Merge pull request #19 from songguocola/main

[Add] Multimodal Dialog API

Author: kevinlin09

dashscope/multimodal/__init__.py

@@ -0,0 +1,56 @@
+# dialog_state.py
+
+from enum import Enum
+
+
+class DialogState(Enum):
+    """
+    对话状态枚举类，定义了对话机器人可能处于的不同状态。
+
+    Attributes:
+        IDLE (str): 表示机器人处于空闲状态。
+        LISTENING (str): 表示机器人正在监听用户输入。
+        THINKING (str): 表示机器人正在思考。
+        RESPONDING (str): 表示机器人正在生成或回复中。
+    """
+    IDLE = 'Idle'
+    LISTENING = 'Listening'
+    THINKING = 'Thinking'
+    RESPONDING = 'Responding'
+
+
+class StateMachine:
+    """
+    状态机类，用于管理机器人的状态转换。
+
+    Attributes:
+        current_state (DialogState): 当前状态。
+    """
+
+    def __init__(self):
+        # 初始化状态机时设置初始状态为IDLE
+        self.current_state = DialogState.IDLE
+
+    def change_state(self, new_state: str) -> None:
+        """
+        更改当前状态到指定的新状态。
+
+        Args:
+            new_state (str): 要切换到的新状态。
+
+        Raises:
+            ValueError: 如果尝试切换到一个无效的状态，则抛出此异常。
+        """
+        if new_state in [state.value for state in DialogState]:
+            self.current_state = DialogState(new_state)
+        else:
+            raise ValueError("无效的状态类型")
+
+    def get_current_state(self) -> DialogState:
+        """
+        获取当前状态。
+
+        Returns:
+            DialogState: 当前状态。
+        """
+        return self.current_state

dashscope/multimodal/dialog_state.py

@@ -0,0 +1,27 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+# -*- coding: utf-8 -*-
+
+# multimodal conversation request directive
+
+class RequestToRespondType:
+    TRANSCRIPT = 'transcript'
+    PROMPT = 'prompt'
+
+
+# multimodal conversation response directive
+RESPONSE_NAME_TASK_STARTED = "task-started"
+RESPONSE_NAME_RESULT_GENERATED = "result-generated"
+RESPONSE_NAME_TASK_FINISHED = "task-finished"
+
+RESPONSE_NAME_TASK_FAILED = "TaskFailed"
+RESPONSE_NAME_STARTED = "Started"
+RESPONSE_NAME_STOPPED = "Stopped"
+RESPONSE_NAME_STATE_CHANGED = "DialogStateChanged"
+RESPONSE_NAME_REQUEST_ACCEPTED = "RequestAccepted"
+RESPONSE_NAME_SPEECH_STARTED = "SpeechStarted"
+RESPONSE_NAME_SPEECH_ENDED = "SpeechEnded"  # 服务端检测到asr语音尾点时下发此事件,可选事件
+RESPONSE_NAME_RESPONDING_STARTED = "RespondingStarted"  # AI语音应答开始，sdk要准备接收服务端下发的语音数据
+RESPONSE_NAME_RESPONDING_ENDED = "RespondingEnded"  # AI语音应答结束
+RESPONSE_NAME_SPEECH_CONTENT = "SpeechContent"  # 用户语音识别出的文本，流式全量输出
+RESPONSE_NAME_RESPONDING_CONTENT = "RespondingContent"  # 统对外输出的文本，流式全量输出
+RESPONSE_NAME_ERROR = "Error"  # 服务端对话中报错