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:
    """
    Convert speech audio (bytes) to text using Gemini API
    """
    client = get_gemini_client()

    # Wrap audio bytes correctly for Gemini
    contents = [types.File(data=audio_bytes, mime_type="audio/wav")]

    response = client.models.generate_content(
        model="gemini-2.5-flash",
        contents=contents
    )

    return response.text