Skip to content

armgit / dify

Public Imported from GitHub
0

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.

7.8 KB
Raw Blame
"""Installed-app audio admission, error translation, and response serialization."""

import logging
from collections.abc import Callable
from functools import wraps
from http import HTTPStatus

from flask import Response, request
from flask_restx import Resource

from controllers.common.audio_response import audio_binary_response
from controllers.common.controller_schemas import TextToAudioPayload
from controllers.common.errors import InternalServerError
from controllers.common.fields import AudioBinaryResponse, AudioTranscriptResponse
from controllers.common.schema import register_response_schema_models, register_schema_model
from controllers.console import console_ns
from controllers.console.app.error import (
    AppUnavailableError,
    AudioTooLargeError,
    CompletionRequestError,
    NoAudioUploadedError,
    ProviderModelCurrentlyNotSupportError,
    ProviderNotInitializeError,
    ProviderNotSupportSpeechToTextError,
    ProviderNotSupportTextToSpeechError,
    ProviderQuotaExceededError,
    SpeechToTextDisabledError,
    UnsupportedAudioTypeError,
)
from controllers.console.explore.error import AgentVersionNotFoundHTTPError
from controllers.console.explore.installed_app_admission import get_installed_app
from controllers.console.flask_admission import console_account_admission
from controllers.console.wraps import model_validate
from core.base.tts.audio_mime import SUPPORTED_TTS_AUDIO_MIME_TYPES
from core.errors.error import ModelCurrentlyNotSupportError, ProviderTokenNotInitError, QuotaExceededError
from extensions.ext_application_services import application_services
from graphon.model_runtime.errors.invoke import InvokeError
from libs.exception import BaseHTTPException
from libs.external_api import error_responses
from libs.flask_restx_compat import BINARY_RESPONSE_MEDIA_TYPES_VENDOR_KEY
from libs.helper import dump_response
from machinery.context import RequestContext
from services.agent.errors import AgentVersionNotFoundError
from services.app_definition_query_service import AppDefinitionUnavailableError
from services.audio_types import AudioAppRef, AudioUpload
from services.errors.app_model_config import AppModelConfigBrokenError
from services.errors.audio import (
    AudioTooLargeServiceError,
    NoAudioUploadedServiceError,
    ProviderNotSupportSpeechToTextServiceError,
    ProviderNotSupportTextToSpeechServiceError,
    SpeechToTextDisabledServiceError,
    UnsupportedAudioTypeServiceError,
)
from services.installed_app_access_service import InstalledAppRef

logger = logging.getLogger(__name__)

register_schema_model(console_ns, TextToAudioPayload)
register_response_schema_models(console_ns, AudioBinaryResponse, AudioTranscriptResponse)


def _audio_errors[**P, R](view: Callable[P, R]) -> Callable[P, R]:
    @wraps(view)
    def decorated(*args: P.args, **kwargs: P.kwargs) -> R:
        try:
            return view(*args, **kwargs)
        except AppDefinitionUnavailableError as error:
            raise AppUnavailableError() from error
        except AppModelConfigBrokenError as error:
            logger.exception("App model config broken")
            raise AppUnavailableError() from error
        except NoAudioUploadedServiceError as error:
            raise NoAudioUploadedError() from error
        except AudioTooLargeServiceError as error:
            raise AudioTooLargeError(str(error)) from error
        except UnsupportedAudioTypeServiceError as error:
            raise UnsupportedAudioTypeError() from error
        except ProviderNotSupportSpeechToTextServiceError as error:
            raise ProviderNotSupportSpeechToTextError() from error
        except ProviderNotSupportTextToSpeechServiceError as error:
            raise ProviderNotSupportTextToSpeechError() from error
        except SpeechToTextDisabledServiceError as error:
            raise SpeechToTextDisabledError() from error
        except ProviderTokenNotInitError as error:
            raise ProviderNotInitializeError(error.description) from error
        except QuotaExceededError as error:
            raise ProviderQuotaExceededError() from error
        except ModelCurrentlyNotSupportError as error:
            raise ProviderModelCurrentlyNotSupportError() from error
        except InvokeError as error:
            raise CompletionRequestError(error.description) from error
        except AgentVersionNotFoundError as error:
            raise AgentVersionNotFoundHTTPError() from error
        except (BaseHTTPException, ValueError):
            raise
        except Exception as error:
            logger.exception("Installed-app audio operation failed")
            raise InternalServerError() from error

