFunctionOptions): Uint8ArrayReturnType;
+
+    /**
+     * The same as deflate, but create gzip wrapper instead of deflate one.
+     */
+    function gzip(data: Data | string, options?: DeflateFunctionOptions): Uint8ArrayReturnType;
+
+    /**
+     * Decompress data with inflate/ungzip and options. Autodetect format via wrapper header
+     * by default. That's why we don't provide separate ungzip method.
+     */
+    function inflate(data: Data, options: InflateFunctionOptions & { to: "string" }): string;
+    function inflate(data: Data, options?: InflateFunctionOptions): Uint8ArrayReturnType;
+
+    /**
+     * The same as inflate, but creates raw data, without wrapper (header and adler32 crc).
+     */
+    function inflateRaw(data: Data, options: InflateFunctionOptions & { to: "string" }): string;
+    function inflateRaw(data: Data, options?: InflateFunctionOptions): Uint8ArrayReturnType;
+
+    /**
+     * Just shortcut to inflate, because it autodetects format by header.content. Done for convenience.
+     */
+    function ungzip(data: Data, options: InflateFunctionOptions & { to: "string" }): string;
+    function ungzip(data: Data, options?: InflateFunctionOptions): Uint8ArrayReturnType;
+
+    // https://github.com/nodeca/pako/blob/893381abcafa10fa2081ce60dae7d4d8e873a658/lib/deflate.js
+    class Deflate {
+        constructor(options?: DeflateOptions);
+        err: ReturnCodes;
+        msg: string;
+        result: Uint8ArrayReturnType;
+        onData(chunk: Data): void;
+        onEnd(status: number): void;
+        push(data: Data | string, mode?: FlushValues | boolean): boolean;
+    }
+
+    // https://github.com/nodeca/pako/blob/893381abcafa10fa2081ce60dae7d4d8e873a658/lib/inflate.js
+    class Inflate {
+        constructor(options?: InflateOptions);
+        header?: Header | undefined;
+        err: ReturnCodes;
+        msg: string;
+        result: Uint8ArrayReturnType | string;
+        onData(chunk: Data): void;
+        onEnd(status: number): void;
+        push(data: Data, mode?: FlushValues | boolean): boolean;
+    }
+}

测试流式传输uniapp/node_modules/@types/pako/package.json

@@ -0,0 +1,36 @@
+{
+    "name": "@types/pako",
+    "version": "2.0.4",
+    "description": "TypeScript definitions for pako",
+    "homepage": "https://github.com/DefinitelyTyped/DefinitelyTyped/tree/master/types/pako",
+    "license": "MIT",
+    "contributors": [
+        {
+            "name": "Caleb Eggensperger",
+            "githubUsername": "calebegg",
+            "url": "https://github.com/calebegg"
+        },
+        {
+            "name": "Muhammet Öztürk",
+            "githubUsername": "hlthi",
+            "url": "https://github.com/hlthi"
+        },
+        {
+            "name": "Thibault Poisson",
+            "githubUsername": "OrIOg",
+            "url": "https://github.com/OrIOg"
+        }
+    ],
+    "main": "",
+    "types": "index.d.ts",
+    "repository": {
+        "type": "git",
+        "url": "https://github.com/DefinitelyTyped/DefinitelyTyped.git",
+        "directory": "types/pako"
+    },
+    "scripts": {},
+    "dependencies": {},
+    "peerDependencies": {},
+    "typesPublisherContentHash": "1aa5b0407205ecaf32d7e7dfdb7cd3a7a87fad829f92f3c51eea48dc08742b89",
+    "typeScriptVersion": "5.2"
+}

测试流式传输uniapp/node_modules/pako/LICENSE

@@ -0,0 +1,21 @@
+(The MIT License)
+
+Copyright (C) 2014-2017 by Vitaly Puzrin and Andrei Tuputcyn
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

测试流式传输uniapp/node_modules/pako/README.md

@@ -0,0 +1,177 @@
+pako
+==========================================
+
+[![CI](https://github.com/nodeca/pako/workflows/CI/badge.svg)](https://github.com/nodeca/pako/actions)
+[![NPM version](https://img.shields.io/npm/v/pako.svg)](https://www.npmjs.org/package/pako)
+
+> zlib port to javascript, very fast!
+
+__Why pako is cool:__
+
+- Results are binary equal to well known [zlib](http://www.zlib.net/) (now contains ported zlib v1.2.8).
+- Almost as fast in modern JS engines as C implementation (see benchmarks).
+- Works in browsers, you can browserify any separate component.
+
+This project was done to understand how fast JS can be and is it necessary to
+develop native C modules for CPU-intensive tasks. Enjoy the result!
+
+
+__Benchmarks:__
+
+
+node v12.16.3 (zlib 1.2.9), 1mb input sample:
+
+```
+deflate-imaya x 4.75 ops/sec ±4.93% (15 runs sampled)
+deflate-pako x 10.38 ops/sec ±0.37% (29 runs sampled)
+deflate-zlib x 17.74 ops/sec ±0.77% (46 runs sampled)
+gzip-pako x 8.86 ops/sec ±1.41% (29 runs sampled)
+inflate-imaya x 107 ops/sec ±0.69% (77 runs sampled)
+inflate-pako x 131 ops/sec ±1.74% (82 runs sampled)
+inflate-zlib x 258 ops/sec ±0.66% (88 runs sampled)
+ungzip-pako x 115 ops/sec ±1.92% (80 runs sampled)
+```
+
+node v14.15.0 (google's zlib), 1mb output sample:
+
+```
+deflate-imaya x 4.93 ops/sec ±3.09% (16 runs sampled)
+deflate-pako x 10.22 ops/sec ±0.33% (29 runs sampled)
+deflate-zlib x 18.48 ops/sec ±0.24% (48 runs sampled)
+gzip-pako x 10.16 ops/sec ±0.25% (28 runs sampled)
+inflate-imaya x 110 ops/sec ±0.41% (77 runs sampled)
+inflate-pako x 134 ops/sec ±0.66% (83 runs sampled)
+inflate-zlib x 402 ops/sec ±0.74% (87 runs sampled)
+ungzip-pako x 113 ops/sec ±0.62% (80 runs sampled)
+```
+
+zlib's test is partially affected by marshalling (that make sense for inflate only).
+You can change deflate level to 0 in benchmark source, to investigate details.
+For deflate level 6 results can be considered as correct.
+
+__Install:__
+
+```
+npm install pako
+```
+
+
+Examples / API
+--------------
+
+Full docs - http://nodeca.github.io/pako/
+
+```javascript
+const pako = require('pako');
+
+// Deflate
+//
+const input = new Uint8Array();
+//... fill input data here
+const output = pako.deflate(input);
+
+// Inflate (simple wrapper can throw exception on broken stream)
+//
+const compressed = new Uint8Array();
+//... fill data to uncompress here
+try {
+  const result = pako.inflate(compressed);
+  // ... continue processing
+} catch (err) {
+  console.log(err);
+}
+
+//
+// Alternate interface for chunking & without exceptions
+//
+
+const deflator = new pako.Deflate();
+
+deflator.push(chunk1, false);
+deflator.push(chunk2); // second param is false by default.
+...
+deflator.push(chunk_last, true); // `true` says this chunk is last
+
+if (deflator.err) {
+  console.log(deflator.msg);
+}
+
+const output = deflator.result;
+
+
+const inflator = new pako.Inflate();
+
+inflator.push(chunk1);
+inflator.push(chunk2);
+...
+inflator.push(chunk_last); // no second param because end is auto-detected
+
+if (inflator.err) {
+  console.log(inflator.msg);
+}
+
+const output = inflator.result;
+```
+
+Sometime you can wish to work with strings. For example, to send
+stringified objects to server. Pako's deflate detects input data type, and
+automatically recode strings to utf-8 prior to compress. Inflate has special
+option, to say compressed data has utf-8 encoding and should be recoded to
+javascript's utf-16.
+
+```javascript
+const pako = require('pako');
+
+const test = { my: 'super', puper: [456, 567], awesome: 'pako' };
+
+const compressed = pako.deflate(JSON.stringify(test));
+
+const restored = JSON.parse(pako.inflate(compressed, { to: 'string' }));
+```
+
+
+Notes
+-----
+
+Pako does not contain some specific zlib functions:
+
+- __deflate__ -  methods `deflateCopy`, `deflateBound`, `deflateParams`,
+  `deflatePending`, `deflatePrime`, `deflateTune`.
+- __inflate__ - methods `inflateCopy`, `inflateMark`,
+  `inflatePrime`, `inflateGetDictionary`, `inflateSync`, `inflateSyncPoint`, `inflateUndermine`.
+- High level inflate/deflate wrappers (classes) may not support some flush
+  modes.
+
+
+pako for enterprise
+-------------------
+
+Available as part of the Tidelift Subscription
+
+The maintainers of pako and thousands of other packages are working with Tidelift to deliver commercial support and maintenance for the open source dependencies you use to build your applications. Save time, reduce risk, and improve code health, while paying the maintainers of the exact dependencies you use. [Learn more.](https://tidelift.com/subscription/pkg/npm-pako?utm_source=npm-pako&utm_medium=referral&utm_campaign=enterprise&utm_term=repo)
+
+
+Authors
+-------
+
+- Andrey Tupitsin [@anrd83](https://github.com/andr83)
+- Vitaly Puzrin [@puzrin](https://github.com/puzrin)
+
+Personal thanks to:
+
+- Vyacheslav Egorov ([@mraleph](https://github.com/mraleph)) for his awesome
+  tutorials about optimising JS code for v8, [IRHydra](http://mrale.ph/irhydra/)
+  tool and his advices.
+- David Duponchel ([@dduponchel](https://github.com/dduponchel)) for help with
+  testing.
+
+Original implementation (in C):
+
+- [zlib](http://zlib.net/) by Jean-loup Gailly and Mark Adler.
+
+
+License
+-------
+
+- MIT - all files, except `/lib/zlib` folder
+- ZLIB - `/lib/zlib` content

测试流式传输uniapp/node_modules/pako/index.js

@@ -0,0 +1,18 @@
+// Top level file is just a mixin of submodules & constants
+'use strict';
+
+const { Deflate, deflate, deflateRaw, gzip } = require('./lib/deflate');
+
+const { Inflate, inflate, inflateRaw, ungzip } = require('./lib/inflate');
+
+const constants = require('./lib/zlib/constants');
+
+module.exports.Deflate = Deflate;
+module.exports.deflate = deflate;
+module.exports.deflateRaw = deflateRaw;
+module.exports.gzip = gzip;
+module.exports.Inflate = Inflate;
+module.exports.inflate = inflate;
+module.exports.inflateRaw = inflateRaw;
+module.exports.ungzip = ungzip;
+module.exports.constants = constants;

