operating-room-monitor-server/app/services/baidu_speech.py

from __future__ import annotations

from threading import Lock
from typing import Any

from aip import AipSpeech

from app.config import Settings, settings as _default_settings


class BaiduSpeechNotConfiguredError(RuntimeError):
    """未配置 BAIDU_SPEECH_APP_ID / API_KEY / SECRET_KEY 时调用接口会抛出。"""


class BaiduSpeechService:
    """百度短语音识别（asr）与在线语音合成（synthesis），基于 `baidu-aip` 的 `AipSpeech`。"""

    def __init__(self, app_settings: Settings | None = None) -> None:
        self._s = app_settings or _default_settings
        self._client: AipSpeech | None = None
        self._lock = Lock()

    @property
    def configured(self) -> bool:
        return self._s.baidu_speech_configured

    def _client_or_raise(self) -> AipSpeech:
        if not self.configured:
            raise BaiduSpeechNotConfiguredError(
                "百度语音未配置：请设置 BAIDU_SPEECH_APP_ID、BAIDU_SPEECH_API_KEY、"
                "BAIDU_SPEECH_SECRET_KEY"
            )
        with self._lock:
            if self._client is None:
                client = AipSpeech(
                    self._s.baidu_speech_app_id,
                    self._s.baidu_speech_api_key,
                    self._s.baidu_speech_secret_key,
                )
                if self._s.baidu_speech_connection_timeout_ms is not None:
                    client.setConnectionTimeoutInMillis(
                        self._s.baidu_speech_connection_timeout_ms
                    )
                if self._s.baidu_speech_socket_timeout_ms is not None:
                    client.setSocketTimeoutInMillis(self._s.baidu_speech_socket_timeout_ms)
                self._client = client
            return self._client

    def asr(
        self,
        speech: bytes | None = None,
        format: str = "pcm",
        rate: int = 16000,
        options: dict[str, Any] | None = None,
    ) -> dict[str, Any]:
        """短语音识别。返回百度 JSON（含 `err_no`、`result` 等）。

        固定使用普通话模型（`dev_pid` 来自配置），避免未传参时误用服务端默认导致偏英语等结果。
        """
        merged: dict[str, Any] = dict(options or {})
        merged["dev_pid"] = int(self._s.baidu_speech_asr_dev_pid)
        return self._client_or_raise().asr(speech, format, rate, merged)

    def synthesis(
        self,
        text: str,
        lang: str = "zh",
        ctp: int = 1,
        options: dict[str, Any] | None = None,
    ) -> bytes | dict[str, Any]:
        """在线语音合成。成功为音频二进制；失败为错误信息 dict。"""
        return self._client_or_raise().synthesis(text, lang, ctp, options)