MCP Server Whisper

"""Transcription service - orchestrates transcription operations.""" from typing import Literal from ..infrastructure import FileSystemRepository, OpenAIClientWrapper, SecurePathResolver from ..models import ChatResult, TranscriptionResult class TranscriptionService: """Service for audio transcription operations.""" def __init__( self, file_repo: FileSystemRepository, openai_client: OpenAIClientWrapper, path_resolver: SecurePathResolver ): """Initialize the transcription service. Args: ---- file_repo: File system repository for I/O operations. openai_client: OpenAI API client wrapper. path_resolver: Secure path resolver for filename to path conversion. """ self.file_repo = file_repo self.openai_client = openai_client self.path_resolver = path_resolver async def transcribe_audio( self, filename: str, model: str = "gpt-4o-mini-transcribe", response_format: str = "text", prompt: str | None = None, timestamp_granularities: list[Literal["word", "segment"]] | None = None, ) -> TranscriptionResult: """Transcribe audio file using OpenAI's transcription API. Args: ---- filename: Name of the audio file. model: Transcription model to use. response_format: Format of the response. prompt: Optional prompt to guide transcription. timestamp_granularities: Optional timestamp granularities. Returns: ------- TranscriptionResult: Transcription result with typed fields. """ # Resolve filename to path file_path = self.path_resolver.resolve_input(filename) # Read audio file audio_bytes = await self.file_repo.read_audio_file(file_path) # Transcribe using OpenAI result_dict = await self.openai_client.transcribe_audio( audio_bytes=audio_bytes, filename=filename, model=model, # type: ignore response_format=response_format, # type: ignore prompt=prompt, timestamp_granularities=timestamp_granularities, ) # Convert dict to typed response model return TranscriptionResult(**result_dict) async def chat_with_audio( self, filename: str, model: str = "gpt-4o-audio-preview-2024-12-17", system_prompt: str | None = None, user_prompt: str | None = None, ) -> ChatResult: """Chat with audio using GPT-4o audio models. Args: ---- filename: Name of the audio file. model: Audio chat model to use. system_prompt: Optional system prompt. user_prompt: Optional user text prompt. Returns: ------- ChatResult: Chat response with typed text field. """ # Resolve filename to path file_path = self.path_resolver.resolve_input(filename) # Validate format ext = file_path.suffix.lower().replace(".", "") if ext not in ["mp3", "wav"]: raise ValueError(f"Expected mp3 or wav extension, but got {ext}") # Read audio file audio_bytes = await self.file_repo.read_audio_file(file_path) # Chat with audio using OpenAI result_dict = await self.openai_client.chat_with_audio( audio_bytes=audio_bytes, audio_format=ext, # type: ignore model=model, # type: ignore system_prompt=system_prompt, user_prompt=user_prompt, ) # Convert dict to typed response model return ChatResult(**result_dict)