app-backup.py

import spaces  # 추가
import gradio as gr
import os
import asyncio
import torch
import io
import json
import re
import httpx
import tempfile
import wave
import base64
import numpy as np
import soundfile as sf
import subprocess
import shutil
from dataclasses import dataclass
from typing import List, Tuple, Dict, Optional
from pathlib import Path
from threading import Thread
from dotenv import load_dotenv

# Edge TTS imports
import edge_tts
from pydub import AudioSegment

# OpenAI imports
from openai import OpenAI

# Transformers imports (for legacy local mode)
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    TextIteratorStreamer,
    BitsAndBytesConfig,
)

# Llama CPP imports (for new local mode)
try:
    from llama_cpp import Llama
    from llama_cpp_agent import LlamaCppAgent, MessagesFormatterType
    from llama_cpp_agent.providers import LlamaCppPythonProvider
    from llama_cpp_agent.chat_history import BasicChatHistory
    from llama_cpp_agent.chat_history.messages import Roles
    from huggingface_hub import hf_hub_download
    LLAMA_CPP_AVAILABLE = True
except ImportError:
    LLAMA_CPP_AVAILABLE = False

# Spark TTS imports
try:
    from huggingface_hub import snapshot_download
    SPARK_AVAILABLE = True
except:
    SPARK_AVAILABLE = False

# MeloTTS imports (for local mode)
try:
    # unidic 다운로드를 조건부로 처리
    if not os.path.exists("/usr/local/lib/python3.10/site-packages/unidic"):
        try:
            os.system("python -m unidic download")
        except:
            pass
    from melo.api import TTS as MeloTTS
    MELO_AVAILABLE = True
except:
    MELO_AVAILABLE = False

load_dotenv()


@dataclass
class ConversationConfig:
    max_words: int = 6000
    prefix_url: str = "https://r.jina.ai/"
    api_model_name: str = "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo"
    legacy_local_model_name: str = "NousResearch/Hermes-2-Pro-Llama-3-8B"
    # 새로운 로컬 모델 설정
    local_model_name: str = "Private-BitSix-Mistral-Small-3.1-24B-Instruct-2503.gguf"
    local_model_repo: str = "ginigen/Private-BitSix-Mistral-Small-3.1-24B-Instruct-2503"


class UnifiedAudioConverter:
    def __init__(self, config: ConversationConfig):
        self.config = config
        self.llm_client = None
        self.legacy_local_model = None
        self.legacy_tokenizer = None
        # 새로운 로컬 LLM 관련
        self.local_llm = None
        self.local_llm_model = None
        self.melo_models = None
        self.spark_model_dir = None
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        
    def initialize_api_mode(self, api_key: str):
        """Initialize API mode with Together API (now fallback)"""
        self.llm_client = OpenAI(api_key=api_key, base_url="https://api.together.xyz/v1")
        
    @spaces.GPU(duration=120)
    def initialize_local_mode(self):
        """Initialize new local mode with Llama CPP"""
        if not LLAMA_CPP_AVAILABLE:
            raise RuntimeError("Llama CPP dependencies not available. Please install llama-cpp-python and llama-cpp-agent.")
        
        if self.local_llm is None or self.local_llm_model != self.config.local_model_name:
            try:
                # 모델 다운로드
                model_path = hf_hub_download(
                    repo_id=self.config.local_model_repo,
                    filename=self.config.local_model_name,
                    local_dir="./models"
                )
                
                model_path_local = os.path.join("./models", self.config.local_model_name)
                
                if not os.path.exists(model_path_local):
                    raise RuntimeError(f"Model file not found at {model_path_local}")
                
                # Llama 모델 초기화
                self.local_llm = Llama(
                    model_path=model_path_local,
                    flash_attn=True,
                    n_gpu_layers=81 if torch.cuda.is_available() else 0,
                    n_batch=1024,
                    n_ctx=8192,
                )
                self.local_llm_model = self.config.local_model_name
                print(f"Local LLM initialized: {model_path_local}")
                
            except Exception as e:
                print(f"Failed to initialize local LLM: {e}")
                raise RuntimeError(f"Failed to initialize local LLM: {e}")

