openai

@define
class AzureOpenAiTextToSpeechDriver(OpenAiTextToSpeechDriver):
    """Azure OpenAi Text to Speech Driver.

    Attributes:
        azure_deployment: An optional Azure OpenAi deployment id. Defaults to the model name.
        azure_endpoint: An Azure OpenAi endpoint.
        azure_ad_token: An optional Azure Active Directory token.
        azure_ad_token_provider: An optional Azure Active Directory token provider.
        api_version: An Azure OpenAi API version.
        client: An `openai.AzureOpenAI` client.
    """

    model: str = field(default="tts", kw_only=True, metadata={"serializable": True})
    azure_deployment: str = field(
        kw_only=True,
        default=Factory(lambda self: self.model, takes_self=True),
        metadata={"serializable": True},
    )
    azure_endpoint: str = field(kw_only=True, metadata={"serializable": True})
    azure_ad_token: Optional[str] = field(kw_only=True, default=None, metadata={"serializable": False})
    azure_ad_token_provider: Optional[Callable[[], str]] = field(
        kw_only=True,
        default=None,
        metadata={"serializable": False},
    )
    api_version: str = field(default="2024-07-01-preview", kw_only=True, metadata={"serializable": True})
    _client: Optional[openai.AzureOpenAI] = field(
        default=None, kw_only=True, alias="client", metadata={"serializable": False}
    )

    @lazy_property()
    def client(self) -> openai.AzureOpenAI:
        return openai.AzureOpenAI(
            organization=self.organization,
            api_key=self.api_key,
            api_version=self.api_version,
            azure_endpoint=self.azure_endpoint,
            azure_deployment=self.azure_deployment,
            azure_ad_token=self.azure_ad_token,
            azure_ad_token_provider=self.azure_ad_token_provider,
        )

@lazy_property()
def client(self) -> openai.AzureOpenAI:
    return openai.AzureOpenAI(
        organization=self.organization,
        api_key=self.api_key,
        api_version=self.api_version,
        azure_endpoint=self.azure_endpoint,
        azure_deployment=self.azure_deployment,
        azure_ad_token=self.azure_ad_token,
        azure_ad_token_provider=self.azure_ad_token_provider,
    )

@define
class OpenAiTextToSpeechDriver(BaseTextToSpeechDriver):
    model: str = field(default="tts-1", kw_only=True, metadata={"serializable": True})
    voice: Literal["alloy", "echo", "fable", "onyx", "nova", "shimmer"] = field(
        default="alloy",
        kw_only=True,
        metadata={"serializable": True},
    )
    format: Literal["mp3", "opus", "aac", "flac"] = field(default="mp3", kw_only=True, metadata={"serializable": True})
    api_type: Optional[str] = field(default=openai.api_type, kw_only=True)
    api_version: Optional[str] = field(default=openai.api_version, kw_only=True, metadata={"serializable": True})
    base_url: Optional[str] = field(default=None, kw_only=True, metadata={"serializable": True})
    api_key: Optional[str] = field(default=None, kw_only=True)
    organization: Optional[str] = field(default=openai.organization, kw_only=True, metadata={"serializable": True})
    _client: Optional[openai.OpenAI] = field(
        default=None, kw_only=True, alias="client", metadata={"serializable": False}
    )

    @lazy_property()
    def client(self) -> openai.OpenAI:
        return openai.OpenAI(
            api_key=self.api_key,
            base_url=self.base_url,
            organization=self.organization,
        )

    def try_text_to_audio(self, prompts: list[str]) -> AudioArtifact:
        response = self.client.audio.speech.create(
            input=". ".join(prompts),
            voice=self.voice,
            model=self.model,
            response_format=self.format,
        )

        return AudioArtifact(value=response.content, format=self.format)

@lazy_property()
def client(self) -> openai.OpenAI:
    return openai.OpenAI(
        api_key=self.api_key,
        base_url=self.base_url,
        organization=self.organization,
    )

def try_text_to_audio(self, prompts: list[str]) -> AudioArtifact:
    response = self.client.audio.speech.create(
        input=". ".join(prompts),
        voice=self.voice,
        model=self.model,
        response_format=self.format,
    )

    return AudioArtifact(value=response.content, format=self.format)