测试流式传输uniapp/node_modules/pako/lib/deflate.js

@@ -0,0 +1,380 @@
+'use strict';
+
+
+const zlib_deflate = require('./zlib/deflate');
+const utils        = require('./utils/common');
+const strings      = require('./utils/strings');
+const msg          = require('./zlib/messages');
+const ZStream      = require('./zlib/zstream');
+
+const toString = Object.prototype.toString;
+
+/* Public constants ==========================================================*/
+/* ===========================================================================*/
+
+const {
+  Z_NO_FLUSH, Z_SYNC_FLUSH, Z_FULL_FLUSH, Z_FINISH,
+  Z_OK, Z_STREAM_END,
+  Z_DEFAULT_COMPRESSION,
+  Z_DEFAULT_STRATEGY,
+  Z_DEFLATED
+} = require('./zlib/constants');
+
+/* ===========================================================================*/
+
+
+/**
+ * class Deflate
+ *
+ * Generic JS-style wrapper for zlib calls. If you don't need
+ * streaming behaviour - use more simple functions: [[deflate]],
+ * [[deflateRaw]] and [[gzip]].
+ **/
+
+/* internal
+ * Deflate.chunks -> Array
+ *
+ * Chunks of output data, if [[Deflate#onData]] not overridden.
+ **/
+
+/**
+ * Deflate.result -> Uint8Array
+ *
+ * Compressed result, generated by default [[Deflate#onData]]
+ * and [[Deflate#onEnd]] handlers. Filled after you push last chunk
+ * (call [[Deflate#push]] with `Z_FINISH` / `true` param).
+ **/
+
+/**
+ * Deflate.err -> Number
+ *
+ * Error code after deflate finished. 0 (Z_OK) on success.
+ * You will not need it in real life, because deflate errors
+ * are possible only on wrong options or bad `onData` / `onEnd`
+ * custom handlers.
+ **/
+
+/**
+ * Deflate.msg -> String
+ *
+ * Error message, if [[Deflate.err]] != 0
+ **/
+
+
+/**
+ * new Deflate(options)
+ * - options (Object): zlib deflate options.
+ *
+ * Creates new deflator instance with specified params. Throws exception
+ * on bad params. Supported options:
+ *
+ * - `level`
+ * - `windowBits`
+ * - `memLevel`
+ * - `strategy`
+ * - `dictionary`
+ *
+ * [http://zlib.net/manual.html#Advanced](http://zlib.net/manual.html#Advanced)
+ * for more information on these.
+ *
+ * Additional options, for internal needs:
+ *
+ * - `chunkSize` - size of generated data chunks (16K by default)
+ * - `raw` (Boolean) - do raw deflate
+ * - `gzip` (Boolean) - create gzip wrapper
+ * - `header` (Object) - custom header for gzip
+ *   - `text` (Boolean) - true if compressed data believed to be text
+ *   - `time` (Number) - modification time, unix timestamp
+ *   - `os` (Number) - operation system code
+ *   - `extra` (Array) - array of bytes with extra data (max 65536)
+ *   - `name` (String) - file name (binary string)
+ *   - `comment` (String) - comment (binary string)
+ *   - `hcrc` (Boolean) - true if header crc should be added
+ *
+ * ##### Example:
+ *
+ * ```javascript
+ * const pako = require('pako')
+ *   , chunk1 = new Uint8Array([1,2,3,4,5,6,7,8,9])
+ *   , chunk2 = new Uint8Array([10,11,12,13,14,15,16,17,18,19]);
+ *
+ * const deflate = new pako.Deflate({ level: 3});
+ *
+ * deflate.push(chunk1, false);
+ * deflate.push(chunk2, true);  // true -> last chunk
+ *
+ * if (deflate.err) { throw new Error(deflate.err); }
+ *
+ * console.log(deflate.result);
+ * ```
+ **/
+function Deflate(options) {
+  this.options = utils.assign({
+    level: Z_DEFAULT_COMPRESSION,
+    method: Z_DEFLATED,
+    chunkSize: 16384,
+    windowBits: 15,
+    memLevel: 8,
+    strategy: Z_DEFAULT_STRATEGY
+  }, options || {});
+
+  let opt = this.options;
+
+  if (opt.raw && (opt.windowBits > 0)) {
+    opt.windowBits = -opt.windowBits;
+  }
+
+  else if (opt.gzip && (opt.windowBits > 0) && (opt.windowBits < 16)) {
+    opt.windowBits += 16;
+  }
+
+  this.err    = 0;      // error code, if happens (0 = Z_OK)
+  this.msg    = '';     // error message
+  this.ended  = false;  // used to avoid multiple onEnd() calls
+  this.chunks = [];     // chunks of compressed data
+
+  this.strm = new ZStream();
+  this.strm.avail_out = 0;
+
+  let status = zlib_deflate.deflateInit2(
+    this.strm,
+    opt.level,
+    opt.method,
+    opt.windowBits,
+    opt.memLevel,
+    opt.strategy
+  );
+
+  if (status !== Z_OK) {
+    throw new Error(msg[status]);
+  }
+
+  if (opt.header) {
+    zlib_deflate.deflateSetHeader(this.strm, opt.header);
+  }
+
+  if (opt.dictionary) {
+    let dict;
+    // Convert data if needed
+    if (typeof opt.dictionary === 'string') {
+      // If we need to compress text, change encoding to utf8.
+      dict = strings.string2buf(opt.dictionary);
+    } else if (toString.call(opt.dictionary) === '[object ArrayBuffer]') {
+      dict = new Uint8Array(opt.dictionary);
+    } else {
+      dict = opt.dictionary;
+    }
+
+    status = zlib_deflate.deflateSetDictionary(this.strm, dict);
+
+    if (status !== Z_OK) {
+      throw new Error(msg[status]);
+    }
+
+    this._dict_set = true;
+  }
+}
+
+/**
+ * Deflate#push(data[, flush_mode]) -> Boolean
+ * - data (Uint8Array|ArrayBuffer|String): input data. Strings will be
+ *   converted to utf8 byte sequence.
+ * - flush_mode (Number|Boolean): 0..6 for corresponding Z_NO_FLUSH..Z_TREE modes.
+ *   See constants. Skipped or `false` means Z_NO_FLUSH, `true` means Z_FINISH.
+ *
+ * Sends input data to deflate pipe, generating [[Deflate#onData]] calls with
+ * new compressed chunks. Returns `true` on success. The last data block must
+ * have `flush_mode` Z_FINISH (or `true`). That will flush internal pending
+ * buffers and call [[Deflate#onEnd]].
+ *
+ * On fail call [[Deflate#onEnd]] with error code and return false.
+ *
+ * ##### Example
+ *
+ * ```javascript
+ * push(chunk, false); // push one of data chunks
+ * ...
+ * push(chunk, true);  // push last chunk
+ * ```
+ **/
+Deflate.prototype.push = function (data, flush_mode) {
+  const strm = this.strm;
+  const chunkSize = this.options.chunkSize;
+  let status, _flush_mode;
+
+  if (this.ended) { return false; }
+
+  if (flush_mode === ~~flush_mode) _flush_mode = flush_mode;
+  else _flush_mode = flush_mode === true ? Z_FINISH : Z_NO_FLUSH;
+
+  // Convert data if needed
+  if (typeof data === 'string') {
+    // If we need to compress text, change encoding to utf8.
+    strm.input = strings.string2buf(data);
+  } else if (toString.call(data) === '[object ArrayBuffer]') {
+    strm.input = new Uint8Array(data);
+  } else {
+    strm.input = data;
+  }
+
+  strm.next_in = 0;
+  strm.avail_in = strm.input.length;
+
+  for (;;) {
+    if (strm.avail_out === 0) {
+      strm.output = new Uint8Array(chunkSize);
+      strm.next_out = 0;
+      strm.avail_out = chunkSize;
+    }
+
+    // Make sure avail_out > 6 to avoid repeating markers
+    if ((_flush_mode === Z_SYNC_FLUSH || _flush_mode === Z_FULL_FLUSH) && strm.avail_out <= 6) {
+      this.onData(strm.output.subarray(0, strm.next_out));
+      strm.avail_out = 0;
+      continue;
+    }
+
+    status = zlib_deflate.deflate(strm, _flush_mode);
+
+    // Ended => flush and finish
+    if (status === Z_STREAM_END) {
+      if (strm.next_out > 0) {
+        this.onData(strm.output.subarray(0, strm.next_out));
+      }
+      status = zlib_deflate.deflateEnd(this.strm);
+      this.onEnd(status);
+      this.ended = true;
+      return status === Z_OK;
+    }
+
+    // Flush if out buffer full
+    if (strm.avail_out === 0) {
+      this.onData(strm.output);
+      continue;
+    }
+
+    // Flush if requested and has data
+    if (_flush_mode > 0 && strm.next_out > 0) {
+      this.onData(strm.output.subarray(0, strm.next_out));
+      strm.avail_out = 0;
+      continue;
+    }
+
+    if (strm.avail_in === 0) break;
+  }
+
+  return true;
+};
+
+
+/**
+ * Deflate#onData(chunk) -> Void
+ * - chunk (Uint8Array): output data.
+ *
+ * By default, stores data blocks in `chunks[]` property and glue
+ * those in `onEnd`. Override this handler, if you need another behaviour.
+ **/
+Deflate.prototype.onData = function (chunk) {
+  this.chunks.push(chunk);
+};
+
+
+/**
+ * Deflate#onEnd(status) -> Void
+ * - status (Number): deflate status. 0 (Z_OK) on success,
+ *   other if not.
+ *
+ * Called once after you tell deflate that the input stream is
+ * complete (Z_FINISH). By default - join collected chunks,
+ * free memory and fill `results` / `err` properties.
+ **/
+Deflate.prototype.onEnd = function (status) {
+  // On success - join
+  if (status === Z_OK) {
+    this.result = utils.flattenChunks(this.chunks);
+  }
+  this.chunks = [];
+  this.err = status;
+  this.msg = this.strm.msg;
+};
+
+
+/**
+ * deflate(data[, options]) -> Uint8Array
+ * - data (Uint8Array|ArrayBuffer|String): input data to compress.
+ * - options (Object): zlib deflate options.
+ *
+ * Compress `data` with deflate algorithm and `options`.
+ *
+ * Supported options are:
+ *
+ * - level
+ * - windowBits
+ * - memLevel
+ * - strategy
+ * - dictionary
+ *
+ * [http://zlib.net/manual.html#Advanced](http://zlib.net/manual.html#Advanced)
+ * for more information on these.
+ *
+ * Sugar (options):
+ *
+ * - `raw` (Boolean) - say that we work with raw stream, if you don't wish to specify
+ *   negative windowBits implicitly.
+ *
+ * ##### Example:
+ *
+ * ```javascript
+ * const pako = require('pako')
+ * const data = new Uint8Array([1,2,3,4,5,6,7,8,9]);
+ *
+ * console.log(pako.deflate(data));
+ * ```
+ **/
+function deflate(input, options) {
+  const deflator = new Deflate(options);
+
+  deflator.push(input, true);
+
+  // That will never happens, if you don't cheat with options :)
+  if (deflator.err) { throw deflator.msg || msg[deflator.err]; }
+
+  return deflator.result;
+}
+
+
+/**
+ * deflateRaw(data[, options]) -> Uint8Array
+ * - data (Uint8Array|ArrayBuffer|String): input data to compress.
+ * - options (Object): zlib deflate options.
+ *
+ * The same as [[deflate]], but creates raw data, without wrapper
+ * (header and adler32 crc).
+ **/
+function deflateRaw(input, options) {
+  options = options || {};
+  options.raw = true;
+  return deflate(input, options);
+}
+
+
+/**
+ * gzip(data[, options]) -> Uint8Array
+ * - data (Uint8Array|ArrayBuffer|String): input data to compress.
+ * - options (Object): zlib deflate options.
+ *
+ * The same as [[deflate]], but create gzip wrapper instead of
+ * deflate one.
+ **/
+function gzip(input, options) {
+  options = options || {};
+  options.gzip = true;
+  return deflate(input, options);
+}
+
+
+module.exports.Deflate = Deflate;
+module.exports.deflate = deflate;
+module.exports.deflateRaw = deflateRaw;
+module.exports.gzip = gzip;
+module.exports.constants = require('./zlib/constants');