    @spaces.GPU(duration=60)
    def initialize_legacy_local_mode(self):
        """Initialize legacy local mode with Hugging Face model (fallback)"""
        if self.legacy_local_model is None:
            quantization_config = BitsAndBytesConfig(
                load_in_4bit=True, 
                bnb_4bit_compute_dtype=torch.float16
            )
            self.legacy_local_model = AutoModelForCausalLM.from_pretrained(
                self.config.legacy_local_model_name, 
                quantization_config=quantization_config
            )
            self.legacy_tokenizer = AutoTokenizer.from_pretrained(
                self.config.legacy_local_model_name,
                revision='8ab73a6800796d84448bc936db9bac5ad9f984ae'
            )

    def initialize_spark_tts(self):
        """Initialize Spark TTS model by downloading if needed"""
        if not SPARK_AVAILABLE:
            raise RuntimeError("Spark TTS dependencies not available")
        
        model_dir = "pretrained_models/Spark-TTS-0.5B"
        
        # Check if model exists, if not download it
        if not os.path.exists(model_dir):
            print("Downloading Spark-TTS model...")
            try:
                os.makedirs("pretrained_models", exist_ok=True)
                snapshot_download(
                    "SparkAudio/Spark-TTS-0.5B", 
                    local_dir=model_dir
                )
                print("Spark-TTS model downloaded successfully")
            except Exception as e:
                raise RuntimeError(f"Failed to download Spark-TTS model: {e}")
        
        self.spark_model_dir = model_dir
        
        # Check if we have the CLI inference script
        if not os.path.exists("cli/inference.py"):
            print("Warning: Spark-TTS CLI not found. Please clone the Spark-TTS repository.")

    @spaces.GPU(duration=60)
    def initialize_melo_tts(self):
        """Initialize MeloTTS models"""        
        if MELO_AVAILABLE and self.melo_models is None:
            self.melo_models = {"EN": MeloTTS(language="EN", device=self.device)}

    def fetch_text(self, url: str) -> str:
        """Fetch text content from URL"""
        if not url:
            raise ValueError("URL cannot be empty")
            
        if not url.startswith("http://") and not url.startswith("https://"):
            raise ValueError("URL must start with 'http://' or 'https://'")

        full_url = f"{self.config.prefix_url}{url}"
        try:
            response = httpx.get(full_url, timeout=60.0)
            response.raise_for_status()
            return response.text
        except httpx.HTTPError as e:
            raise RuntimeError(f"Failed to fetch URL: {e}")

    def _get_messages_formatter_type(self, model_name):
        """Get appropriate message formatter for the model"""
        if "Mistral" in model_name or "BitSix" in model_name:
            return MessagesFormatterType.CHATML
        else:
            return MessagesFormatterType.LLAMA_3

    def _build_prompt(self, text: str, language: str = "English") -> str:
        """Build prompt for conversation generation"""
        if language == "Korean":
            template = """
            {
                "conversation": [
                    {"speaker": "", "text": ""},
                    {"speaker": "", "text": ""}
                ]
            }
            """
            return (
                f"{text}\n\n제공된 텍스트를 두 명의 전문가 간의 짧고 유익하며 명확한 "
                f"팟캐스트 대화로 변환해주세요. 톤은 전문적이고 매력적이어야 합니다. "
                f"다음 형식을 준수하고 JSON만 반환해주세요:\n{template}"
            )
        else:
            template = """
            {
                "conversation": [
                    {"speaker": "", "text": ""},
                    {"speaker": "", "text": ""}
                ]
            }
            """
            return (
                f"{text}\n\nConvert the provided text into a short, informative and crisp "
                f"podcast conversation between two experts. The tone should be "
                f"professional and engaging. Please adhere to the following "
                f"format and return ONLY the JSON:\n{template}"
            )

    def _build_messages_for_local(self, text: str, language: str = "English") -> List[Dict]:
        """Build messages for local LLM"""
        if language == "Korean":
            system_message = "당신은 한국어로 팟캐스트 대화를 생성하는 전문가입니다. 자연스럽고 유익한 한국어 대화를 만들어주세요."
        else:
            system_message = "You are an expert at creating podcast conversations in English. Create natural and informative English conversations."
        
