tools/audio_transcribe.py

import os
import requests
import tempfile

def audio_transcribe(audio_url: str) -> str:
    """
    Transcribe an audio file from a URL to text using HuggingFace Whisper API.
    Uses HF_TOKEN for authentication when downloading the file.
    """
    try:
        headers = {"Authorization": f"Bearer {os.getenv('HF_TOKEN')}"}
        response = requests.get(audio_url, headers=headers, timeout=15)
        response.raise_for_status()
        
        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp:
            tmp.write(response.content)
            tmp_path = tmp.name
        
        api_url = "https://api-inference.huggingface.co/models/openai/whisper-base"
        api_headers = {"Authorization": f"Bearer {os.getenv('HF_TOKEN')}"}
        with open(tmp_path, "rb") as f:
            api_response = requests.post(api_url, headers=api_headers, data=f, timeout=30)
        os.unlink(tmp_path)
        
        if api_response.status_code == 200:
            result = api_response.json()
            transcript = result.get("text", "").strip()
            return transcript if transcript else "No speech detected."
        else:
            return f"Transcription API error (HTTP {api_response.status_code}): {api_response.text}"
    except requests.exceptions.RequestException as e:
        return f"Failed to download audio file: {str(e)}"
    except Exception as e:
        return f"Transcription failed: {str(e)}"