from openai import OpenAI

client = OpenAI(base_url="https://api.naga.ac/v1", api_key="YOUR_API_KEY")

with open("audio.mp3", "rb") as f:
    transcription = client.audio.transcriptions.create(
        model="cohere-transcribe-03-2026",
        file=f,
        prompt=None,
        language=None,
    )
print(transcription.text)