Kris
7a972055b7
新增大量渠道适配器相关的协议、策略、工具类与基础设施代码，包括： 1. 多协议定义：认证、消息、配置、网关等核心接口 2. 策略模块：上下文、群聊、去重、防抖等业务策略 3. 工具集：重试、去重、文本分块、消息格式化等SDK工具 4. 基础设施：外部进程管理、事件广播、熔断机制等 5. 账户与管道系统：账户管理、消息处理管道实现 6. 运行时服务：状态收集、维护任务、日志等后台服务
61 lines
1.6 KiB
Python
61 lines
1.6 KiB
Python
from abc import ABC, abstractmethod
|
|
|
|
from pydantic import BaseModel
|
|
|
|
from yuxi.utils.logging_config import logger
|
|
|
|
|
|
class VoiceSegment(BaseModel):
|
|
audio_data: bytes
|
|
sample_rate: int = 16000
|
|
channels: int = 1
|
|
encoding: str = "pcm_s16le"
|
|
duration_ms: int = 0
|
|
|
|
|
|
class STTResult(BaseModel):
|
|
text: str
|
|
confidence: float = 0.0
|
|
language: str | None = None
|
|
|
|
|
|
class STTProvider(ABC):
|
|
@abstractmethod
|
|
async def transcribe(self, segment: VoiceSegment) -> STTResult: ...
|
|
|
|
|
|
class TTSProvider(ABC):
|
|
@abstractmethod
|
|
async def synthesize(self, text: str, voice: str = "default") -> bytes: ...
|
|
|
|
|
|
class VoicePolicy:
|
|
def __init__(self):
|
|
self._stt_provider: STTProvider | None = None
|
|
self._tts_provider: TTSProvider | None = None
|
|
|
|
def configure_stt(self, provider: STTProvider) -> None:
|
|
self._stt_provider = provider
|
|
|
|
def configure_tts(self, provider: TTSProvider) -> None:
|
|
self._tts_provider = provider
|
|
|
|
async def process_voice(self, segment: VoiceSegment) -> str | None:
|
|
if not self._stt_provider:
|
|
return None
|
|
try:
|
|
result = await self._stt_provider.transcribe(segment)
|
|
except Exception as e:
|
|
logger.error(f"STT transcription failed: {e}")
|
|
return None
|
|
return result.text or None
|
|
|
|
async def text_to_speech(self, text: str) -> bytes | None:
|
|
if not self._tts_provider:
|
|
return None
|
|
try:
|
|
return await self._tts_provider.synthesize(text)
|
|
except Exception as e:
|
|
logger.error(f"TTS synthesis failed: {e}")
|
|
return None
|