测试流式传输uniapp/node_modules/pako/lib/inflate.js

@@ -0,0 +1,419 @@
+'use strict';
+
+
+const zlib_inflate = require('./zlib/inflate');
+const utils        = require('./utils/common');
+const strings      = require('./utils/strings');
+const msg          = require('./zlib/messages');
+const ZStream      = require('./zlib/zstream');
+const GZheader     = require('./zlib/gzheader');
+
+const toString = Object.prototype.toString;
+
+/* Public constants ==========================================================*/
+/* ===========================================================================*/
+
+const {
+  Z_NO_FLUSH, Z_FINISH,
+  Z_OK, Z_STREAM_END, Z_NEED_DICT, Z_STREAM_ERROR, Z_DATA_ERROR, Z_MEM_ERROR
+} = require('./zlib/constants');
+
+/* ===========================================================================*/
+
+
+/**
+ * class Inflate
+ *
+ * Generic JS-style wrapper for zlib calls. If you don't need
+ * streaming behaviour - use more simple functions: [[inflate]]
+ * and [[inflateRaw]].
+ **/
+
+/* internal
+ * inflate.chunks -> Array
+ *
+ * Chunks of output data, if [[Inflate#onData]] not overridden.
+ **/
+
+/**
+ * Inflate.result -> Uint8Array|String
+ *
+ * Uncompressed result, generated by default [[Inflate#onData]]
+ * and [[Inflate#onEnd]] handlers. Filled after you push last chunk
+ * (call [[Inflate#push]] with `Z_FINISH` / `true` param).
+ **/
+
+/**
+ * Inflate.err -> Number
+ *
+ * Error code after inflate finished. 0 (Z_OK) on success.
+ * Should be checked if broken data possible.
+ **/
+
+/**
+ * Inflate.msg -> String
+ *
+ * Error message, if [[Inflate.err]] != 0
+ **/
+
+
+/**
+ * new Inflate(options)
+ * - options (Object): zlib inflate options.
+ *
+ * Creates new inflator instance with specified params. Throws exception
+ * on bad params. Supported options:
+ *
+ * - `windowBits`
+ * - `dictionary`
+ *
+ * [http://zlib.net/manual.html#Advanced](http://zlib.net/manual.html#Advanced)
+ * for more information on these.
+ *
+ * Additional options, for internal needs:
+ *
+ * - `chunkSize` - size of generated data chunks (16K by default)
+ * - `raw` (Boolean) - do raw inflate
+ * - `to` (String) - if equal to 'string', then result will be converted
+ *   from utf8 to utf16 (javascript) string. When string output requested,
+ *   chunk length can differ from `chunkSize`, depending on content.
+ *
+ * By default, when no options set, autodetect deflate/gzip data format via
+ * wrapper header.
+ *
+ * ##### Example:
+ *
+ * ```javascript
+ * const pako = require('pako')
+ * const chunk1 = new Uint8Array([1,2,3,4,5,6,7,8,9])
+ * const chunk2 = new Uint8Array([10,11,12,13,14,15,16,17,18,19]);
+ *
+ * const inflate = new pako.Inflate({ level: 3});
+ *
+ * inflate.push(chunk1, false);
+ * inflate.push(chunk2, true);  // true -> last chunk
+ *
+ * if (inflate.err) { throw new Error(inflate.err); }
+ *
+ * console.log(inflate.result);
+ * ```
+ **/
+function Inflate(options) {
+  this.options = utils.assign({
+    chunkSize: 1024 * 64,
+    windowBits: 15,
+    to: ''
+  }, options || {});
+
+  const opt = this.options;
+
+  // Force window size for `raw` data, if not set directly,
+  // because we have no header for autodetect.
+  if (opt.raw && (opt.windowBits >= 0) && (opt.windowBits < 16)) {
+    opt.windowBits = -opt.windowBits;
+    if (opt.windowBits === 0) { opt.windowBits = -15; }
+  }
+
+  // If `windowBits` not defined (and mode not raw) - set autodetect flag for gzip/deflate
+  if ((opt.windowBits >= 0) && (opt.windowBits < 16) &&
+      !(options && options.windowBits)) {
+    opt.windowBits += 32;
+  }
+
+  // Gzip header has no info about windows size, we can do autodetect only
+  // for deflate. So, if window size not set, force it to max when gzip possible
+  if ((opt.windowBits > 15) && (opt.windowBits < 48)) {
+    // bit 3 (16) -> gzipped data
+    // bit 4 (32) -> autodetect gzip/deflate
+    if ((opt.windowBits & 15) === 0) {
+      opt.windowBits |= 15;
+    }
+  }
+
+  this.err    = 0;      // error code, if happens (0 = Z_OK)
+  this.msg    = '';     // error message
+  this.ended  = false;  // used to avoid multiple onEnd() calls
+  this.chunks = [];     // chunks of compressed data
+
+  this.strm   = new ZStream();
+  this.strm.avail_out = 0;
+
+  let status  = zlib_inflate.inflateInit2(
+    this.strm,
+    opt.windowBits
+  );
+
+  if (status !== Z_OK) {
+    throw new Error(msg[status]);
+  }
+
+  this.header = new GZheader();
+
+  zlib_inflate.inflateGetHeader(this.strm, this.header);
+
+  // Setup dictionary
+  if (opt.dictionary) {
+    // Convert data if needed
+    if (typeof opt.dictionary === 'string') {
+      opt.dictionary = strings.string2buf(opt.dictionary);
+    } else if (toString.call(opt.dictionary) === '[object ArrayBuffer]') {
+      opt.dictionary = new Uint8Array(opt.dictionary);
+    }
+    if (opt.raw) { //In raw mode we need to set the dictionary early
+      status = zlib_inflate.inflateSetDictionary(this.strm, opt.dictionary);
+      if (status !== Z_OK) {
+        throw new Error(msg[status]);
+      }
+    }
+  }
+}
+
+/**
+ * Inflate#push(data[, flush_mode]) -> Boolean
+ * - data (Uint8Array|ArrayBuffer): input data
+ * - flush_mode (Number|Boolean): 0..6 for corresponding Z_NO_FLUSH..Z_TREE
+ *   flush modes. See constants. Skipped or `false` means Z_NO_FLUSH,
+ *   `true` means Z_FINISH.
+ *
+ * Sends input data to inflate pipe, generating [[Inflate#onData]] calls with
+ * new output chunks. Returns `true` on success. If end of stream detected,
+ * [[Inflate#onEnd]] will be called.
+ *
+ * `flush_mode` is not needed for normal operation, because end of stream
+ * detected automatically. You may try to use it for advanced things, but
+ * this functionality was not tested.
+ *
+ * On fail call [[Inflate#onEnd]] with error code and return false.
+ *
+ * ##### Example
+ *
+ * ```javascript
+ * push(chunk, false); // push one of data chunks
+ * ...
+ * push(chunk, true);  // push last chunk
+ * ```
+ **/
+Inflate.prototype.push = function (data, flush_mode) {
+  const strm = this.strm;
+  const chunkSize = this.options.chunkSize;
+  const dictionary = this.options.dictionary;
+  let status, _flush_mode, last_avail_out;
+
+  if (this.ended) return false;
+
+  if (flush_mode === ~~flush_mode) _flush_mode = flush_mode;
+  else _flush_mode = flush_mode === true ? Z_FINISH : Z_NO_FLUSH;
+
+  // Convert data if needed
+  if (toString.call(data) === '[object ArrayBuffer]') {
+    strm.input = new Uint8Array(data);
+  } else {
+    strm.input = data;
+  }
+
+  strm.next_in = 0;
+  strm.avail_in = strm.input.length;
+
+  for (;;) {
+    if (strm.avail_out === 0) {
+      strm.output = new Uint8Array(chunkSize);
+      strm.next_out = 0;
+      strm.avail_out = chunkSize;
+    }
+
+    status = zlib_inflate.inflate(strm, _flush_mode);
+
+    if (status === Z_NEED_DICT && dictionary) {
+      status = zlib_inflate.inflateSetDictionary(strm, dictionary);
+
+      if (status === Z_OK) {
+        status = zlib_inflate.inflate(strm, _flush_mode);
+      } else if (status === Z_DATA_ERROR) {
+        // Replace code with more verbose
+        status = Z_NEED_DICT;
+      }
+    }
+
+    // Skip snyc markers if more data follows and not raw mode
+    while (strm.avail_in > 0 &&
+           status === Z_STREAM_END &&
+           strm.state.wrap > 0 &&
+           data[strm.next_in] !== 0)
+    {
+      zlib_inflate.inflateReset(strm);
+      status = zlib_inflate.inflate(strm, _flush_mode);
+    }
+
+    switch (status) {
+      case Z_STREAM_ERROR:
+      case Z_DATA_ERROR:
+      case Z_NEED_DICT:
+      case Z_MEM_ERROR:
+        this.onEnd(status);
+        this.ended = true;
+        return false;
+    }
+
+    // Remember real `avail_out` value, because we may patch out buffer content
+    // to align utf8 strings boundaries.
+    last_avail_out = strm.avail_out;
+
+    if (strm.next_out) {
+      if (strm.avail_out === 0 || status === Z_STREAM_END) {
+
+        if (this.options.to === 'string') {
+
+          let next_out_utf8 = strings.utf8border(strm.output, strm.next_out);
+
+          let tail = strm.next_out - next_out_utf8;
+          let utf8str = strings.buf2string(strm.output, next_out_utf8);
+
+          // move tail & realign counters
+          strm.next_out = tail;
+          strm.avail_out = chunkSize - tail;
+          if (tail) strm.output.set(strm.output.subarray(next_out_utf8, next_out_utf8 + tail), 0);
+
+          this.onData(utf8str);
+
+        } else {
+          this.onData(strm.output.length === strm.next_out ? strm.output : strm.output.subarray(0, strm.next_out));
+        }
+      }
+    }
+
+    // Must repeat iteration if out buffer is full
+    if (status === Z_OK && last_avail_out === 0) continue;
+
+    // Finalize if end of stream reached.
+    if (status === Z_STREAM_END) {
+      status = zlib_inflate.inflateEnd(this.strm);
+      this.onEnd(status);
+      this.ended = true;
+      return true;
+    }
+
+    if (strm.avail_in === 0) break;
+  }
+
+  return true;
+};
+
+
+/**
+ * Inflate#onData(chunk) -> Void
+ * - chunk (Uint8Array|String): output data. When string output requested,
+ *   each chunk will be string.
+ *
+ * By default, stores data blocks in `chunks[]` property and glue
+ * those in `onEnd`. Override this handler, if you need another behaviour.
+ **/
+Inflate.prototype.onData = function (chunk) {
+  this.chunks.push(chunk);
+};
+
+
+/**
+ * Inflate#onEnd(status) -> Void
+ * - status (Number): inflate status. 0 (Z_OK) on success,
+ *   other if not.
+ *
+ * Called either after you tell inflate that the input stream is
+ * complete (Z_FINISH). By default - join collected chunks,
+ * free memory and fill `results` / `err` properties.
+ **/
+Inflate.prototype.onEnd = function (status) {
+  // On success - join
+  if (status === Z_OK) {
+    if (this.options.to === 'string') {
+      this.result = this.chunks.join('');
+    } else {
+      this.result = utils.flattenChunks(this.chunks);
+    }
+  }
+  this.chunks = [];
+  this.err = status;
+  this.msg = this.strm.msg;
+};
+
+
+/**
+ * inflate(data[, options]) -> Uint8Array|String
+ * - data (Uint8Array|ArrayBuffer): input data to decompress.
+ * - options (Object): zlib inflate options.
+ *
+ * Decompress `data` with inflate/ungzip and `options`. Autodetect
+ * format via wrapper header by default. That's why we don't provide
+ * separate `ungzip` method.
+ *
+ * Supported options are:
+ *
+ * - windowBits
+ *
+ * [http://zlib.net/manual.html#Advanced](http://zlib.net/manual.html#Advanced)
+ * for more information.
+ *
+ * Sugar (options):
+ *
+ * - `raw` (Boolean) - say that we work with raw stream, if you don't wish to specify
+ *   negative windowBits implicitly.
+ * - `to` (String) - if equal to 'string', then result will be converted
+ *   from utf8 to utf16 (javascript) string. When string output requested,
+ *   chunk length can differ from `chunkSize`, depending on content.
+ *
+ *
+ * ##### Example:
+ *
+ * ```javascript
+ * const pako = require('pako');
+ * const input = pako.deflate(new Uint8Array([1,2,3,4,5,6,7,8,9]));
+ * let output;
+ *
+ * try {
+ *   output = pako.inflate(input);
+ * } catch (err) {
+ *   console.log(err);
+ * }
+ * ```
+ **/
+function inflate(input, options) {
+  const inflator = new Inflate(options);
+
+  inflator.push(input);
+
+  // That will never happens, if you don't cheat with options :)
+  if (inflator.err) throw inflator.msg || msg[inflator.err];
+
+  return inflator.result;
+}
+
+
+/**
+ * inflateRaw(data[, options]) -> Uint8Array|String
+ * - data (Uint8Array|ArrayBuffer): input data to decompress.
+ * - options (Object): zlib inflate options.
+ *
+ * The same as [[inflate]], but creates raw data, without wrapper
+ * (header and adler32 crc).
+ **/
+function inflateRaw(input, options) {
+  options = options || {};
+  options.raw = true;
+  return inflate(input, options);
+}
+
+
+/**
+ * ungzip(data[, options]) -> Uint8Array|String
+ * - data (Uint8Array|ArrayBuffer): input data to decompress.
+ * - options (Object): zlib inflate options.
+ *
+ * Just shortcut to [[inflate]], because it autodetects format
+ * by header.content. Done for convenience.
+ **/
+
+
+module.exports.Inflate = Inflate;
+module.exports.inflate = inflate;
+module.exports.inflateRaw = inflateRaw;
+module.exports.ungzip = inflate;
+module.exports.constants = require('./zlib/constants');

测试流式传输uniapp/node_modules/pako/lib/utils/common.js

@@ -0,0 +1,48 @@
+'use strict';
+
+
+const _has = (obj, key) => {
+  return Object.prototype.hasOwnProperty.call(obj, key);
+};
+
+module.exports.assign = function (obj /*from1, from2, from3, ...*/) {
+  const sources = Array.prototype.slice.call(arguments, 1);
+  while (sources.length) {
+    const source = sources.shift();
+    if (!source) { continue; }
+
+    if (typeof source !== 'object') {
+      throw new TypeError(source + 'must be non-object');
+    }
+
+    for (const p in source) {
+      if (_has(source, p)) {
+        obj[p] = source[p];
+      }
+    }
+  }
+
+  return obj;
+};
+
+
+// Join array of chunks to single array.
+module.exports.flattenChunks = (chunks) => {
+  // calculate data length
+  let len = 0;
+
+  for (let i = 0, l = chunks.length; i < l; i++) {
+    len += chunks[i].length;
+  }
+
+  // join chunks
+  const result = new Uint8Array(len);
+
+  for (let i = 0, pos = 0, l = chunks.length; i < l; i++) {
+    let chunk = chunks[i];
+    result.set(chunk, pos);
+    pos += chunk.length;
+  }
+
+  return result;
+};

测试流式传输uniapp/node_modules/pako/lib/utils/strings.js

@@ -0,0 +1,174 @@
+// String encode/decode helpers
+'use strict';
+
+
+// Quick check if we can use fast array to bin string conversion
+//
+// - apply(Array) can fail on Android 2.2
+// - apply(Uint8Array) can fail on iOS 5.1 Safari
+//
+let STR_APPLY_UIA_OK = true;
+
+try { String.fromCharCode.apply(null, new Uint8Array(1)); } catch (__) { STR_APPLY_UIA_OK = false; }
+
+
+// Table with utf8 lengths (calculated by first byte of sequence)
+// Note, that 5 & 6-byte values and some 4-byte values can not be represented in JS,
+// because max possible codepoint is 0x10ffff
+const _utf8len = new Uint8Array(256);
+for (let q = 0; q < 256; q++) {
+  _utf8len[q] = (q >= 252 ? 6 : q >= 248 ? 5 : q >= 240 ? 4 : q >= 224 ? 3 : q >= 192 ? 2 : 1);
+}
+_utf8len[254] = _utf8len[254] = 1; // Invalid sequence start
+
+
+// convert string to array (typed, when possible)
+module.exports.string2buf = (str) => {
+  if (typeof TextEncoder === 'function' && TextEncoder.prototype.encode) {
+    return new TextEncoder().encode(str);
+  }
+
+  let buf, c, c2, m_pos, i, str_len = str.length, buf_len = 0;
+
+  // count binary size
+  for (m_pos = 0; m_pos < str_len; m_pos++) {
+    c = str.charCodeAt(m_pos);
+    if ((c & 0xfc00) === 0xd800 && (m_pos + 1 < str_len)) {
+      c2 = str.charCodeAt(m_pos + 1);
+      if ((c2 & 0xfc00) === 0xdc00) {
+        c = 0x10000 + ((c - 0xd800) << 10) + (c2 - 0xdc00);
+        m_pos++;
+      }
+    }
+    buf_len += c < 0x80 ? 1 : c < 0x800 ? 2 : c < 0x10000 ? 3 : 4;
+  }
+
+  // allocate buffer
+  buf = new Uint8Array(buf_len);
+
+  // convert
+  for (i = 0, m_pos = 0; i < buf_len; m_pos++) {
+    c = str.charCodeAt(m_pos);
+    if ((c & 0xfc00) === 0xd800 && (m_pos + 1 < str_len)) {
+      c2 = str.charCodeAt(m_pos + 1);
+      if ((c2 & 0xfc00) === 0xdc00) {
+        c = 0x10000 + ((c - 0xd800) << 10) + (c2 - 0xdc00);
+        m_pos++;
+      }
+    }
+    if (c < 0x80) {
+      /* one byte */
+      buf[i++] = c;
+    } else if (c < 0x800) {
+      /* two bytes */
+      buf[i++] = 0xC0 | (c >>> 6);
+      buf[i++] = 0x80 | (c & 0x3f);
+    } else if (c < 0x10000) {
+      /* three bytes */
+      buf[i++] = 0xE0 | (c >>> 12);
+      buf[i++] = 0x80 | (c >>> 6 & 0x3f);
+      buf[i++] = 0x80 | (c & 0x3f);
+    } else {
+      /* four bytes */
+      buf[i++] = 0xf0 | (c >>> 18);
+      buf[i++] = 0x80 | (c >>> 12 & 0x3f);
+      buf[i++] = 0x80 | (c >>> 6 & 0x3f);
+      buf[i++] = 0x80 | (c & 0x3f);
+    }
+  }
+
+  return buf;
+};
+
+// Helper
+const buf2binstring = (buf, len) => {
+  // On Chrome, the arguments in a function call that are allowed is `65534`.
+  // If the length of the buffer is smaller than that, we can use this optimization,
+  // otherwise we will take a slower path.
+  if (len < 65534) {
+    if (buf.subarray && STR_APPLY_UIA_OK) {
+      return String.fromCharCode.apply(null, buf.length === len ? buf : buf.subarray(0, len));
+    }
+  }
+
+  let result = '';
+  for (let i = 0; i < len; i++) {
+    result += String.fromCharCode(buf[i]);
+  }
+  return result;
+};
+
+
+// convert array to string
+module.exports.buf2string = (buf, max) => {
+  const len = max || buf.length;
+
+  if (typeof TextDecoder === 'function' && TextDecoder.prototype.decode) {
+    return new TextDecoder().decode(buf.subarray(0, max));
+  }
+
+  let i, out;
+
+  // Reserve max possible length (2 words per char)
+  // NB: by unknown reasons, Array is significantly faster for
+  //     String.fromCharCode.apply than Uint16Array.
+  const utf16buf = new Array(len * 2);
+
+  for (out = 0, i = 0; i < len;) {
+    let c = buf[i++];
+    // quick process ascii
+    if (c < 0x80) { utf16buf[out++] = c; continue; }
+
+    let c_len = _utf8len[c];
+    // skip 5 & 6 byte codes
+    if (c_len > 4) { utf16buf[out++] = 0xfffd; i += c_len - 1; continue; }
+
+    // apply mask on first byte
+    c &= c_len === 2 ? 0x1f : c_len === 3 ? 0x0f : 0x07;
+    // join the rest
+    while (c_len > 1 && i < len) {
+      c = (c << 6) | (buf[i++] & 0x3f);
+      c_len--;
+    }
+
+    // terminated by end of string?
+    if (c_len > 1) { utf16buf[out++] = 0xfffd; continue; }
+
+    if (c < 0x10000) {
+      utf16buf[out++] = c;
+    } else {
+      c -= 0x10000;
+      utf16buf[out++] = 0xd800 | ((c >> 10) & 0x3ff);
+      utf16buf[out++] = 0xdc00 | (c & 0x3ff);
+    }
+  }
+
+  return buf2binstring(utf16buf, out);
+};
+
+
+// Calculate max possible position in utf8 buffer,
+// that will not break sequence. If that's not possible
+// - (very small limits) return max size as is.
+//
+// buf[] - utf8 bytes array
+// max   - length limit (mandatory);
+module.exports.utf8border = (buf, max) => {
+
+  max = max || buf.length;
+  if (max > buf.length) { max = buf.length; }
+
+  // go back from last position, until start of sequence found
+  let pos = max - 1;
+  while (pos >= 0 && (buf[pos] & 0xC0) === 0x80) { pos--; }
+
+  // Very small and broken sequence,
+  // return max, because we should return something anyway.
+  if (pos < 0) { return max; }
+
+  // If we came to start of buffer - that means buffer is too small,
+  // return max too.
+  if (pos === 0) { return max; }
+
+  return (pos + _utf8len[buf[pos]] > max) ? pos : max;
+};

