services/stt_service.py

from services.gemini_client import get_gemini_client
from google.genai import types

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

    # Correctly wrap audio bytes using types.File
    audio_file = types.File(data=audio_bytes, mime_type="audio/wav")

    response = client.models.generate_content(
        model="gemini-2.5-flash",
        contents=[audio_file]  # <-- pass as a list of types.File
    )

    return response.text