src/stt/deepgram_rest.py

import logging
from deepgram import AsyncDeepgramClient
from src.config import DEEPGRAM_API_KEY, DEEPGRAM_LANGUAGE

logger = logging.getLogger(__name__)


async def transcribe_audio(data: bytes, mimetype: str = "audio/wav") -> dict:
    """
    Transcribes a full audio file using Deepgram pre-recorded API.
    Returns dict with 'text' (full transcript) and 'duration' (seconds).
    """
    client = AsyncDeepgramClient(api_key=DEEPGRAM_API_KEY)
    response = await client.listen.v1.media.transcribe_file(
        request=data,
        model="nova-2",
        language=DEEPGRAM_LANGUAGE,
        smart_format=True,
    )

    try:
        transcript = response.results.channels[0].alternatives[0].transcript
    except (AttributeError, IndexError):
        transcript = ""

    try:
        duration = response.metadata.duration
    except AttributeError:
        duration = None

    logger.info("STT transcript (%ss): %s", duration, transcript[:80])
    return {"text": transcript, "language": DEEPGRAM_LANGUAGE, "duration": duration}