测试流式传输uniapp/node_modules/pako/lib/zlib/README

@@ -0,0 +1,59 @@
+Content of this folder follows zlib C sources as close as possible.
+That's intended to simplify maintainability and guarantee equal API
+and result.
+
+Key differences:
+
+- Everything is in JavaScript.
+- No platform-dependent blocks.
+- Some things like crc32 rewritten to keep size small and make JIT
+  work better.
+- Some code is different due missed features in JS (macros, pointers,
+  structures, header files)
+- Specific API methods are not implemented (see notes in root readme)
+
+This port is based on zlib 1.2.8.
+
+This port is under zlib license (see below) with contribution and addition of javascript
+port under expat license (see LICENSE at root of project)
+
+Copyright:
+(C) 1995-2013 Jean-loup Gailly and Mark Adler
+(C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+
+
+From zlib's README
+=============================================================================
+
+Acknowledgments:
+
+  The deflate format used by zlib was defined by Phil Katz.  The deflate and
+  zlib specifications were written by L.  Peter Deutsch.  Thanks to all the
+  people who reported problems and suggested various improvements in zlib; they
+  are too numerous to cite here.
+
+Copyright notice:
+
+ (C) 1995-2013 Jean-loup Gailly and Mark Adler
+
+Copyright (c) <''year''> <''copyright holders''>
+
+This software is provided 'as-is', without any express or implied
+warranty. In no event will the authors be held liable for any damages
+arising from the use of this software.
+
+Permission is granted to anyone to use this software for any purpose,
+including commercial applications, and to alter it and redistribute it
+freely, subject to the following restrictions:
+
+1. The origin of this software must not be misrepresented; you must not
+   claim that you wrote the original software. If you use this software
+   in a product, an acknowledgment in the product documentation would be
+   appreciated but is not required.
+2. Altered source versions must be plainly marked as such, and must not be
+   misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu

测试流式传输uniapp/node_modules/pako/lib/zlib/adler32.js

@@ -0,0 +1,51 @@
+'use strict';
+
+// Note: adler32 takes 12% for level 0 and 2% for level 6.
+// It isn't worth it to make additional optimizations as in original.
+// Small size is preferable.
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+const adler32 = (adler, buf, len, pos) => {
+  let s1 = (adler & 0xffff) |0,
+      s2 = ((adler >>> 16) & 0xffff) |0,
+      n = 0;
+
+  while (len !== 0) {
+    // Set limit ~ twice less than 5552, to keep
+    // s2 in 31-bits, because we force signed ints.
+    // in other case %= will fail.
+    n = len > 2000 ? 2000 : len;
+    len -= n;
+
+    do {
+      s1 = (s1 + buf[pos++]) |0;
+      s2 = (s2 + s1) |0;
+    } while (--n);
+
+    s1 %= 65521;
+    s2 %= 65521;
+  }
+
+  return (s1 | (s2 << 16)) |0;
+};
+
+
+module.exports = adler32;

测试流式传输uniapp/node_modules/pako/lib/zlib/constants.js

@@ -0,0 +1,68 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+module.exports = {
+
+  /* Allowed flush values; see deflate() and inflate() below for details */
+  Z_NO_FLUSH:         0,
+  Z_PARTIAL_FLUSH:    1,
+  Z_SYNC_FLUSH:       2,
+  Z_FULL_FLUSH:       3,
+  Z_FINISH:           4,
+  Z_BLOCK:            5,
+  Z_TREES:            6,
+
+  /* Return codes for the compression/decompression functions. Negative values
+  * are errors, positive values are used for special but normal events.
+  */
+  Z_OK:               0,
+  Z_STREAM_END:       1,
+  Z_NEED_DICT:        2,
+  Z_ERRNO:           -1,
+  Z_STREAM_ERROR:    -2,
+  Z_DATA_ERROR:      -3,
+  Z_MEM_ERROR:       -4,
+  Z_BUF_ERROR:       -5,
+  //Z_VERSION_ERROR: -6,
+
+  /* compression levels */
+  Z_NO_COMPRESSION:         0,
+  Z_BEST_SPEED:             1,
+  Z_BEST_COMPRESSION:       9,
+  Z_DEFAULT_COMPRESSION:   -1,
+
+
+  Z_FILTERED:               1,
+  Z_HUFFMAN_ONLY:           2,
+  Z_RLE:                    3,
+  Z_FIXED:                  4,
+  Z_DEFAULT_STRATEGY:       0,
+
+  /* Possible values of the data_type field (though see inflate()) */
+  Z_BINARY:                 0,
+  Z_TEXT:                   1,
+  //Z_ASCII:                1, // = Z_TEXT (deprecated)
+  Z_UNKNOWN:                2,
+
+  /* The deflate compression method */
+  Z_DEFLATED:               8
+  //Z_NULL:                 null // Use -1 or null inline, depending on var type
+};

测试流式传输uniapp/node_modules/pako/lib/zlib/crc32.js

@@ -0,0 +1,59 @@
+'use strict';
+
+// Note: we can't get significant speed boost here.
+// So write code to minimize size - no pregenerated tables
+// and array tools dependencies.
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+// Use ordinary array, since untyped makes no boost here
+const makeTable = () => {
+  let c, table = [];
+
+  for (var n = 0; n < 256; n++) {
+    c = n;
+    for (var k = 0; k < 8; k++) {
+      c = ((c & 1) ? (0xEDB88320 ^ (c >>> 1)) : (c >>> 1));
+    }
+    table[n] = c;
+  }
+
+  return table;
+};
+
+// Create table on load. Just 255 signed longs. Not a problem.
+const crcTable = new Uint32Array(makeTable());
+
+
+const crc32 = (crc, buf, len, pos) => {
+  const t = crcTable;
+  const end = pos + len;
+
+  crc ^= -1;
+
+  for (let i = pos; i < end; i++) {
+    crc = (crc >>> 8) ^ t[(crc ^ buf[i]) & 0xFF];
+  }
+
+  return (crc ^ (-1)); // >>> 0;
+};
+
+
+module.exports = crc32;

测试流式传输uniapp/node_modules/pako/lib/zlib/gzheader.js

@@ -0,0 +1,58 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+function GZheader() {
+  /* true if compressed data believed to be text */
+  this.text       = 0;
+  /* modification time */
+  this.time       = 0;
+  /* extra flags (not used when writing a gzip file) */
+  this.xflags     = 0;
+  /* operating system */
+  this.os         = 0;
+  /* pointer to extra field or Z_NULL if none */
+  this.extra      = null;
+  /* extra field length (valid if extra != Z_NULL) */
+  this.extra_len  = 0; // Actually, we don't need it in JS,
+                       // but leave for few code modifications
+
+  //
+  // Setup limits is not necessary because in js we should not preallocate memory
+  // for inflate use constant limit in 65536 bytes
+  //
+
+  /* space at extra (only when reading header) */
+  // this.extra_max  = 0;
+  /* pointer to zero-terminated file name or Z_NULL */
+  this.name       = '';
+  /* space at name (only when reading header) */
+  // this.name_max   = 0;
+  /* pointer to zero-terminated comment or Z_NULL */
+  this.comment    = '';
+  /* space at comment (only when reading header) */
+  // this.comm_max   = 0;
+  /* true if there was or will be a header crc */
+  this.hcrc       = 0;
+  /* true when done reading gzip header (not used when writing a gzip file) */
+  this.done       = false;
+}
+
+module.exports = GZheader;

测试流式传输uniapp/node_modules/pako/lib/zlib/inffast.js

@@ -0,0 +1,344 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+// See state defs from inflate.js
+const BAD = 16209;       /* got a data error -- remain here until reset */
+const TYPE = 16191;      /* i: waiting for type bits, including last-flag bit */
+
+/*
+   Decode literal, length, and distance codes and write out the resulting
+   literal and match bytes until either not enough input or output is
+   available, an end-of-block is encountered, or a data error is encountered.
+   When large enough input and output buffers are supplied to inflate(), for
+   example, a 16K input buffer and a 64K output buffer, more than 95% of the
+   inflate execution time is spent in this routine.
+
+   Entry assumptions:
+
+        state.mode === LEN
+        strm.avail_in >= 6
+        strm.avail_out >= 258
+        start >= strm.avail_out
+        state.bits < 8
+
+   On return, state.mode is one of:
+
+        LEN -- ran out of enough output space or enough available input
+        TYPE -- reached end of block code, inflate() to interpret next block
+        BAD -- error in block data
+
+   Notes:
+
+    - The maximum input bits used by a length/distance pair is 15 bits for the
+      length code, 5 bits for the length extra, 15 bits for the distance code,
+      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
+      Therefore if strm.avail_in >= 6, then there is enough input to avoid
+      checking for available input while decoding.
+
+    - The maximum bytes that a single length/distance pair can output is 258
+      bytes, which is the maximum length that can be coded.  inflate_fast()
+      requires strm.avail_out >= 258 for each loop to avoid checking for
+      output space.
+ */
+module.exports = function inflate_fast(strm, start) {
+  let _in;                    /* local strm.input */
+  let last;                   /* have enough input while in < last */
+  let _out;                   /* local strm.output */
+  let beg;                    /* inflate()'s initial strm.output */
+  let end;                    /* while out < end, enough space available */
+//#ifdef INFLATE_STRICT
+  let dmax;                   /* maximum distance from zlib header */
+//#endif
+  let wsize;                  /* window size or zero if not using window */
+  let whave;                  /* valid bytes in the window */
+  let wnext;                  /* window write index */
+  // Use `s_window` instead `window`, avoid conflict with instrumentation tools
+  let s_window;               /* allocated sliding window, if wsize != 0 */
+  let hold;                   /* local strm.hold */
+  let bits;                   /* local strm.bits */
+  let lcode;                  /* local strm.lencode */
+  let dcode;                  /* local strm.distcode */
+  let lmask;                  /* mask for first level of length codes */
+  let dmask;                  /* mask for first level of distance codes */
+  let here;                   /* retrieved table entry */
+  let op;                     /* code bits, operation, extra bits, or */
+                              /*  window position, window bytes to copy */
+  let len;                    /* match length, unused bytes */
+  let dist;                   /* match distance */
+  let from;                   /* where to copy match from */
+  let from_source;
+
+
+  let input, output; // JS specific, because we have no pointers
+
+  /* copy state to local variables */
+  const state = strm.state;
+  //here = state.here;
+  _in = strm.next_in;
+  input = strm.input;
+  last = _in + (strm.avail_in - 5);
+  _out = strm.next_out;
+  output = strm.output;
+  beg = _out - (start - strm.avail_out);
+  end = _out + (strm.avail_out - 257);
+//#ifdef INFLATE_STRICT
+  dmax = state.dmax;
+//#endif
+  wsize = state.wsize;
+  whave = state.whave;
+  wnext = state.wnext;
+  s_window = state.window;
+  hold = state.hold;
+  bits = state.bits;
+  lcode = state.lencode;
+  dcode = state.distcode;
+  lmask = (1 << state.lenbits) - 1;
+  dmask = (1 << state.distbits) - 1;
+
+
+  /* decode literals and length/distances until end-of-block or not enough
+     input data or output space */
+
+  top:
+  do {
+    if (bits < 15) {
+      hold += input[_in++] << bits;
+      bits += 8;
+      hold += input[_in++] << bits;
+      bits += 8;
+    }
+
+    here = lcode[hold & lmask];
+
+    dolen:
+    for (;;) { // Goto emulation
+      op = here >>> 24/*here.bits*/;
+      hold >>>= op;
+      bits -= op;
+      op = (here >>> 16) & 0xff/*here.op*/;
+      if (op === 0) {                          /* literal */
+        //Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+        //        "inflate:         literal '%c'\n" :
+        //        "inflate:         literal 0x%02x\n", here.val));
+        output[_out++] = here & 0xffff/*here.val*/;
+      }
+      else if (op & 16) {                     /* length base */
+        len = here & 0xffff/*here.val*/;
+        op &= 15;                           /* number of extra bits */
+        if (op) {
+          if (bits < op) {
+            hold += input[_in++] << bits;
+            bits += 8;
+          }
+          len += hold & ((1 << op) - 1);
+          hold >>>= op;
+          bits -= op;
+        }
+        //Tracevv((stderr, "inflate:         length %u\n", len));
+        if (bits < 15) {
+          hold += input[_in++] << bits;
+          bits += 8;
+          hold += input[_in++] << bits;
+          bits += 8;
+        }
+        here = dcode[hold & dmask];
+
+        dodist:
+        for (;;) { // goto emulation
+          op = here >>> 24/*here.bits*/;
+          hold >>>= op;
+          bits -= op;
+          op = (here >>> 16) & 0xff/*here.op*/;
+
+          if (op & 16) {                      /* distance base */
+            dist = here & 0xffff/*here.val*/;
+            op &= 15;                       /* number of extra bits */
+            if (bits < op) {
+              hold += input[_in++] << bits;
+              bits += 8;
+              if (bits < op) {
+                hold += input[_in++] << bits;
+                bits += 8;
+              }
+            }
+            dist += hold & ((1 << op) - 1);
+//#ifdef INFLATE_STRICT
+            if (dist > dmax) {
+              strm.msg = 'invalid distance too far back';
+              state.mode = BAD;
+              break top;
+            }
+//#endif
+            hold >>>= op;
+            bits -= op;
+            //Tracevv((stderr, "inflate:         distance %u\n", dist));
+            op = _out - beg;                /* max distance in output */
+            if (dist > op) {                /* see if copy from window */
+              op = dist - op;               /* distance back in window */
+              if (op > whave) {
+                if (state.sane) {
+                  strm.msg = 'invalid distance too far back';
+                  state.mode = BAD;
+                  break top;
+                }
+
+// (!) This block is disabled in zlib defaults,
+// don't enable it for binary compatibility
+//#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+//                if (len <= op - whave) {
+//                  do {
+//                    output[_out++] = 0;
+//                  } while (--len);
+//                  continue top;
+//                }
+//                len -= op - whave;
+//                do {
+//                  output[_out++] = 0;
+//                } while (--op > whave);
+//                if (op === 0) {
+//                  from = _out - dist;
+//                  do {
+//                    output[_out++] = output[from++];
+//                  } while (--len);
+//                  continue top;
+//                }
+//#endif
+              }
+              from = 0; // window index
+              from_source = s_window;
+              if (wnext === 0) {           /* very common case */
+                from += wsize - op;
+                if (op < len) {         /* some from window */
+                  len -= op;
+                  do {
+                    output[_out++] = s_window[from++];
+                  } while (--op);
+                  from = _out - dist;  /* rest from output */
+                  from_source = output;
+                }
+              }
+              else if (wnext < op) {      /* wrap around window */
+                from += wsize + wnext - op;
+                op -= wnext;
+                if (op < len) {         /* some from end of window */
+                  len -= op;
+                  do {
+                    output[_out++] = s_window[from++];
+                  } while (--op);
+                  from = 0;
+                  if (wnext < len) {  /* some from start of window */
+                    op = wnext;
+                    len -= op;
+                    do {
+                      output[_out++] = s_window[from++];
+                    } while (--op);
+                    from = _out - dist;      /* rest from output */
+                    from_source = output;
+                  }
+                }
+              }
+              else {                      /* contiguous in window */
+                from += wnext - op;
+                if (op < len) {         /* some from window */
+                  len -= op;
+                  do {
+                    output[_out++] = s_window[from++];
+                  } while (--op);
+                  from = _out - dist;  /* rest from output */
+                  from_source = output;
+                }
+              }
+              while (len > 2) {
+                output[_out++] = from_source[from++];
+                output[_out++] = from_source[from++];
+                output[_out++] = from_source[from++];
+                len -= 3;
+              }
+              if (len) {
+                output[_out++] = from_source[from++];
+                if (len > 1) {
+                  output[_out++] = from_source[from++];
+                }
+              }
+            }
+            else {
+              from = _out - dist;          /* copy direct from output */
+              do {                        /* minimum length is three */
+                output[_out++] = output[from++];
+                output[_out++] = output[from++];
+                output[_out++] = output[from++];
+                len -= 3;
+              } while (len > 2);
+              if (len) {
+                output[_out++] = output[from++];
+                if (len > 1) {
+                  output[_out++] = output[from++];
+                }
+              }
+            }
+          }
+          else if ((op & 64) === 0) {          /* 2nd level distance code */
+            here = dcode[(here & 0xffff)/*here.val*/ + (hold & ((1 << op) - 1))];
+            continue dodist;
+          }
+          else {
+            strm.msg = 'invalid distance code';
+            state.mode = BAD;
+            break top;
+          }
+
+          break; // need to emulate goto via "continue"
+        }
+      }
+      else if ((op & 64) === 0) {              /* 2nd level length code */
+        here = lcode[(here & 0xffff)/*here.val*/ + (hold & ((1 << op) - 1))];
+        continue dolen;
+      }
+      else if (op & 32) {                     /* end-of-block */
+        //Tracevv((stderr, "inflate:         end of block\n"));
+        state.mode = TYPE;
+        break top;
+      }
+      else {
+        strm.msg = 'invalid literal/length code';
+        state.mode = BAD;
+        break top;
+      }
+
+      break; // need to emulate goto via "continue"
+    }
+  } while (_in < last && _out < end);
+
+  /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
+  len = bits >> 3;
+  _in -= len;
+  bits -= len << 3;
+  hold &= (1 << bits) - 1;
+
+  /* update state and return */
+  strm.next_in = _in;
+  strm.next_out = _out;
+  strm.avail_in = (_in < last ? 5 + (last - _in) : 5 - (_in - last));
+  strm.avail_out = (_out < end ? 257 + (end - _out) : 257 - (_out - end));
+  state.hold = hold;
+  state.bits = bits;
+  return;
+};