        return [
            {"role": "system", "content": system_message},
            {"role": "user", "content": self._build_prompt(text, language)}
        ]

    @spaces.GPU(duration=120)
    def translate_to_korean(self, conversation_json: Dict) -> Dict:
        """Translate English conversation to Korean using local LLM"""
        try:
            self.initialize_local_mode()
            
            chat_template = self._get_messages_formatter_type(self.config.local_model_name)
            provider = LlamaCppPythonProvider(self.local_llm)

            system_message = """당신은 전문 번역가입니다. 영어 팟캐스트 대화를 자연스러운 한국어 구어체로 번역해주세요. 
            한국적 표현을 사용하고, 팟캐스트에 적합한 친근하고 자연스러운 말투로 번역하세요. 
            전문 용어는 한국어로 적절히 의역하거나 설명을 추가하세요. JSON 형식을 유지하세요."""

            agent = LlamaCppAgent(
                provider,
                system_prompt=system_message,
                predefined_messages_formatter_type=chat_template,
                debug_output=False
            )
            
            settings = provider.get_provider_default_settings()
            settings.temperature = 0.7
            settings.top_k = 40
            settings.top_p = 0.95
            settings.max_tokens = 2048
            settings.repeat_penalty = 1.1
            settings.stream = False

            messages = BasicChatHistory()
            
            # 번역할 대화를 JSON 문자열로 변환
            conversation_str = json.dumps(conversation_json, ensure_ascii=False, indent=2)
            
            prompt = f"""다음 영어 팟캐스트 대화를 자연스러운 한국어 구어체로 번역해주세요. 
            한국적 표현과 친근한 말투를 사용하고, JSON 형식을 그대로 유지하세요:

            {conversation_str}

            번역된 결과를 JSON 형식으로만 반환해주세요."""
            
            response = agent.get_chat_response(
                prompt,
                llm_sampling_settings=settings,
                chat_history=messages,
                returns_streaming_generator=False,
                print_output=False
            )

            # JSON 파싱
            pattern = r"\{(?:[^{}]|(?:\{[^{}]*\}))*\}"
            json_match = re.search(pattern, response)
            
            if json_match:
                translated_json = json.loads(json_match.group())
                return translated_json
            else:
                print("번역 실패, 원본 반환")
                return conversation_json
                
        except Exception as e:
            print(f"번역 중 오류 발생: {e}, 원본 반환")
            return conversation_json

    @spaces.GPU(duration=120)
    def extract_conversation_local(self, text: str, language: str = "English", progress=None) -> Dict:
        """Extract conversation using new local LLM (primary method)"""
        try:
            # 먼저 새로운 로컬 LLM 시도
            self.initialize_local_mode()
            
            chat_template = self._get_messages_formatter_type(self.config.local_model_name)
            provider = LlamaCppPythonProvider(self.local_llm)

            # 영어로 대화 생성 (일단 영어로 생성하고 한국어 선택시 번역)
            system_message = "You are an expert at creating podcast conversations in English. Create natural and informative English conversations. Respond only in JSON format."

            agent = LlamaCppAgent(
                provider,
                system_prompt=system_message,
                predefined_messages_formatter_type=chat_template,
                debug_output=False
            )
            
            settings = provider.get_provider_default_settings()
            settings.temperature = 0.7
            settings.top_k = 40
            settings.top_p = 0.95
            settings.max_tokens = 2048
            settings.repeat_penalty = 1.1
            settings.stream = False

            messages = BasicChatHistory()
            
