services/stt_service.py

from services.gemini_client import get_gemini_client


async def speech_to_text(audio_bytes: bytes) -> str:
    client = get_gemini_client()

    response = client.models.generate_content(
        model="gemini-2.5-flash",
        contents=[
            {
                "mime_type": "audio/wav",
                "data": audio_bytes
            }
        ]
    )

    text = response.text
    return text