测试流式传输uniapp/node_modules/pako/lib/zlib/inftrees.js

@@ -0,0 +1,340 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+const MAXBITS = 15;
+const ENOUGH_LENS = 852;
+const ENOUGH_DISTS = 592;
+//const ENOUGH = (ENOUGH_LENS+ENOUGH_DISTS);
+
+const CODES = 0;
+const LENS = 1;
+const DISTS = 2;
+
+const lbase = new Uint16Array([ /* Length codes 257..285 base */
+  3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+  35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0
+]);
+
+const lext = new Uint8Array([ /* Length codes 257..285 extra */
+  16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
+  19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 72, 78
+]);
+
+const dbase = new Uint16Array([ /* Distance codes 0..29 base */
+  1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+  257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+  8193, 12289, 16385, 24577, 0, 0
+]);
+
+const dext = new Uint8Array([ /* Distance codes 0..29 extra */
+  16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22,
+  23, 23, 24, 24, 25, 25, 26, 26, 27, 27,
+  28, 28, 29, 29, 64, 64
+]);
+
+const inflate_table = (type, lens, lens_index, codes, table, table_index, work, opts) =>
+{
+  const bits = opts.bits;
+      //here = opts.here; /* table entry for duplication */
+
+  let len = 0;               /* a code's length in bits */
+  let sym = 0;               /* index of code symbols */
+  let min = 0, max = 0;          /* minimum and maximum code lengths */
+  let root = 0;              /* number of index bits for root table */
+  let curr = 0;              /* number of index bits for current table */
+  let drop = 0;              /* code bits to drop for sub-table */
+  let left = 0;                   /* number of prefix codes available */
+  let used = 0;              /* code entries in table used */
+  let huff = 0;              /* Huffman code */
+  let incr;              /* for incrementing code, index */
+  let fill;              /* index for replicating entries */
+  let low;               /* low bits for current root entry */
+  let mask;              /* mask for low root bits */
+  let next;             /* next available space in table */
+  let base = null;     /* base value table to use */
+//  let shoextra;    /* extra bits table to use */
+  let match;                  /* use base and extra for symbol >= match */
+  const count = new Uint16Array(MAXBITS + 1); //[MAXBITS+1];    /* number of codes of each length */
+  const offs = new Uint16Array(MAXBITS + 1); //[MAXBITS+1];     /* offsets in table for each length */
+  let extra = null;
+
+  let here_bits, here_op, here_val;
+
+  /*
+   Process a set of code lengths to create a canonical Huffman code.  The
+   code lengths are lens[0..codes-1].  Each length corresponds to the
+   symbols 0..codes-1.  The Huffman code is generated by first sorting the
+   symbols by length from short to long, and retaining the symbol order
+   for codes with equal lengths.  Then the code starts with all zero bits
+   for the first code of the shortest length, and the codes are integer
+   increments for the same length, and zeros are appended as the length
+   increases.  For the deflate format, these bits are stored backwards
+   from their more natural integer increment ordering, and so when the
+   decoding tables are built in the large loop below, the integer codes
+   are incremented backwards.
+
+   This routine assumes, but does not check, that all of the entries in
+   lens[] are in the range 0..MAXBITS.  The caller must assure this.
+   1..MAXBITS is interpreted as that code length.  zero means that that
+   symbol does not occur in this code.
+
+   The codes are sorted by computing a count of codes for each length,
+   creating from that a table of starting indices for each length in the
+   sorted table, and then entering the symbols in order in the sorted
+   table.  The sorted table is work[], with that space being provided by
+   the caller.
+
+   The length counts are used for other purposes as well, i.e. finding
+   the minimum and maximum length codes, determining if there are any
+   codes at all, checking for a valid set of lengths, and looking ahead
+   at length counts to determine sub-table sizes when building the
+   decoding tables.
+   */
+
+  /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
+  for (len = 0; len <= MAXBITS; len++) {
+    count[len] = 0;
+  }
+  for (sym = 0; sym < codes; sym++) {
+    count[lens[lens_index + sym]]++;
+  }
+
+  /* bound code lengths, force root to be within code lengths */
+  root = bits;
+  for (max = MAXBITS; max >= 1; max--) {
+    if (count[max] !== 0) { break; }
+  }
+  if (root > max) {
+    root = max;
+  }
+  if (max === 0) {                     /* no symbols to code at all */
+    //table.op[opts.table_index] = 64;  //here.op = (var char)64;    /* invalid code marker */
+    //table.bits[opts.table_index] = 1;   //here.bits = (var char)1;
+    //table.val[opts.table_index++] = 0;   //here.val = (var short)0;
+    table[table_index++] = (1 << 24) | (64 << 16) | 0;
+
+
+    //table.op[opts.table_index] = 64;
+    //table.bits[opts.table_index] = 1;
+    //table.val[opts.table_index++] = 0;
+    table[table_index++] = (1 << 24) | (64 << 16) | 0;
+
+    opts.bits = 1;
+    return 0;     /* no symbols, but wait for decoding to report error */
+  }
+  for (min = 1; min < max; min++) {
+    if (count[min] !== 0) { break; }
+  }
+  if (root < min) {
+    root = min;
+  }
+
+  /* check for an over-subscribed or incomplete set of lengths */
+  left = 1;
+  for (len = 1; len <= MAXBITS; len++) {
+    left <<= 1;
+    left -= count[len];
+    if (left < 0) {
+      return -1;
+    }        /* over-subscribed */
+  }
+  if (left > 0 && (type === CODES || max !== 1)) {
+    return -1;                      /* incomplete set */
+  }
+
+  /* generate offsets into symbol table for each length for sorting */
+  offs[1] = 0;
+  for (len = 1; len < MAXBITS; len++) {
+    offs[len + 1] = offs[len] + count[len];
+  }
+
+  /* sort symbols by length, by symbol order within each length */
+  for (sym = 0; sym < codes; sym++) {
+    if (lens[lens_index + sym] !== 0) {
+      work[offs[lens[lens_index + sym]]++] = sym;
+    }
+  }
+
+  /*
+   Create and fill in decoding tables.  In this loop, the table being
+   filled is at next and has curr index bits.  The code being used is huff
+   with length len.  That code is converted to an index by dropping drop
+   bits off of the bottom.  For codes where len is less than drop + curr,
+   those top drop + curr - len bits are incremented through all values to
+   fill the table with replicated entries.
+
+   root is the number of index bits for the root table.  When len exceeds
+   root, sub-tables are created pointed to by the root entry with an index
+   of the low root bits of huff.  This is saved in low to check for when a
+   new sub-table should be started.  drop is zero when the root table is
+   being filled, and drop is root when sub-tables are being filled.
+
+   When a new sub-table is needed, it is necessary to look ahead in the
+   code lengths to determine what size sub-table is needed.  The length
+   counts are used for this, and so count[] is decremented as codes are
+   entered in the tables.
+
+   used keeps track of how many table entries have been allocated from the
+   provided *table space.  It is checked for LENS and DIST tables against
+   the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
+   the initial root table size constants.  See the comments in inftrees.h
+   for more information.
+
+   sym increments through all symbols, and the loop terminates when
+   all codes of length max, i.e. all codes, have been processed.  This
+   routine permits incomplete codes, so another loop after this one fills
+   in the rest of the decoding tables with invalid code markers.
+   */
+
+  /* set up for code type */
+  // poor man optimization - use if-else instead of switch,
+  // to avoid deopts in old v8
+  if (type === CODES) {
+    base = extra = work;    /* dummy value--not used */
+    match = 20;
+
+  } else if (type === LENS) {
+    base = lbase;
+    extra = lext;
+    match = 257;
+
+  } else {                    /* DISTS */
+    base = dbase;
+    extra = dext;
+    match = 0;
+  }
+
+  /* initialize opts for loop */
+  huff = 0;                   /* starting code */
+  sym = 0;                    /* starting code symbol */
+  len = min;                  /* starting code length */
+  next = table_index;              /* current table to fill in */
+  curr = root;                /* current table index bits */
+  drop = 0;                   /* current bits to drop from code for index */
+  low = -1;                   /* trigger new sub-table when len > root */
+  used = 1 << root;          /* use root table entries */
+  mask = used - 1;            /* mask for comparing low */
+
+  /* check available table space */
+  if ((type === LENS && used > ENOUGH_LENS) ||
+    (type === DISTS && used > ENOUGH_DISTS)) {
+    return 1;
+  }
+
+  /* process all codes and make table entries */
+  for (;;) {
+    /* create table entry */
+    here_bits = len - drop;
+    if (work[sym] + 1 < match) {
+      here_op = 0;
+      here_val = work[sym];
+    }
+    else if (work[sym] >= match) {
+      here_op = extra[work[sym] - match];
+      here_val = base[work[sym] - match];
+    }
+    else {
+      here_op = 32 + 64;         /* end of block */
+      here_val = 0;
+    }
+
+    /* replicate for those indices with low len bits equal to huff */
+    incr = 1 << (len - drop);
+    fill = 1 << curr;
+    min = fill;                 /* save offset to next table */
+    do {
+      fill -= incr;
+      table[next + (huff >> drop) + fill] = (here_bits << 24) | (here_op << 16) | here_val |0;
+    } while (fill !== 0);
+
+    /* backwards increment the len-bit code huff */
+    incr = 1 << (len - 1);
+    while (huff & incr) {
+      incr >>= 1;
+    }
+    if (incr !== 0) {
+      huff &= incr - 1;
+      huff += incr;
+    } else {
+      huff = 0;
+    }
+
+    /* go to next symbol, update count, len */
+    sym++;
+    if (--count[len] === 0) {
+      if (len === max) { break; }
+      len = lens[lens_index + work[sym]];
+    }
+
+    /* create new sub-table if needed */
+    if (len > root && (huff & mask) !== low) {
+      /* if first time, transition to sub-tables */
+      if (drop === 0) {
+        drop = root;
+      }
+
+      /* increment past last table */
+      next += min;            /* here min is 1 << curr */
+
+      /* determine length of next table */
+      curr = len - drop;
+      left = 1 << curr;
+      while (curr + drop < max) {
+        left -= count[curr + drop];
+        if (left <= 0) { break; }
+        curr++;
+        left <<= 1;
+      }
+
+      /* check for enough space */
+      used += 1 << curr;
+      if ((type === LENS && used > ENOUGH_LENS) ||
+        (type === DISTS && used > ENOUGH_DISTS)) {
+        return 1;
+      }
+
+      /* point entry in root table to sub-table */
+      low = huff & mask;
+      /*table.op[low] = curr;
+      table.bits[low] = root;
+      table.val[low] = next - opts.table_index;*/
+      table[low] = (root << 24) | (curr << 16) | (next - table_index) |0;
+    }
+  }
+
+  /* fill in remaining table entry if code is incomplete (guaranteed to have
+   at most one remaining entry, since if the code is incomplete, the
+   maximum code length that was allowed to get this far is one bit) */
+  if (huff !== 0) {
+    //table.op[next + huff] = 64;            /* invalid code marker */
+    //table.bits[next + huff] = len - drop;
+    //table.val[next + huff] = 0;
+    table[next + huff] = ((len - drop) << 24) | (64 << 16) |0;
+  }
+
+  /* set return parameters */
+  //opts.table_index += used;
+  opts.bits = root;
+  return 0;
+};
+
+
+module.exports = inflate_table;

