src/stt.py

# File: src/stt.py
# Purpose: Convert audio file to transcript using OpenAI Whisper (optional, not used on HF Spaces)

try:
    import whisper  # type: ignore[import]
    WHISPER_AVAILABLE = True
except ImportError:
    WHISPER_AVAILABLE = False

from pathlib import Path
import sys
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from config import WHISPER_MODEL_SIZE

_model = None


def get_model():
    if not WHISPER_AVAILABLE:
        raise RuntimeError("openai-whisper is not installed. Run: pip install openai-whisper")
    global _model
    if _model is None:
        print(f"Loading Whisper model: {WHISPER_MODEL_SIZE}")
        _model = whisper.load_model(WHISPER_MODEL_SIZE)
    return _model


def transcribe_audio(audio_path: str) -> str:
    model = get_model()
    result = model.transcribe(audio_path, fp16=False, language="en")
    transcript = result["text"].strip()
    print(f"[STT] Transcript: {transcript}")
    return transcript


def transcribe_from_bland_webhook(audio_url: str) -> str:
    import requests
    import tempfile
    import os

    response = requests.get(audio_url, timeout=30)
    response.raise_for_status()

    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp:
        tmp.write(response.content)
        tmp_path = tmp.name

    try:
        transcript = transcribe_audio(tmp_path)
    finally:
        os.unlink(tmp_path)

    return transcript


if __name__ == "__main__":
    if len(sys.argv) > 1:
        text = transcribe_audio(sys.argv[1])
        print(f"Result: {text}")
    else:
        print("Usage: python src/stt.py <audio_file.wav>")