            # 영어 프롬프트로 생성
            prompt = self._build_prompt(text, "English")
            response = agent.get_chat_response(
                prompt,
                llm_sampling_settings=settings,
                chat_history=messages,
                returns_streaming_generator=False,
                print_output=False
            )

            # JSON 파싱
            pattern = r"\{(?:[^{}]|(?:\{[^{}]*\}))*\}"
            json_match = re.search(pattern, response)
            
            if json_match:
                conversation_json = json.loads(json_match.group())
                
                # 한국어가 선택된 경우 번역 수행
                if language == "Korean":
                    print("한국어 번역 중...")
                    conversation_json = self.translate_to_korean(conversation_json)
                
                return conversation_json
            else:
                raise ValueError("No valid JSON found in local LLM response")
                
        except Exception as e:
            print(f"Local LLM failed: {e}, falling back to legacy local method")
            return self.extract_conversation_legacy_local(text, language, progress)

    @spaces.GPU(duration=120)
    def extract_conversation_legacy_local(self, text: str, language: str = "English", progress=None) -> Dict:
        """Extract conversation using legacy local model (fallback)"""
        try:
            self.initialize_legacy_local_mode()
            
            # 영어로 대화 생성
            system_message = "You are an expert at creating podcast conversations in English. Create natural and informative English conversations."

            chat = [
                {"role": "system", "content": system_message},
                {"role": "user", "content": self._build_prompt(text, "English")}
            ]

            terminators = [
                self.legacy_tokenizer.eos_token_id,
                self.legacy_tokenizer.convert_tokens_to_ids("<|eot_id|>")
            ]

            messages = self.legacy_tokenizer.apply_chat_template(
                chat, tokenize=False, add_generation_prompt=True
            )
            model_inputs = self.legacy_tokenizer([messages], return_tensors="pt").to(self.device)
            
            streamer = TextIteratorStreamer(
                self.legacy_tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True
            )
            
            generate_kwargs = dict(
                model_inputs,
                streamer=streamer,
                max_new_tokens=4000,
                do_sample=True,
                temperature=0.9,
                eos_token_id=terminators,
            )

            t = Thread(target=self.legacy_local_model.generate, kwargs=generate_kwargs)
            t.start()

            partial_text = ""
            for new_text in streamer:
                partial_text += new_text

            pattern = r"\{(?:[^{}]|(?:\{[^{}]*\}))*\}"
            json_match = re.search(pattern, partial_text)
            
            if json_match:
                conversation_json = json.loads(json_match.group())
                
                # 한국어가 선택된 경우 번역 수행
                if language == "Korean":
                    print("한국어 번역 중...")
                    conversation_json = self.translate_to_korean(conversation_json)
                
                return conversation_json
            else:
                raise ValueError("No valid JSON found in legacy local response")
                
        except Exception as e:
            print(f"Legacy local model also failed: {e}")
            # Return default template
            if language == "Korean":
                return {
                    "conversation": [
                        {"speaker": "진행자", "text": "안녕하세요, 팟캐스트에 오신 것을 환영합니다."},
                        {"speaker": "게스트", "text": "안녕하세요, 초대해 주셔서 감사합니다."}
                    ]
                }
            else:
                return {
                    "conversation": [
                        {"speaker": "Host", "text": "Welcome to our podcast."},
                        {"speaker": "Guest", "text": "Thank you for having me."}
                    ]
                }

    def extract_conversation_api(self, text: str, language: str = "English") -> Dict:
        """Extract conversation using API (fallback method)"""
        if not self.llm_client:
            raise RuntimeError("API mode not initialized")

        try:
            # 영어로 대화 생성
            system_message = "You are an expert at creating podcast conversations in English. Create natural and informative English conversations."