测试流式传输uniapp/node_modules/pako/lib/zlib/messages.js

@@ -0,0 +1,32 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+module.exports = {
+  2:      'need dictionary',     /* Z_NEED_DICT       2  */
+  1:      'stream end',          /* Z_STREAM_END      1  */
+  0:      '',                    /* Z_OK              0  */
+  '-1':   'file error',          /* Z_ERRNO         (-1) */
+  '-2':   'stream error',        /* Z_STREAM_ERROR  (-2) */
+  '-3':   'data error',          /* Z_DATA_ERROR    (-3) */
+  '-4':   'insufficient memory', /* Z_MEM_ERROR     (-4) */
+  '-5':   'buffer error',        /* Z_BUF_ERROR     (-5) */
+  '-6':   'incompatible version' /* Z_VERSION_ERROR (-6) */
+};

测试流式传输uniapp/node_modules/pako/lib/zlib/zstream.js

@@ -0,0 +1,47 @@
+'use strict';
+
+// (C) 1995-2013 Jean-loup Gailly and Mark Adler
+// (C) 2014-2017 Vitaly Puzrin and Andrey Tupitsin
+//
+// This software is provided 'as-is', without any express or implied
+// warranty. In no event will the authors be held liable for any damages
+// arising from the use of this software.
+//
+// Permission is granted to anyone to use this software for any purpose,
+// including commercial applications, and to alter it and redistribute it
+// freely, subject to the following restrictions:
+//
+// 1. The origin of this software must not be misrepresented; you must not
+//   claim that you wrote the original software. If you use this software
+//   in a product, an acknowledgment in the product documentation would be
+//   appreciated but is not required.
+// 2. Altered source versions must be plainly marked as such, and must not be
+//   misrepresented as being the original software.
+// 3. This notice may not be removed or altered from any source distribution.
+
+function ZStream() {
+  /* next input byte */
+  this.input = null; // JS specific, because we have no pointers
+  this.next_in = 0;
+  /* number of bytes available at input */
+  this.avail_in = 0;
+  /* total number of input bytes read so far */
+  this.total_in = 0;
+  /* next output byte should be put there */
+  this.output = null; // JS specific, because we have no pointers
+  this.next_out = 0;
+  /* remaining free space at output */
+  this.avail_out = 0;
+  /* total number of bytes output so far */
+  this.total_out = 0;
+  /* last error message, NULL if no error */
+  this.msg = ''/*Z_NULL*/;
+  /* not visible by applications */
+  this.state = null;
+  /* best guess about the data type: binary or text */
+  this.data_type = 2/*Z_UNKNOWN*/;
+  /* adler32 value of the uncompressed data */
+  this.adler = 0;
+}
+
+module.exports = ZStream;

测试流式传输uniapp/node_modules/pako/package.json

@@ -0,0 +1,64 @@
+{
+  "name": "pako",
+  "description": "zlib port to javascript - fast, modularized, with browser support",
+  "version": "2.1.0",
+  "keywords": [
+    "zlib",
+    "deflate",
+    "inflate",
+    "gzip"
+  ],
+  "contributors": [
+    "Andrei Tuputcyn (https://github.com/andr83)",
+    "Vitaly Puzrin (https://github.com/puzrin)",
+    "Friedel Ziegelmayer (https://github.com/dignifiedquire)",
+    "Kirill Efimov (https://github.com/Kirill89)",
+    "Jean-loup Gailly",
+    "Mark Adler"
+  ],
+  "files": [
+    "index.js",
+    "dist/",
+    "lib/"
+  ],
+  "license": "(MIT AND Zlib)",
+  "repository": "nodeca/pako",
+  "module": "./dist/pako.esm.mjs",
+  "exports": {
+    ".": {
+      "import": "./dist/pako.esm.mjs",
+      "require": "./index.js"
+    },
+    "./package.json": "./package.json",
+    "./dist/*": "./dist/*",
+    "./lib/*": "./lib/*",
+    "./lib/zlib/*": "./lib/zlib/*",
+    "./lib/utils/*": "./lib/utils/*"
+  },
+  "scripts": {
+    "lint": "eslint .",
+    "test": "npm run lint && mocha",
+    "coverage": "npm run lint && nyc mocha && nyc report --reporter html",
+    "build": "rollup -c",
+    "build_fixtures": "node support/build_fixtures.js",
+    "doc": "node support/build_doc.js",
+    "gh-doc": "npm run doc && gh-pages -d doc -f",
+    "prepublishOnly": "npm run gh-doc"
+  },
+  "devDependencies": {
+    "@babel/preset-env": "^7.12.1",
+    "@rollup/plugin-babel": "^5.2.1",
+    "@rollup/plugin-commonjs": "^16.0.0",
+    "@rollup/plugin-node-resolve": "^10.0.0",
+    "eslint": "^7.13.0",
+    "gh-pages": "^3.1.0",
+    "mocha": "^8.2.1",
+    "multiparty": "^4.1.3",
+    "ndoc": "^6.0.0",
+    "nyc": "^15.1.0",
+    "rollup": "^2.33.1",
+    "rollup-plugin-terser": "^7.0.2",
+    "shelljs": "^0.8.4"
+  },
+  "dependencies": {}
+}

测试流式传输uniapp/package-lock.json

@@ -0,0 +1,28 @@
+{
+  "name": "测试流式传输uniapp",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "dependencies": {
+        "pako": "^2.1.0"
+      },
+      "devDependencies": {
+        "@types/pako": "^2.0.4"
+      }
+    },
+    "node_modules/@types/pako": {
+      "version": "2.0.4",
+      "resolved": "https://registry.npmmirror.com/@types/pako/-/pako-2.0.4.tgz",
+      "integrity": "sha512-VWDCbrLeVXJM9fihYodcLiIv0ku+AlOa/TQ1SvYOaBuyrSKgEcro95LJyIsJ4vSo6BXIxOKxiJAat04CmST9Fw==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/pako": {
+      "version": "2.1.0",
+      "resolved": "https://registry.npmmirror.com/pako/-/pako-2.1.0.tgz",
+      "integrity": "sha512-w+eufiZ1WuJYgPXbV/PO3NCMEc3xqylkKHzp8bxp1uW4qaSNQUkwmLLEc3kKsfz8lpV1F8Ht3U1Cm+9Srog2ug==",
+      "license": "(MIT AND Zlib)"
+    }
+  }
+}

测试流式传输uniapp/package.json

@@ -0,0 +1,8 @@
+{
+  "dependencies": {
+    "pako": "^2.1.0"
+  },
+  "devDependencies": {
+    "@types/pako": "^2.0.4"
+  }
+}

