MCP Server Whisper

"""Transcription related Pydantic models.""" from typing import Literal, Optional from openai.types import AudioModel, AudioResponseFormat from pydantic import Field from ..constants import ENHANCEMENT_PROMPTS, AudioChatModel, EnhancementType from .base import BaseInputPath class TranscribeAudioInputParamsBase(BaseInputPath): """Base params for transcribing audio with audio-to-text model.""" model: AudioModel = Field( default="gpt-4o-mini-transcribe", description="The transcription model to use (e.g., 'whisper-1', 'gpt-4o-transcribe', 'gpt-4o-mini-transcribe')", ) response_format: AudioResponseFormat = Field( "text", description="The response format of the transcription model. " 'Use `verbose_json` with `model="whisper-1"` for timestamps. ' "`gpt-4o-transcribe` and `gpt-4o-mini-transcribe` only support `text` and `json`.", ) timestamp_granularities: list[Literal["word", "segment"]] | None = Field( None, description="""The timestamp granularities to populate for this transcription. `response_format` must be set `verbose_json` to use timestamp granularities. Either or both of these options are supported: `word`, or `segment`. Note: There is no additional latency for segment timestamps, but generating word timestamp incurs additional latency. """, ) class TranscribeAudioInputParams(TranscribeAudioInputParamsBase): """Params for transcribing audio with audio-to-text model.""" prompt: str | None = Field( None, description="""An optional prompt to guide the transcription model's output. Effective prompts can: 1. Correct specific words/acronyms: Include technical terms or names that might be misrecognized Example: "The transcript discusses OpenAI's DALL·E and GPT-4 technology" 2. Maintain context from previous segments: Include the last part of previous transcript Note: Model only considers final 224 tokens of the prompt 3. Enforce punctuation: Include properly punctuated example text Example: "Hello, welcome to my lecture. Today, we'll discuss..." 4. Preserve filler words: Include example with verbal hesitations Example: "Umm, let me think like, hmm... Okay, here's what I'm thinking" 5. Set writing style: Use examples in desired format (simplified/traditional, formal/casual) The model will try to match the style and formatting of your prompt.""", ) class ChatWithAudioInputParams(BaseInputPath): """Params for transcribing audio with LLM using custom prompt.""" system_prompt: Optional[str] = Field(default=None, description="Custom system prompt to use.") user_prompt: Optional[str] = Field(default=None, description="Custom user prompt to use.") model: AudioChatModel = Field( default="gpt-4o-audio-preview-2024-12-17", description="The audio LLM model to use for transcription" ) class TranscribeWithEnhancementInputParams(TranscribeAudioInputParamsBase): """Params for transcribing audio with LLM using template prompt.""" enhancement_type: EnhancementType = Field( default="detailed", description="Type of enhancement to apply to the transcription: " "detailed, storytelling, professional, or analytical.", ) def to_transcribe_audio_input_params(self) -> TranscribeAudioInputParams: """Transfer audio with LLM using custom prompt.""" return TranscribeAudioInputParams( input_file_path=self.input_file_path, prompt=ENHANCEMENT_PROMPTS[self.enhancement_type], model=self.model, timestamp_granularities=self.timestamp_granularities, response_format=self.response_format, )