            chat_completion = self.llm_client.chat.completions.create(
                messages=[
                    {"role": "system", "content": system_message},
                    {"role": "user", "content": self._build_prompt(text, "English")}
                ],
                model=self.config.api_model_name,
            )

            pattern = r"\{(?:[^{}]|(?:\{[^{}]*\}))*\}"
            json_match = re.search(pattern, chat_completion.choices[0].message.content)

            if not json_match:
                raise ValueError("No valid JSON found in response")

            conversation_json = json.loads(json_match.group())
            
            # 한국어가 선택된 경우 번역 수행
            if language == "Korean":
                print("한국어 번역 중...")
                conversation_json = self.translate_to_korean(conversation_json)
            
            return conversation_json
        except Exception as e:
            raise RuntimeError(f"Failed to extract conversation: {e}")

    def parse_conversation_text(self, conversation_text: str) -> Dict:
        """Parse conversation text back to JSON format"""
        lines = conversation_text.strip().split('\n')
        conversation_data = {"conversation": []}
        
        for line in lines:
            if ':' in line:
                speaker, text = line.split(':', 1)
                conversation_data["conversation"].append({
                    "speaker": speaker.strip(),
                    "text": text.strip()
                })
        
        return conversation_data

    async def text_to_speech_edge(self, conversation_json: Dict, language: str = "English") -> Tuple[str, str]:
        """Convert text to speech using Edge TTS"""
        output_dir = Path(self._create_output_directory())
        filenames = []

        try:
            # 언어별 음성 설정
            if language == "Korean":
                voices = [
                    "ko-KR-SunHiNeural",  # 여성 음성 (자연스러운 한국어)
                    "ko-KR-HyunsuNeural"  # 남성 음성 (자연스러운 한국어)
                ]
            else:
                voices = [
                    "en-US-AvaMultilingualNeural",    # 여성 음성
                    "en-US-AndrewMultilingualNeural"  # 남성 음성
                ]

            for i, turn in enumerate(conversation_json["conversation"]):
                filename = output_dir / f"output_{i}.wav"
                voice = voices[i % len(voices)]

                tmp_path = await self._generate_audio_edge(turn["text"], voice)
                os.rename(tmp_path, filename)
                filenames.append(str(filename))

            # Combine audio files
            final_output = os.path.join(output_dir, "combined_output.wav")
            self._combine_audio_files(filenames, final_output)
            
            # Generate conversation text
            conversation_text = "\n".join(
                f"{turn.get('speaker', f'Speaker {i+1}')}: {turn['text']}" 
                for i, turn in enumerate(conversation_json["conversation"])
            )
            
            return final_output, conversation_text
        except Exception as e:
            raise RuntimeError(f"Failed to convert text to speech: {e}")

    async def _generate_audio_edge(self, text: str, voice: str) -> str:
        """Generate audio using Edge TTS"""
        if not text.strip():
            raise ValueError("Text cannot be empty")
            
        voice_short_name = voice.split(" - ")[0] if " - " in voice else voice
        communicate = edge_tts.Communicate(text, voice_short_name)

        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
            tmp_path = tmp_file.name
            await communicate.save(tmp_path)

        return tmp_path

    @spaces.GPU(duration=60)
    def text_to_speech_spark(self, conversation_json: Dict, language: str = "English", progress=None) -> Tuple[str, str]:
        """Convert text to speech using Spark TTS CLI"""
        if not SPARK_AVAILABLE or not self.spark_model_dir:
            raise RuntimeError("Spark TTS not available")

        try:
            output_dir = self._create_output_directory()
            audio_files = []
            