    return decorated


@console_ns.route(
    "/installed-apps/<uuid:installed_app_id>/audio-to-text",
    endpoint="installed_app_audio",
)
class ChatAudioApi(Resource):
    @error_responses(
        AgentVersionNotFoundHTTPError,
        AppUnavailableError,
        AudioTooLargeError,
        CompletionRequestError,
        NoAudioUploadedError,
        ProviderModelCurrentlyNotSupportError,
        ProviderNotInitializeError,
        ProviderNotSupportSpeechToTextError,
        ProviderQuotaExceededError,
        SpeechToTextDisabledError,
        UnsupportedAudioTypeError,
    )
    @console_ns.response(HTTPStatus.OK, "Success", console_ns.models[AudioTranscriptResponse.__name__])
    @console_account_admission()
    @get_installed_app
    def post(self, request_context: RequestContext, installed_app: InstalledAppRef) -> dict[str, object]:
        # Let Flask handle multipart parsing errors before mapping audio operation failures.
        file = request.files.get("file")
        audio = AudioUpload(stream=file.stream, mime_type=file.mimetype) if file is not None else None
        return self._transcribe(installed_app, audio)

    @_audio_errors
    def _transcribe(self, installed_app: InstalledAppRef, audio: AudioUpload | None) -> dict[str, object]:
        """Transcribe an upload; None lets the audio service report a missing file."""
        transcript = application_services().app_audio.transcript_asr(
            app=AudioAppRef(
                app_id=installed_app.app_id,
                tenant_id=installed_app.app_owner_tenant_id,
                app_mode=installed_app.app_mode,
            ),
            audio=audio,
        )
        return dump_response(AudioTranscriptResponse, transcript)


@console_ns.route(
    "/installed-apps/<uuid:installed_app_id>/text-to-audio",
    endpoint="installed_app_text",
)
class ChatTextApi(Resource):
    @error_responses(
        AgentVersionNotFoundHTTPError,
        AppUnavailableError,
        CompletionRequestError,
        ProviderModelCurrentlyNotSupportError,
        ProviderNotInitializeError,
        ProviderNotSupportTextToSpeechError,
        ProviderQuotaExceededError,
    )
    @console_ns.expect(console_ns.models[TextToAudioPayload.__name__])
    @console_ns.doc(
        produces=list(SUPPORTED_TTS_AUDIO_MIME_TYPES),
        vendor={BINARY_RESPONSE_MEDIA_TYPES_VENDOR_KEY: list(SUPPORTED_TTS_AUDIO_MIME_TYPES)},
    )
    @console_ns.response(HTTPStatus.OK, "Success", console_ns.models[AudioBinaryResponse.__name__])
    @console_account_admission()
    @get_installed_app
    @model_validate(TextToAudioPayload)
    @_audio_errors
    def post(
        self, payload: TextToAudioPayload, request_context: RequestContext, installed_app: InstalledAppRef
    ) -> Response | None:
        output = application_services().app_audio.transcript_tts(
            app=AudioAppRef(
                app_id=installed_app.app_id,
                tenant_id=installed_app.app_owner_tenant_id,
                app_mode=installed_app.app_mode,
            ),
            account_id=request_context.account_id,
            text=payload.text,
            voice=payload.voice,
            message_id=payload.message_id,
        )
        # response-contract:ignore audio_binary_response
        return audio_binary_response(output)