测试流式传输uniapp/pages/index/ProtocolCodec.ts

@@ -0,0 +1,338 @@
+// 导入 GZIP 处理库（浏览器/Node.js 通用，需提前安装：npm install pako @types/pako）
+// import * as pako from "pako";
+
+/**
+ * 协议常量类：存储协议核心配置（不可修改，确保前后端一致）
+ */
+export class ProtocolConst {
+  /** 协议版本：v1（4位，对应字节3~4的低4位，剩余位保留用于后续升级） */
+  static readonly PROTOCOL_VERSION = 0b0001;
+
+  /** 头部固定长度：8字节（字节1~8，结构严格定义，不可修改） */
+  static readonly HEADER_SIZE = 8;
+
+  /** 最大包体大小：10MB（防止内存溢出，与 Python 端 MAX_BODY_SIZE 一致） */
+  static readonly MAX_BODY_SIZE = 1024 * 1024 * 10;
+
+  /** 字符串编码格式：UTF-8（统一前后端字符串编解码，避免乱码） */
+  static readonly STRING_ENCODING = "utf-8" as const; // as const 固定字符串类型，避免类型拓宽
+}
+
+/**
+ * 消息类型枚举（4位，存储在字节1高4位）
+ * 与 Python 端 MessageType 枚举值完全对齐，支持16种类型（当前用5种，剩余预留扩展）
+ */
+export enum MessageType {
+  PING = 0b0000, // 心跳消息（用于检测连接存活）
+  AUDIO_DATA = 0b0001, // 纯音频数据（PCM 二进制流，对应 RAW 序列化）
+  TEXT_MESSAGE = 0b0010, // 纯文本消息（对应 STRING 序列化，无 JSON 包装）
+  AUDIO_TEXT_MIX = 0b0011, // 音频+文本混合数据（复杂结构，对应 JSON 序列化）
+  CONTROL_CMD = 0b0100, // 控制指令（暂停/继续/停止等，对应 JSON 序列化）
+}
+
+/**
+ * 序列化方式枚举（4位，存储在字节2高4位）
+ * 按数据类型选择最优序列化方式，减少开销
+ */
+export enum SerializationType {
+  RAW = 0b0000, // 原始二进制（无需编解码，直接透传，适合音频等二进制数据）
+  JSON = 0b0001, // JSON 格式（适合字典、列表等复杂结构，如控制指令、混合数据）
+  STRING = 0b0010, // 直接字符串（UTF-8 编码，无 JSON 包装，适合纯文本消息）
+}
+
+/**
+ * 压缩方式枚举（4位，存储在字节2低4位）
+ * 按需选择压缩策略，平衡性能和传输体积
+ */
+export enum CompressionType {
+  NONE = 0b0000, // 无压缩（小数据包如控制指令、短文本，避免压缩开销）
+  GZIP = 0b0001, // GZIP 压缩（大数据包如长文本、混合数据，减少网络传输量）
+}
+
+/**
+ * 控制指令枚举（配合 MessageType.CONTROL_CMD 使用）
+ * 定义业务层面的控制指令，与 Python 端控制指令值一致
+ */
+export enum ControlCommand {
+  HEARTBEAT = 0b0001, // 心跳响应（回复 PING 消息）
+  PAUSE = 0b0010, // 暂停指令（如暂停音频播放）
+  RESUME = 0b0011, // 继续指令（如恢复音频播放）
+  STOP = 0b0100, // 停止指令（如停止音频传输）
+}
+
+/**
+ * 解包返回结果接口（强类型约束，明确返回数据结构）
+ * 让 IDE 提供自动提示，避免类型错误
+ */
+export interface UnpackedResult {
+  msgType: MessageType; // 消息类型（枚举值，便于逻辑判断）
+  msgTypeName: keyof typeof MessageType; // 消息类型名称（字符串，如 "TEXT_MESSAGE"，便于日志打印）
+  serialization: SerializationType; // 序列化方式（枚举值）
+  serializationName: keyof typeof SerializationType; // 序列化方式名称（字符串）
+  compression: CompressionType; // 压缩方式（枚举值）
+  compressionName: keyof typeof CompressionType; // 压缩方式名称（字符串）
+  body: Uint8Array | string | object | unknown[]; // 解包后的原始包体（根据序列化方式动态变化）
+}
+
+/**
+ * 打包入参类型别名（简化方法入参类型定义，提高可读性）
+ */
+// 包体支持的类型：二进制、字符串、对象、数组
+type PackBody = Uint8Array | string | object | unknown[];
+// 序列化方式可选值：枚举值、null、undefined（null/undefined 时自动推导）
+type OptionalSerialization = SerializationType | null | undefined;
+
+/**
+ * 协议编解码工具类（静态类，无需实例化，提供 pack/unpack 静态方法）
+ * 核心功能：将业务数据打包为符合协议的二进制包，或解析二进制包为业务数据
+ * 特点：与 Python/JS 端完全兼容，性能高效，类型安全
+ */
+export class ProtocolCodec {
+  /**
+   * 打包协议包：将业务数据按协议格式封装为二进制包
+   * @param msgType 消息类型（必须指定，枚举值）
+   * @param body 业务数据（根据消息类型对应不同类型，如字符串、对象、Uint8Array）
+   * @param serialization 序列化方式（可选，默认自动推导：纯文本→STRING，音频→RAW，复杂结构→JSON）
+   * @param compression 压缩方式（可选，默认无压缩）
+   * @returns 完整协议包（Uint8Array 类型，便于网络传输）
+   * @throws 类型错误、不支持的枚举值、包体过大等异常
+   */
+  static pack(
+    msgType: MessageType,
+    body: PackBody,
+    serialization: OptionalSerialization = null,
+    compression: CompressionType = CompressionType.NONE
+  ): Uint8Array {
+    // 1. 自动推导序列化方式（减少调用方心智负担，按消息类型默认最优解）
+    if (serialization === null || serialization === undefined) {
+      if (msgType === MessageType.AUDIO_DATA) {
+        // 音频数据→RAW 序列化（无需编解码，性能最优）
+        serialization = SerializationType.RAW;
+      } else if (msgType === MessageType.TEXT_MESSAGE) {
+        // 纯文本消息→STRING 序列化（直接 UTF-8 编码，无冗余）
+        serialization = SerializationType.STRING;
+      } else if (msgType === MessageType.CONTROL_CMD || msgType === MessageType.AUDIO_TEXT_MIX) {
+        // 控制指令/混合数据→JSON 序列化（支持复杂结构）
+        serialization = SerializationType.JSON;
+      } else {
+        // 非法消息类型：抛出异常，提前阻断错误
+        throw new Error(`不支持的消息类型：${MessageType[msgType]}（值：${msgType}）`);
+      }
+    }
+
+    // 2. 序列化包体：将业务数据转为二进制（按序列化方式处理）
+    let serializedBody: Uint8Array; // 序列化后的二进制数据
+    const textEncoder = new TextEncoder(ProtocolConst.STRING_ENCODING); // UTF-8 编码器
+
+    switch (serialization) {
+      case SerializationType.RAW:
+        // RAW 序列化：必须传入 Uint8Array（二进制数据直接透传）
+        if (!(body instanceof Uint8Array)) {
+          throw new TypeError(`RAW 序列化要求 body 必须是 Uint8Array 类型，当前传入：${typeof body}`);
+        }
+        serializedBody = body;
+        break;
+
+      case SerializationType.STRING:
+        // STRING 序列化：必须传入字符串，直接 UTF-8 编码（无 JSON 包装）
+        if (typeof body !== "string") {
+          throw new TypeError(`STRING 序列化要求 body 必须是 string 类型，当前传入：${typeof body}`);
+        }
+        serializedBody = textEncoder.encode(body);
+        break;
+
+      case SerializationType.JSON:
+        // JSON 序列化：支持字符串、对象、数组，转为 JSON 字符串后 UTF-8 编码
+        if (typeof body === "string") {
+          // 已为字符串，直接编码
+          serializedBody = textEncoder.encode(body);
+        } else if (typeof body === "object" && body !== null) {
+          // 对象/数组→JSON 字符串→编码（用 JSON.stringify 序列化）
+          const jsonStr = JSON.stringify(body);
+          serializedBody = textEncoder.encode(jsonStr);
+        } else {
+          throw new TypeError(`JSON 序列化要求 body 必须是 string/object/array 类型，当前传入：${typeof body}`);
+        }
+        break;
+
+      default:
+        // 非法序列化方式：抛出异常
+        throw new Error(`不支持的序列化方式：${SerializationType[serialization]}（值：${serialization}）`);
+    }
+
+    // 3. 压缩包体：按指定压缩方式处理（GZIP 或无压缩）
+    let compressedBody: Uint8Array; // 压缩后的二进制数据
+    if (compression === CompressionType.NONE) {
+      // 无压缩：直接透传序列化后的二进制数据
+      compressedBody = serializedBody;
+    } else {
+      // 非法压缩方式：抛出异常
+      throw new Error(`不支持的压缩方式：${CompressionType[compression]}（值：${compression}）`);
+    }
+
+    // 4. 校验包体大小：防止超过最大限制（避免内存溢出）
+    const bodyLen = compressedBody.length; // 压缩后的包体长度
+    if (bodyLen > ProtocolConst.MAX_BODY_SIZE) {
+      throw new Error(
+        `包体过大（${bodyLen}字节），最大支持${ProtocolConst.MAX_BODY_SIZE}字节（10MB）`
+      );
+    }
+
+    // 5. 构造头部：8字节固定结构（严格按协议定义，大端序，与 Python 端一致）
+    const header = new Uint8Array(ProtocolConst.HEADER_SIZE); // 头部缓冲区（8字节）
+
+    // 字节1：消息类型(4位) + 保留位(4位)
+    // 消息类型左移4位（占高4位），保留位填0（后续扩展用）
+    header[0] = (msgType << 4) | 0x00;
+
+    // 字节2：序列化方式(4位) + 压缩方式(4位)
+    // 序列化方式左移4位（占高4位），压缩方式占低4位（&0x0F 确保仅4位）
+    header[1] = (serialization << 4) | (compression & 0x0F);
+
+    // 字节3~4：协议版本（16位大端序）
+    // 大端序：高位字节在前，低位字节在后（网络传输标准字节序）
+    header[2] = (ProtocolConst.PROTOCOL_VERSION >> 8) & 0xFF; // 版本高8位（当前版本为1，高8位为0）
+    header[3] = ProtocolConst.PROTOCOL_VERSION & 0xFF; // 版本低8位（当前版本为1）
+
+    // 字节5~8：包体长度（32位大端序）
+    // 32位整数：最高位字节（第5字节）→ 最低位字节（第8字节）
+    header[4] = (bodyLen >> 24) & 0xFF; // 包体长度第24~31位
+    header[5] = (bodyLen >> 16) & 0xFF; // 包体长度第16~23位
+    header[6] = (bodyLen >> 8) & 0xFF; // 包体长度第8~15位
+    header[7] = bodyLen & 0xFF; // 包体长度第0~7位
+
+    // 6. 拼接头部和包体：生成完整协议包
+    const totalLen = ProtocolConst.HEADER_SIZE + bodyLen; // 总长度 = 头部8字节 + 包体长度
+    const packet = new Uint8Array(totalLen); // 完整包缓冲区
+
+    packet.set(header, 0); // 从索引0开始写入头部（占前8字节）
+    packet.set(compressedBody, ProtocolConst.HEADER_SIZE); // 从索引8开始写入包体
+
+    return packet; // 返回完整协议包（Uint8Array，网络传输高效）
+  }
+
+  /**
+   * 解包协议包：将二进制协议包解析为业务数据
+   * @param packet 完整协议包（网络接收的二进制数据，支持 Uint8Array 或 ArrayBuffer）
+   * @returns 结构化解包结果（UnpackedResult 接口，包含消息类型、包体等信息）
+   * @throws 包长度过短、版本不匹配、解压缩失败、反序列化失败等异常
+   */
+  static unpack(packet: Uint8Array | ArrayBuffer): UnpackedResult {
+    // 统一数据类型：将 ArrayBuffer 转为 Uint8Array（便于按字节操作）
+    const uint8Packet = packet instanceof ArrayBuffer 
+      ? new Uint8Array(packet) 
+      : packet;
+
+    // 1. 校验包长度：至少包含8字节头部
+    if (uint8Packet.length < ProtocolConst.HEADER_SIZE) {
+      throw new Error(
+        `包长度过短（${uint8Packet.length}字节），至少需要${ProtocolConst.HEADER_SIZE}字节头部`
+      );
+    }
+
+    // 2. 拆分头部和包体：前8字节为头部，剩余为包体
+    const header = uint8Packet.subarray(0, ProtocolConst.HEADER_SIZE); // 头部（0~7索引）
+    const bodyBuffer = uint8Packet.subarray(ProtocolConst.HEADER_SIZE); // 包体（8索引开始）
+
+    // 3. 解析头部字段（按协议结构逐字节解析，大端序）
+    // 字节1：消息类型(4位) + 保留位(4位)
+    const byte1 = header[0];
+    const msgType = (byte1 >> 4) & 0x0F; // 右移4位取高4位（消息类型），&0x0F 确保仅4位
+    // 校验消息类型合法性（防止非法包注入）
+    if (!Object.values(MessageType).includes(msgType as MessageType)) {
+      throw new Error(`非法消息类型：${msgType}（无对应枚举值）`);
+    }
+
+    // 字节2：序列化方式(4位) + 压缩方式(4位)
+    const byte2 = header[1];
+    const serialization = (byte2 >> 4) & 0x0F; // 右移4位取高4位（序列化方式）
+    const compression = byte2 & 0x0F; // 取低4位（压缩方式）
+    // 校验序列化方式合法性
+    if (!Object.values(SerializationType).includes(serialization as SerializationType)) {
+      throw new Error(`非法序列化方式：${serialization}（无对应枚举值）`);
+    }
+    // 校验压缩方式合法性
+    if (!Object.values(CompressionType).includes(compression as CompressionType)) {
+      throw new Error(`非法压缩方式：${compression}（无对应枚举值）`);
+    }
+
+    // 字节3~4：协议版本（16位大端序）
+    const version = (header[2] << 8) | header[3]; // 大端序解析：高位字节<<8 + 低位字节
+    // 校验版本兼容性（仅支持当前协议版本）
+    if (version !== ProtocolConst.PROTOCOL_VERSION) {
+      throw new Error(
+        `协议版本不匹配：收到v${version}，当前支持v${ProtocolConst.PROTOCOL_VERSION}`
+      );
+    }
+
+    // 字节5~8：包体长度（32位大端序）
+    const bodyLen = (header[4] << 24) | (header[5] << 16) | (header[6] << 8) | header[7];
+    // 校验包体长度一致性（头部声明长度 vs 实际包体长度）
+    if (bodyBuffer.length !== bodyLen) {
+      throw new Error(
+        `包体长度不匹配：头部声明${bodyLen}字节，实际接收${bodyBuffer.length}字节（可能包丢失或非法包）`
+      );
+    }
+
+    // 4. 解压包体：按头部指定的压缩方式解压
+    let decompressedBody: Uint8Array; // 解压后的二进制数据
+    if (compression === CompressionType.NONE) {
+      // 无压缩：直接透传包体数据
+      decompressedBody = bodyBuffer;
+    } else {
+      // 此处理论上不会触发（已提前校验压缩方式合法性）
+      throw new Error(`不支持的压缩方式：${CompressionType[compression]}（值：${compression}）`);
+    }
+
+    // 5. 反序列化包体：将二进制数据转为业务数据（按序列化方式处理）
+    let body: UnpackedResult["body"]; // 反序列化后的原始业务数据
+    const textDecoder = new TextDecoder(ProtocolConst.STRING_ENCODING); // UTF-8 解码器
+
+    switch (serialization) {
+      case SerializationType.RAW:
+        // RAW 反序列化：直接返回 Uint8Array（二进制数据，如音频 PCM）
+        body = decompressedBody;
+        break;
+
+      case SerializationType.STRING:
+        // STRING 反序列化：UTF-8 解码为字符串（无 JSON 解析步骤）
+        try {
+          body = textDecoder.decode(decompressedBody);
+        } catch (e) {
+          throw new Error(`STRING 反序列化失败：${(e as Error).message}（UTF-8 解码错误，可能是非法字符串数据）`);
+        }
+        break;
+
+      case SerializationType.JSON:
+        // JSON 反序列化：先 UTF-8 解码为字符串，再 JSON.parse 转为对象/数组
+        try {
+          const jsonStr = textDecoder.decode(decompressedBody); // 二进制→JSON 字符串
+          body = JSON.parse(jsonStr); // JSON 字符串→对象/数组
+        } catch (e) {
+          if (e instanceof TypeError) {
+            throw new Error(`JSON 反序列化失败：${ProtocolConst.STRING_ENCODING} 解码错误（非法 UTF-8 数据）`);
+          } else if (e instanceof SyntaxError) {
+            throw new Error(`JSON 反序列化失败：格式错误（${(e as Error).message}），请检查 JSON 语法`);
+          } else {
+            throw new Error(`JSON 反序列化失败：${(e as Error).message}`);
+          }
+        }
+        break;
+
+      default:
+        // 此处理论上不会触发（已提前校验序列化方式合法性）
+        throw new Error(`不支持的序列化方式：${SerializationType[serialization]}（值：${serialization}）`);
+    }
+
+    // 6. 构造并返回解包结果（利用枚举反向映射获取名称，便于日志和展示）
+    return {
+      msgType: msgType as MessageType, // 消息类型枚举值
+      msgTypeName: MessageType[msgType] as keyof typeof MessageType, // 消息类型名称（如 "TEXT_MESSAGE"）
+      serialization: serialization as SerializationType, // 序列化方式枚举值
+      serializationName: SerializationType[serialization] as keyof typeof SerializationType, // 序列化方式名称
+      compression: compression as CompressionType, // 压缩方式枚举值
+      compressionName: CompressionType[compression] as keyof typeof CompressionType, // 压缩方式名称
+      body: body, // 反序列化后的原始业务数据
+    };
+  }
+}

测试流式传输uniapp/pages/index/index - 副本.vue

@@ -0,0 +1,237 @@
+<template>
+	<view class="container">
+		<button @click="onStartRecord" :disabled="isRecording">开始录音</button>
+		<button @click="onStopRecord" :disabled="!isRecording">停止录音</button>
+		<view class="tip">{{ status }}</view>
+		<view class="tip">当前分贝：{{ currentDecibels }}</view>
+		<yao-RecordFrame ref="recordFrame" @onFrameRecorded="frameRecorded" @currentDecibels="onCurrentDecibels"
+			@onStop="stopIt">
+		</yao-RecordFrame>
+		<sdx-StreamPlayer>xx</sdx-StreamPlayer>
+	</view>
+</template>
+
+<script>
+	export default {
+		data() {
+			return {
+				status: "未录音",
+				currentDecibels: 0,
+				isRecording: false,
+				ws: null, // WebSocket 实例
+				audioContext: null, // 音频上下文
+				scriptProcessor: null, // 音频处理节点
+				audioBufferSource: null, // 音频源节点
+				frameBufferList: [], // 缓存音频帧
+				wsUrl: "ws://172.16.89.58:8000/ws/audio", // 替换为实际后端地址
+			};
+		},
+		onUnload() {
+			// 页面卸载时清理资源
+			this.stopRecordAndClean();
+		},
+		methods: {
+
+
+			// 申请录音权限
+			async applyRecordPermission() {
+				try {
+					const res = await uni.requestPermissions({
+						scope: "scope.record"
+					});
+					const isGranted = res[0].grantStatus === 1;
+					if (!isGranted) {
+						uni.showToast({
+							title: "请授予录音权限",
+							icon: "none"
+						});
+					}
+					return isGranted;
+				} catch (e) {
+					console.error("申请权限失败：", e);
+					return false;
+				}
+			},
+			// ArrayBuffer 转字符串
+			// 兼容的 ArrayBuffer 转字符串方法
+			arrayBufferToStringCompat(buffer) {
+			    // 方法1: 使用 String.fromCharCode 和 Uint8Array
+			    const uint8Array = new Uint8Array(buffer);
+			    let str = '';
+			    for (let i = 0; i < uint8Array.length; i++) {
+			        str += String.fromCharCode(uint8Array[i]);
+			    }
+			    return str;
+			    
+			    // 方法2: 或者使用更简洁的方式
+			    // return String.fromCharCode.apply(null, new Uint8Array(buffer));
+			},
+			// 开始录音
+			async onStartRecord() {
+				// 1. 申请权限
+				// const hasPermission = await this.applyRecordPermission();
+				// if (!hasPermission) return;
+				this.ws = uni.connectSocket({
+					url: this.wsUrl, //仅为示例，并非真实接口地址。
+					complete: () => {
+						console.log('complete');
+					},
+					success: () => {
+						console.log('web');
+						this.ws.onOpen((res) => {
+							console.log('WebSocket连接已打开', res);
+							this.ws.onMessage((res) => {
+								let messageData = res.data;
+								// 处理不同类型的数据
+								if (typeof messageData === 'string') {
+									// 文本数据
+									try {
+										const parsedData = JSON.parse(messageData);
+										this.handleMessage(parsedData);
+									} catch (e) {
+										this.handleMessage(messageData);
+									}
+								} else if (messageData instanceof ArrayBuffer) {
+									// 二进制数据
+									try {
+										
+										// const str = this.arrayBufferToStringCompat(messageData);
+										// console.log('转换后的字符串:', str);
+										
+									} catch (e) {
+										console.log('二进制数据解析失败:', e);
+									}
+								}
+							})
+						});
+					},
+					fail: () => {
+						console.log('fail');
+					},
+				});
+
+				// await this.initWebSocket();
+
+				// 3. 启动录音
+				try {
+					this.$refs.recordFrame.start({
+						sampleRate: 16000,
+						frameSize: 1024,
+						gain: 1.0,
+						onFrameRecorded: ({
+							isLastFrame,
+							frameBuffer
+						}) => {
+							// 处理帧数据（如实时上传/渲染）
+							// console.log('帧数据:', frameData);
+							this.ws.send({
+								data: frameBuffer
+							});
+						},
+						onDecibels: (decibels) => {
+							// 处理分贝数据（如实时更新UI）
+							// console.log('当前分贝:', decibels);
+							this.currentDecibels = decibels;
+						}
+					});
+					this.isRecording = true;
+					this.status = "录音中...";
+				} catch (e) {
+					console.error("启动录音失败：", e);
+					this.status = "启动录音失败";
+					this.isRecording = false;
+				}
+			},
+			
+			// 停止录音
+			onStopRecord() {
+				this.stopRecordAndClean();
+			},
+
+			// 停止录音并清理资源
+			stopRecordAndClean() {
+				if (this.isRecording) {
+					// 停止录音组件
+					this.$refs.recordFrame.stop();
+					this.isRecording = false;
+					this.status = "已停止录音";
+				}
+
+				// 关闭 WebSocket
+				if (this.ws) {
+					this.ws.close();
+					this.ws = null;
+				}
+
+				// 清理音频上下文
+				if (this.audioContext) {
+					this.audioContext.close();
+					this.audioContext = null;
+					this.scriptProcessor = null;
+					this.frameBufferList = [];
+				}
+
+				// 重置状态
+				this.currentDecibels = 0;
+			},
+
+			// 接收音频帧并处理
+			frameRecorded({
+				isLastFrame,
+				frameBuffer
+			}) {
+				// console.log("收到音频帧：", isLastFrame, frameBuffer.length);
+
+				// 2. 通过 WebSocket 发送给后端
+				if (this.ws) {
+					try {
+						this.ws.send({
+							data: frameBuffer
+						});
+					} catch (e) {
+						console.error("发送音频帧失败：", e);
+					}
+				}
+
+
+			},
+
+			// 监听分贝值
+			onCurrentDecibels(decibels) {
+				// this.currentDecibels = decibels.toFixed(2);
+				// console.log("当前分贝：", this.currentDecibels);
+			},
+
+			// 录音停止回调
+			stopIt(base64) {
+				this.stopRecordAndClean();
+				console.log("录音停止，最终音频Base64：", base64?.substring(0, 50) + "...");
+			},
+		},
+	};
+</script>
+
+<style scoped>
+	.container {
+		padding: 20rpx;
+	}
+
+	button {
+		margin: 10rpx 0;
+		padding: 15rpx 30rpx;
+		background: #007aff;
+		color: white;
+		border: none;
+		border-radius: 8rpx;
+	}
+
+	button:disabled {
+		background: #ccc;
+	}
+
+	.tip {
+		margin: 15rpx 0;
+		font-size: 28rpx;
+		color: #333;
+	}
+</style>