            # Create different voice characteristics for different speakers
            if language == "Korean":
                voice_configs = [
                    {"prompt_text": "안녕하세요, 오늘 팟캐스트 진행을 맡은 진행자입니다.", "gender": "female"},
                    {"prompt_text": "안녕하세요, 오늘 게스트로 참여하게 되어 기쁩니다.", "gender": "male"}
                ]
            else:
                voice_configs = [
                    {"prompt_text": "Hello, welcome to our podcast. I'm your host today.", "gender": "female"},
                    {"prompt_text": "Thank you for having me. I'm excited to be here.", "gender": "male"}
                ]

            for i, turn in enumerate(conversation_json["conversation"]):
                text = turn["text"]
                if not text.strip():
                    continue
                
                # Use different voice config for each speaker
                voice_config = voice_configs[i % len(voice_configs)]
                
                output_file = os.path.join(output_dir, f"spark_output_{i}.wav")
                
                # Run Spark TTS CLI inference
                cmd = [
                    "python", "-m", "cli.inference",
                    "--text", text,
                    "--device", "0" if torch.cuda.is_available() else "cpu",
                    "--save_dir", output_dir,
                    "--model_dir", self.spark_model_dir,
                    "--prompt_text", voice_config["prompt_text"],
                    "--output_name", f"spark_output_{i}.wav"
                ]
                
                try:
                    # Run the command
                    result = subprocess.run(
                        cmd, 
                        capture_output=True, 
                        text=True, 
                        timeout=60,
                        cwd="."  # Make sure we're in the right directory
                    )
                    
                    if result.returncode == 0:
                        audio_files.append(output_file)
                    else:
                        print(f"Spark TTS error for turn {i}: {result.stderr}")
                        # Create a short silence as fallback
                        silence = np.zeros(int(22050 * 1.0))  # 1 second of silence
                        sf.write(output_file, silence, 22050)
                        audio_files.append(output_file)
                        
                except subprocess.TimeoutExpired:
                    print(f"Spark TTS timeout for turn {i}")
                    # Create silence as fallback
                    silence = np.zeros(int(22050 * 1.0))
                    sf.write(output_file, silence, 22050)
                    audio_files.append(output_file)
                except Exception as e:
                    print(f"Error running Spark TTS for turn {i}: {e}")
                    # Create silence as fallback
                    silence = np.zeros(int(22050 * 1.0))
                    sf.write(output_file, silence, 22050)
                    audio_files.append(output_file)

            # Combine all audio files
            if audio_files:
                final_output = os.path.join(output_dir, "spark_combined.wav")
                self._combine_audio_files(audio_files, final_output)
            else:
                raise RuntimeError("No audio files generated")

            # Generate conversation text
            conversation_text = "\n".join(
                f"{turn.get('speaker', f'Speaker {i+1}')}: {turn['text']}" 
                for i, turn in enumerate(conversation_json["conversation"])
            )
            
            return final_output, conversation_text
            
        except Exception as e:
            raise RuntimeError(f"Failed to convert text to speech with Spark TTS: {e}")

    @spaces.GPU(duration=60)
    def text_to_speech_melo(self, conversation_json: Dict, progress=None) -> Tuple[str, str]:
        """Convert text to speech using MeloTTS"""
        if not MELO_AVAILABLE or not self.melo_models:
            raise RuntimeError("MeloTTS not available")

        speakers = ["EN-Default", "EN-US"]
        combined_audio = AudioSegment.empty()

        for i, turn in enumerate(conversation_json["conversation"]):
            bio = io.BytesIO()
            text = turn["text"]
            speaker = speakers[i % 2]
            speaker_id = self.melo_models["EN"].hps.data.spk2id[speaker]
            
            # Generate audio
            self.melo_models["EN"].tts_to_file(
                text, speaker_id, bio, speed=1.0, 
                pbar=progress.tqdm if progress else None, 
                format="wav"
            )
            
            bio.seek(0)
            audio_segment = AudioSegment.from_file(bio, format="wav")
            combined_audio += audio_segment

        # Save final audio
        final_audio_path = "melo_podcast.mp3"
        combined_audio.export(final_audio_path, format="mp3")
        
