Get Voice Prompt Audio

Download the original prompt audio for a voice from GCS.

Retrieves the preset row from the voice_presets PostgreSQL table using the preset_id resolved from PostgreSQL, reads the prompt_audio_gcs_uri field (format <bucket>/<filename>), downloads the file from GCS, and returns the raw audio bytes with an appropriate Content-Type header derived from the file extension.

Args: voice_id: UUID of the voice whose prompt audio should be fetched. model_version: AI model version used to resolve the preset_id (default async_flash_v1.0). x_api_key: Bearer API key from the x-api-key header. token_service: Injected token validation service. request: Raw FastAPI Request. Returns: Response: Raw audio bytes with Content-Disposition attachment header.

Raises: VoicePreviewNotFound: When no preset exists for the voice/model combination, or when the preset document has no prompt_audio_gcs_uri field.