import subprocess
import sys

# Force upgrade gradio (only for Hugging Face)
import os
if "SPACE_ID" in os.environ:
    subprocess.check_call([sys.executable, "-m", "pip", "install", "--upgrade", "gradio>=4.44.0"])

import gradio as gr
from transformers import (
    pipeline,
    WhisperForConditionalGeneration,
    AutoTokenizer,
    WhisperFeatureExtractor,
    GenerationConfig
)
import traceback

print("🚀 Starting Kurmanji ASR application...")

# Global variables
asr = None

def load_asr_model():
    global asr
    
    try:
        print("📥 Loading Whisper model for Kurmanji...")
        
        # Load generation config and remove forced_decoder_ids
        print("⚙️ Loading generation config...")
        gen_config = GenerationConfig.from_pretrained("amedcj/whisper-kurmanji")
        gen_config.forced_decoder_ids = None
        print("✓ Generation config loaded")
        
        # Load model and set generation config directly
        print("🤖 Loading Whisper model...")
        model = WhisperForConditionalGeneration.from_pretrained("amedcj/whisper-kurmanji")
        model.generation_config = gen_config
        print("✓ Model loaded successfully")
        
        # Load tokenizer explicitly
        print("📝 Loading tokenizer...")
        tokenizer = AutoTokenizer.from_pretrained("amedcj/whisper-kurmanji")
        print("✓ Tokenizer loaded successfully")
        
        # Load feature extractor explicitly
        print("🔍 Loading feature extractor...")
        feature_extractor = WhisperFeatureExtractor.from_pretrained("amedcj/whisper-kurmanji")
        print("✓ Feature extractor loaded successfully")
        
        # Create the pipeline with model, tokenizer and feature extractor
        print("🔧 Creating ASR pipeline...")
        asr = pipeline(
            "automatic-speech-recognition",
            model=model,
            tokenizer=tokenizer,
            feature_extractor=feature_extractor,
            chunk_length_s=10,  # try with 10s chunks
            stride_length_s=(2, 2),  # optional smoothing
            device=-1  # CPU
        )
        print("✅ ASR pipeline created successfully!")

               
    except Exception as e:
        print(f"❌ Error loading ASR model: {e}")
        traceback.print_exc()
        asr = None

# Load the model at startup
load_asr_model()

def transcribe(audio_file):
    print("=== ASR Function Called ===")
    print(f"Audio file: {audio_file}")
    
    try:
        # Check if audio file is provided
        if audio_file is None:
            error_msg = "Ji kerema xwe dosyeyek deng bar bike. / Please upload an audio file."
            print(f"Error: {error_msg}")
            return error_msg
        
        # Check if ASR model is loaded
        if asr is None:
            error_msg = "Model nehatiye barkirin. / ASR model not loaded properly."
            print(f"Error: {error_msg}")
            return error_msg
        
        print("🎵 Processing audio file...")
        
        # Transcribe the audio
        result = asr(audio_file)
        transcription = result["text"]
        
        print(f"✅ Transcription completed: {transcription}")
        return transcription
        
    except Exception as e:
        error_msg = f"Çewtî: {str(e)} / Error: {str(e)}"
        print(f"❌ Error in transcription: {e}")
        traceback.print_exc()
        return error_msg

def clear_inputs():
    return None, ""

# Create Gradio interface with Kurdish elements
print("🎨 Creating Gradio interface...")

demo = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(
        sources=["microphone", "upload"],
        type="filepath",
        label="🎤 Deng Tomar Bike an Dosye Bar Bike / Record Voice or Upload File"
    ),
    outputs=gr.Textbox(
        label="📝 Encam / Result",
        placeholder="Li vir nivîsa wergerandî dê xuya bibe... / Transcribed text will appear here...",
        lines=5,
        show_copy_button=True
    ),
    title="🗣️ Kurmancî ASR - Kurdish Speech Recognition",
    description="""
    **Deng bo Nivîs / Speech to Text**
    
    **Formatên çêdibin:** WAV, MP3, M4A, FLAC
    """,
    submit_btn="Wergerîne / Transcribe",
    clear_btn="Paqij Bike / Clear",
    examples=[
        # You can add example audio files here if you have them
        # ["path/to/example1.wav"],
        # ["path/to/example2.mp3"],
    ],
    cache_examples=False
)

print("🚀 Launching Gradio app...")

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