dify-plugin-sdks/python/examples/openai/models/tts/tts.py

import concurrent.futures
import operator
from collections.abc import Generator
from functools import reduce
from io import BytesIO

from openai import OpenAI
from pydub import AudioSegment

from dify_plugin import TTSModel
from dify_plugin.errors.model import (
    CredentialsValidateFailedError,
    InvokeBadRequestError,
)

from ..common_openai import _CommonOpenAI


class OpenAIText2SpeechModel(_CommonOpenAI, TTSModel):
    """
    Model class for OpenAI Speech to text model.
    """

    def _invoke(
        self,
        model: str,
        tenant_id: str,
        credentials: dict,
        content_text: str,
        voice: str,
        user: str | None = None,
    ) -> bytes | Generator[bytes, None, None]:
        """
        _invoke text2speech model

        :param model: model name
        :param tenant_id: user tenant id
        :param credentials: model credentials
        :param content_text: text content to be translated
        :param voice: model timbre
        :param user: unique user id
        :return: text translated to audio file
        """

        voices = self.get_tts_model_voices(model=model, credentials=credentials)
        if not voices:
            raise InvokeBadRequestError("No voices found for the model")

        if not voice or voice not in [d["value"] for d in voices]:
            voice = self._get_model_default_voice(model, credentials)

        # if streaming:
        return self._tts_invoke_streaming(model=model, credentials=credentials, content_text=content_text, voice=voice)

    def validate_credentials(self, model: str, credentials: dict, user: str | None = None) -> None:
        """
        validate credentials text2speech model

        :param model: model name
        :param credentials: model credentials
        :param user: unique user id
        :return: text translated to audio file
        """
        try:
            self._tts_invoke(
                model=model,
                credentials=credentials,
                content_text="Hello Dify!",
                voice=self._get_model_default_voice(model, credentials),
            )
        except Exception as ex:
            raise CredentialsValidateFailedError(str(ex)) from ex

    def _tts_invoke(self, model: str, credentials: dict, content_text: str, voice: str) -> bytes:
        """
        _tts_invoke text2speech model

        :param model: model name
        :param credentials: model credentials
        :param content_text: text content to be translated
        :param voice: model timbre
        :return: text translated to audio file
        """
        audio_type = self._get_model_audio_type(model, credentials)
        word_limit = self._get_model_word_limit(model, credentials) or 500
        max_workers = self._get_model_workers_limit(model, credentials)
        try:
            sentences = list(self._split_text_into_sentences(org_text=content_text, max_length=word_limit))
            audio_bytes_list = []

            # Create a thread pool and map the function to the list of sentences
            with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
                futures = [
                    executor.submit(
                        self._process_sentence,
                        sentence=sentence,
                        model=model,
                        voice=voice,
                        credentials=credentials,
                    )
                    for sentence in sentences
                ]
                for future in futures:
                    try:
                        if future.result():
                            audio_bytes_list.append(future.result())
                    except Exception as ex:
                        raise InvokeBadRequestError(str(ex)) from ex

            if len(audio_bytes_list) > 0:
                audio_segments = [
                    AudioSegment.from_file(BytesIO(audio_bytes), format=audio_type)
                    for audio_bytes in audio_bytes_list
                    if audio_bytes
                ]
                combined_segment = reduce(operator.add, audio_segments)
                buffer: BytesIO = BytesIO()
                combined_segment.export(buffer, format=audio_type)
                buffer.seek(0)

                return buffer.read()
            else:
                raise InvokeBadRequestError("No audio bytes found")
        except Exception as ex:
            raise InvokeBadRequestError(str(ex)) from ex

    def _tts_invoke_streaming(
        self, model: str, credentials: dict, content_text: str, voice: str
    ) -> Generator[bytes, None, None]:
        """
        _tts_invoke_streaming text2speech model

        :param model: model name
        :param credentials: model credentials
        :param content_text: text content to be translated
        :param voice: model timbre
        :return: text translated to audio file
        """
        try:
            # doc: https://platform.openai.com/docs/guides/text-to-speech
            credentials_kwargs = self._to_credential_kwargs(credentials)
            client = OpenAI(**credentials_kwargs)

            voices = self.get_tts_model_voices(model=model, credentials=credentials)
            if not voices:
                raise InvokeBadRequestError("No voices found for the model")

            if not voice or voice not in voices:
                voice = self._get_model_default_voice(model, credentials)

            word_limit = self._get_model_word_limit(model, credentials) or 500
            if len(content_text) > word_limit:
                sentences = self._split_text_into_sentences(content_text, max_length=word_limit)
                executor = concurrent.futures.ThreadPoolExecutor(max_workers=min(3, len(sentences)))
                futures = [
                    executor.submit(
                        client.audio.speech.with_streaming_response.create,
                        model=model,
                        response_format="mp3",
                        input=sentences[i],
                        voice=voice,  # type: ignore
                    )
                    for i in range(len(sentences))
                ]
                for future in futures:
                    response = future.result()
                    with response as r:
                        yield from r.iter_bytes(1024)

            else:
                response = client.audio.speech.with_streaming_response.create(
                    model=model,
                    voice=voice,  # type: ignore
                    response_format="mp3",
                    input=content_text.strip(),
                )

                with response as r:
                    yield from r.iter_bytes(1024)
        except Exception as ex:
            raise InvokeBadRequestError(str(ex)) from ex

    def _process_sentence(self, sentence: str, model: str, voice, credentials: dict):
        """
        _tts_invoke openai text2speech model api

        :param model: model name
        :param credentials: model credentials
        :param voice: model timbre
        :param sentence: text content to be translated
        :return: text translated to audio file
        """
        # transform credentials to kwargs for model instance
        credentials_kwargs = self._to_credential_kwargs(credentials)
        client = OpenAI(**credentials_kwargs)
        response = client.audio.speech.create(model=model, voice=voice, input=sentence.strip())
        if isinstance(response.read(), bytes):
            return response.read()