app.py

import gradio as gr
import os
from groq import Groq

# Load the API key from the environment variable
api_key = os.getenv('groq_whisper')

if api_key is None:
    raise ValueError("groq_whisper environment variable is not set")

# Initialize the Groq client
client = Groq(api_key=api_key)


def processaudio(audio_data):
    try:
        # Entpacken der Audiodaten (Sample-Rate und Numpy-Array)
        sample_rate, samples = audio_data
        
        # Temporäre Audiodatei erstellen
        with NamedTemporaryFile(suffix=".wav", delete=True) as tmpfile:
            # Audio als WAV-Datei speichern
            sf.write(tmpfile.name, samples, sample_rate)
            
            # Datei erneut öffnen und an Groq senden
            with open(tmpfile.name, "rb") as file:
                transcription = client.audio.transcriptions.create(
                    file=(os.path.basename(tmpfile.name), file.read()),
                    model="whisper-large-v3-turbo",
                    prompt="transcribe",
                    language="de",
                    response_format="json",
                    temperature=0.0
                )
            return transcription.text
            
    except Exception as e:
        return f"Ein Fehler ist aufgetreten: {str(e)}"



def process_audio(file_path):
   
    try:
        # Open the audio file
        with open(file_path, "rb") as file:
            # Create a transcription of the audio file
            transcription = client.audio.transcriptions.create(
                file=(os.path.basename(file_path), file.read()),  # Correct passing of filename
                model="whisper-large-v3-turbo",  # Required model to use for transcription
                prompt="transcribe",  # Optional
                language="de",  # Optional
                response_format="json",  # Optional
                temperature=0.0  # Optional
            )
            # Return the transcription text
            sr_inputs.clear()
            return transcription.text
    except Exception as e:
        return f"An error occurred: {str(e)}"

with gr.Blocks() as speech:  
    with gr.Row():
        sr_outputs = gr.Textbox(label="Transkription")
    with gr.Row():
        sr_inputs = gr.Microphone(type="filepath") 
    sr_inputs.change(process_audio, inputs=sr_inputs, outputs=sr_outputs)

speech.launch()