app.py

import gradio as gr
from typhoon_asr import transcribe
import os

# Global variable to store transcription results
last_transcription = None

def transcribe_audio(audio_path):
    """Transcribe the audio using typhoon_asr"""
    global last_transcription
    
    if not audio_path:
        return "❌ No audio to transcribe. Please upload or record audio first.", ""
    
    try:
        # Show loading message
        status_msg = "🔄 Transcribing audio..."
        
        # Perform transcription (basic only)
        print(f"DEBUG: Starting transcription of {audio_path}")
        result = transcribe(audio_path)
        print(f"DEBUG: Transcription result: {result}")
        
        last_transcription = result
        
        # Handle different result formats
        if isinstance(result, dict) and 'text' in result:
            if hasattr(result['text'], 'text'):
                transcription_text = result['text'].text
            else:
                transcription_text = result['text']
        else:
            transcription_text = str(result)
        
        status_msg = "✅ Transcription completed!"
        return status_msg, transcription_text
        
    except Exception as e:
        error_msg = f"❌ Transcription failed: {str(e)}"
        print(f"DEBUG: Error occurred: {error_msg}")
        return error_msg, ""

def clear_transcription():
    """Clear the transcription"""
    global last_transcription
    last_transcription = None
    return "🗑️ Transcription cleared", ""

def audio_uploaded(audio_path):
    """Called when audio is uploaded - update status and enable button"""
    if audio_path:
        return f"✅ Audio uploaded! Ready to transcribe.", gr.Button(interactive=True)
    else:
        return "❌ No audio uploaded", gr.Button(interactive=False)

# Create the Gradio interface
with gr.Blocks(title="Typhoon ASR API") as demo:
    gr.Markdown("# 🎤 Typhoon ASR Real-Time Transcription")
    gr.Markdown("Upload an audio file or record to get Thai speech transcription")
    
    # Audio input component
    audio_input = gr.Audio(
        sources=["upload", "microphone"], 
        type="filepath", 
        label="Upload Audio File or Record"
    )
    
    # Status display
    status_text = gr.Textbox(
        label="Status", 
        value="Upload or record audio to get started", 
        interactive=False
    )
    
    # Transcription buttons
    gr.Markdown("### Transcription")
    with gr.Row():
        transcribe_btn = gr.Button("🎯 Transcribe", variant="primary", interactive=False)
        clear_btn = gr.Button("🗑️ Clear Result", variant="secondary")
    
    # Transcription result
    transcription_output = gr.Textbox(
        label="Transcription Result",
        lines=10,
        placeholder="Transcription will appear here after uploading/recording and clicking transcribe...",
        interactive=False
    )
    
    # Event handlers
    # When audio changes (uploaded/recorded), update status and enable button
    audio_input.change(
        fn=audio_uploaded,
        inputs=[audio_input],
        outputs=[status_text, transcribe_btn]
    )
    
    # Transcription button click
    transcribe_btn.click(
        fn=transcribe_audio,
        inputs=[audio_input],
        outputs=[status_text, transcription_output]
    )
    
    clear_btn.click(
        fn=clear_transcription,
        outputs=[status_text, transcription_output]
    )

# For API access - this function can be called externally
def api_transcribe(audio_file_path):
    """API endpoint for external calls"""
    return transcribe_audio(audio_file_path)

if __name__ == "__main__":
    demo.launch()