测试流式传输uniapp/pages/index/index.vue

@@ -7,7 +7,7 @@
 		<yao-RecordFrame ref="recordFrame" @onFrameRecorded="frameRecorded" @currentDecibels="onCurrentDecibels"
 			@onStop="stopIt">
 		</yao-RecordFrame>
-		<sdx-StreamPlayer>xx</sdx-StreamPlayer>
+		<sdx-StreamPlayer ref="aaa">xx</sdx-StreamPlayer>
 	</view>
 </template>
 
@@ -23,7 +23,7 @@
 				scriptProcessor: null, // 音频处理节点
 				audioBufferSource: null, // 音频源节点
 				frameBufferList: [], // 缓存音频帧
-				wsUrl: "ws://10.10.10.201:8000/ws/audio", // 替换为实际后端地址
+				wsUrl: "ws://172.16.89.58:8000/ws/audio", // 替换为实际后端地址
 			};
 		},
 		onUnload() {
@@ -80,6 +80,8 @@
 						console.log('web');
 						this.ws.onOpen((res) => {
 							console.log('WebSocket连接已打开', res);
+							
+
 							this.ws.onMessage((res) => {
 								let messageData = res.data;
 								// 处理不同类型的数据
@@ -94,7 +96,7 @@
 								} else if (messageData instanceof ArrayBuffer) {
 									// 二进制数据
 									try {
-										
+										this.$refs.aaa.appendBuffer(messageData)
 										// const str = this.arrayBufferToStringCompat(messageData);
 										// console.log('转换后的字符串:', str);
 										
@@ -142,7 +144,7 @@
 					this.isRecording = false;
 				}
 			},
-
+			
 			// 停止录音
 			onStopRecord() {
 				this.stopRecordAndClean();

测试流式传输uniapp/uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue

@@ -21,22 +21,8 @@
 		},
 		methods: {
 			connect() {
-				// 这里链接socket或者sse,并把接受到的二进制数据不断调用即可, 具体的ws请自行封装
-				// ws = uni.connectSocket({
-				// 	url: 'ws://10.10.10.201:8000/ws/audio',
-				// 	method: 'GET',
-				// 	success() {},
-				// 	fail(e) {
-				// 		console.log(e);
-				// 	}
-				// });
-				// ws.onMessage(arrayBuffer => {
-				// 	console.log('arrayBuffer', arrayBuffer);
-				// 	this.isStreamPlaying = true
-				// 	this.currBuffer= uni.arrayBufferToBase64(arrayBuffer)
-				// })
 				this.ws = uni.connectSocket({
-					url: 'ws://10.10.10.201:8000/ws/audio',
+					url: 'ws://172.16.89.58:8000/ws/audio',
 					complete: () => {
 						console.log('complete');
 					},
@@ -60,10 +46,8 @@
 									this.isStreamPlaying = true
 									this.currBuffer= uni.arrayBufferToBase64(messageData)
 									try {
-										
 										// const str = this.arrayBufferToStringCompat(messageData);
 										// console.log('转换后的字符串:', str);
-										
 									} catch (e) {
 										console.log('二进制数据解析失败:', e);
 									}
@@ -77,6 +61,10 @@
 			},
 			changeStreamPlaying(e) {
 				this.isStreamPlaying = e.type
+			},
+			appendBuffer(messageData) {
+				this.isStreamPlaying = true
+				this.currBuffer= uni.arrayBufferToBase64(messageData)
 			}
 		}
 	}
@@ -99,6 +87,7 @@
 		},
 		methods: {
 			playTTS (base64) {
+				console.log('base64', base64);
 				if(!base64)return
 				const binaryStr = atob(base64)
 				const bytes = new Uint8Array(binaryStr.length)

测试流式传输uniapp/uni_modules/sdx-StreamPlayer/plugins/StreamPlayer.js

@@ -30,6 +30,7 @@ export class StreamPlayer {
 	}
 
 	appendChunk(audioData) {
+		console.log('this.audioQueue', this.audioQueue.length);
 		this.audioQueue.push(audioData);
 		this._processQueue();
 	}

测试流式传输uniapp/uni_modules/yao-RecordFrame/components/yao-RecordFrame/yao-RecordFrame.vue

@@ -1,6 +1,5 @@
 <template>
 	<view :options="options" :change:options="record.startRecord" :status="status" :change:status="record.onStop">
-
 	</view>
 </template>
 <script>

测试流式传输uniapp/unpackage/dist/dev/app-plus/app-renderjs.js

@@ -313,7 +313,7 @@ __renderjsModules["5f91482f"] = (() => {
     default: () => stdin_default
   });
 
-  // F:/aistream-test/测试流式传输uniapp/uni_modules/sdx-StreamPlayer/plugins/StreamPlayer.js
+  // C:/Users/1/Desktop/testAudio/测试流式传输uniapp/uni_modules/sdx-StreamPlayer/plugins/StreamPlayer.js
   var StreamPlayer = class {
     constructor({
       inputSampleRate = 16e3,
@@ -343,6 +343,7 @@ __renderjsModules["5f91482f"] = (() => {
       this.callback = callback;
     }
     appendChunk(audioData) {
+      console.log("this.audioQueue", this.audioQueue.length);
       this.audioQueue.push(audioData);
       this._processQueue();
     }
@@ -496,6 +497,7 @@ __renderjsModules["5f91482f"] = (() => {
     },
     methods: {
       playTTS(base64) {
+        console.log("base64", base64);
         if (!base64)
           return;
         const binaryStr = atob(base64);

测试流式传输uniapp/unpackage/dist/dev/app-plus/app-service.js

@@ -121,7 +121,7 @@ if (uni.restoreGlobal) {
   }
   if (typeof block0$1 === "function")
     block0$1(_sfc_main$3);
-  const __easycom_0 = /* @__PURE__ */ _export_sfc(_sfc_main$3, [["render", _sfc_render$2], ["__file", "F:/aistream-test/测试流式传输uniapp/uni_modules/yao-RecordFrame/components/yao-RecordFrame/yao-RecordFrame.vue"]]);
+  const __easycom_0 = /* @__PURE__ */ _export_sfc(_sfc_main$3, [["render", _sfc_render$2], ["__file", "C:/Users/1/Desktop/testAudio/测试流式传输uniapp/uni_modules/yao-RecordFrame/components/yao-RecordFrame/yao-RecordFrame.vue"]]);
   const block0 = (Comp) => {
     (Comp.$renderjs || (Comp.$renderjs = [])).push("renderJS");
     (Comp.$renderjsModules || (Comp.$renderjsModules = {}))["renderJS"] = "5f91482f";
@@ -137,14 +137,14 @@ if (uni.restoreGlobal) {
     methods: {
       connect() {
         this.ws = uni.connectSocket({
-          url: "ws://10.10.10.201:8000/ws/audio",
+          url: "ws://172.16.89.58:8000/ws/audio",
           complete: () => {
-            formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:41", "complete");
+            formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:27", "complete");
           },
           success: () => {
-            formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:44", "web");
+            formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:30", "web");
             this.ws.onOpen((res) => {
-              formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:46", "WebSocket连接已打开", res);
+              formatAppLog("log", "at uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue:32", "WebSocket连接已打开", res);
               this.ws.onMessage((res2) => {
                 let messageData = res2.data;
                 if (typeof messageData === "string") {
@@ -168,6 +168,10 @@ if (uni.restoreGlobal) {
       },
       changeStreamPlaying(e) {
         this.isStreamPlaying = e.type;
+      },
+      appendBuffer(messageData) {
+        this.isStreamPlaying = true;
+        this.currBuffer = uni.arrayBufferToBase64(messageData);
       }
     }
   };
@@ -200,7 +204,7 @@ if (uni.restoreGlobal) {
   }
   if (typeof block0 === "function")
     block0(_sfc_main$2);
-  const __easycom_1 = /* @__PURE__ */ _export_sfc(_sfc_main$2, [["render", _sfc_render$1], ["__file", "F:/aistream-test/测试流式传输uniapp/uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue"]]);
+  const __easycom_1 = /* @__PURE__ */ _export_sfc(_sfc_main$2, [["render", _sfc_render$1], ["__file", "C:/Users/1/Desktop/testAudio/测试流式传输uniapp/uni_modules/sdx-StreamPlayer/components/sdx-StreamPlayer/sdx-StreamPlayer.vue"]]);
   const _sfc_main$1 = {
     data() {
       return {
@@ -217,7 +221,7 @@ if (uni.restoreGlobal) {
         // 音频源节点
         frameBufferList: [],
         // 缓存音频帧
-        wsUrl: "ws://10.10.10.201:8000/ws/audio"
+        wsUrl: "ws://172.16.89.58:8000/ws/audio"
         // 替换为实际后端地址
       };
     },
@@ -275,12 +279,18 @@ if (uni.restoreGlobal) {
                   } catch (e) {
                     this.handleMessage(messageData);
                   }
+                } else if (messageData instanceof ArrayBuffer) {
+                  try {
+                    this.$refs.aaa.appendBuffer(messageData);
+                  } catch (e) {
+                    formatAppLog("log", "at pages/index/index.vue:104", "二进制数据解析失败:", e);
+                  }
                 }
               });
             });
           },
           fail: () => {
-            formatAppLog("log", "at pages/index/index.vue:109", "fail");
+            formatAppLog("log", "at pages/index/index.vue:111", "fail");
           }
         });
         try {
@@ -303,7 +313,7 @@ if (uni.restoreGlobal) {
           this.isRecording = true;
           this.status = "录音中...";
         } catch (e) {
-          formatAppLog("error", "at pages/index/index.vue:140", "启动录音失败：", e);
+          formatAppLog("error", "at pages/index/index.vue:142", "启动录音失败：", e);
           this.status = "启动录音失败";
           this.isRecording = false;
         }
@@ -342,7 +352,7 @@ if (uni.restoreGlobal) {
               data: frameBuffer
             });
           } catch (e) {
-            formatAppLog("error", "at pages/index/index.vue:192", "发送音频帧失败：", e);
+            formatAppLog("error", "at pages/index/index.vue:194", "发送音频帧失败：", e);
           }
         }
       },
@@ -352,7 +362,7 @@ if (uni.restoreGlobal) {
       // 录音停止回调
       stopIt(base64) {
         this.stopRecordAndClean();
-        formatAppLog("log", "at pages/index/index.vue:208", "录音停止，最终音频Base64：", (base64 == null ? void 0 : base64.substring(0, 50)) + "...");
+        formatAppLog("log", "at pages/index/index.vue:210", "录音停止，最终音频Base64：", (base64 == null ? void 0 : base64.substring(0, 50)) + "...");
       }
     }
   };
@@ -388,16 +398,22 @@ if (uni.restoreGlobal) {
         onCurrentDecibels: $options.onCurrentDecibels,
         onOnStop: $options.stopIt
       }, null, 8, ["onOnFrameRecorded", "onCurrentDecibels", "onOnStop"]),
-      vue.createVNode(_component_sdx_StreamPlayer, null, {
-        default: vue.withCtx(() => [
-          vue.createTextVNode("xx")
-        ]),
-        _: 1
-        /* STABLE */
-      })
+      vue.createVNode(
+        _component_sdx_StreamPlayer,
+        { ref: "aaa" },
+        {
+          default: vue.withCtx(() => [
+            vue.createTextVNode("xx")
+          ]),
+          _: 1
+          /* STABLE */
+        },
+        512
+        /* NEED_PATCH */
+      )
     ]);
   }
-  const PagesIndexIndex = /* @__PURE__ */ _export_sfc(_sfc_main$1, [["render", _sfc_render], ["__scopeId", "data-v-1cf27b2a"], ["__file", "F:/aistream-test/测试流式传输uniapp/pages/index/index.vue"]]);
+  const PagesIndexIndex = /* @__PURE__ */ _export_sfc(_sfc_main$1, [["render", _sfc_render], ["__scopeId", "data-v-1cf27b2a"], ["__file", "C:/Users/1/Desktop/testAudio/测试流式传输uniapp/pages/index/index.vue"]]);
   __definePage("pages/index/index", PagesIndexIndex);
   const _sfc_main = {
     onLaunch: function() {
@@ -410,7 +426,7 @@ if (uni.restoreGlobal) {
       formatAppLog("log", "at App.vue:10", "App Hide");
     }
   };
-  const App = /* @__PURE__ */ _export_sfc(_sfc_main, [["__file", "F:/aistream-test/测试流式传输uniapp/App.vue"]]);
+  const App = /* @__PURE__ */ _export_sfc(_sfc_main, [["__file", "C:/Users/1/Desktop/testAudio/测试流式传输uniapp/App.vue"]]);
   function createApp() {
     const app = vue.createVueApp(App);
     return {