        # Generate conversation text
        conversation_text = "\n".join(
            f"{turn.get('speaker', f'Speaker {i+1}')}: {turn['text']}" 
            for i, turn in enumerate(conversation_json["conversation"])
        )
        
        return final_audio_path, conversation_text

    def _create_output_directory(self) -> str:
        """Create a unique output directory"""
        random_bytes = os.urandom(8)
        folder_name = base64.urlsafe_b64encode(random_bytes).decode("utf-8")
        os.makedirs(folder_name, exist_ok=True)
        return folder_name

    def _combine_audio_files(self, filenames: List[str], output_file: str) -> None:
        """Combine multiple audio files into one"""
        if not filenames:
            raise ValueError("No input files provided")

        try:
            audio_segments = []
            for filename in filenames:
                if os.path.exists(filename):
                    audio_segment = AudioSegment.from_file(filename)
                    audio_segments.append(audio_segment)

            if audio_segments:
                combined = sum(audio_segments)
                combined.export(output_file, format="wav")

            # Clean up temporary files
            for filename in filenames:
                if os.path.exists(filename):
                    os.remove(filename)

        except Exception as e:
            raise RuntimeError(f"Failed to combine audio files: {e}")


# Global converter instance
converter = UnifiedAudioConverter(ConversationConfig())


async def synthesize(article_url: str, mode: str = "Local", tts_engine: str = "Edge-TTS", language: str = "English"):
    """Main synthesis function - Local is now primary, API is fallback"""
    if not article_url:
        return "Please provide a valid URL.", None

    try:
        # Fetch text from URL
        text = converter.fetch_text(article_url)
        
        # Limit text to max words
        words = text.split()
        if len(words) > converter.config.max_words:
            text = " ".join(words[:converter.config.max_words])

        # Extract conversation based on mode
        if mode == "Local":
            # 로컬 모드가 기본 (새로운 Local LLM 사용)
            try:
                conversation_json = converter.extract_conversation_local(text, language)
            except Exception as e:
                print(f"Local mode failed: {e}, trying API fallback")
                # API 폴백
                api_key = os.environ.get("TOGETHER_API_KEY")
                if api_key:
                    converter.initialize_api_mode(api_key)
                    conversation_json = converter.extract_conversation_api(text, language)
                else:
                    raise RuntimeError("Local mode failed and no API key available for fallback")
        else:  # API mode (now secondary)
            api_key = os.environ.get("TOGETHER_API_KEY")
            if not api_key:
                print("API key not found, falling back to local mode")
                conversation_json = converter.extract_conversation_local(text, language)
            else:
                try:
                    converter.initialize_api_mode(api_key)
                    conversation_json = converter.extract_conversation_api(text, language)
                except Exception as e:
                    print(f"API mode failed: {e}, falling back to local mode")
                    conversation_json = converter.extract_conversation_local(text, language)

        # Generate conversation text
        conversation_text = "\n".join(
            f"{turn.get('speaker', f'Speaker {i+1}')}: {turn['text']}" 
            for i, turn in enumerate(conversation_json["conversation"])
        )

        return conversation_text, None
        
    except Exception as e:
        return f"Error: {str(e)}", None


async def regenerate_audio(conversation_text: str, tts_engine: str = "Edge-TTS", language: str = "English"):
    """Regenerate audio from edited conversation text"""
    if not conversation_text.strip():
        return "Please provide conversation text.", None

    try:
        # Parse the conversation text back to JSON format
        conversation_json = converter.parse_conversation_text(conversation_text)
        
        if not conversation_json["conversation"]:
            return "No valid conversation found in the text.", None

        # 한국어인 경우 Edge-TTS만 사용 (다른 TTS는 한국어 지원이 제한적)
        if language == "Korean" and tts_engine != "Edge-TTS":
            return "한국어는 Edge-TTS만 지원됩니다. TTS 엔진이 자동으로 Edge-TTS로 변경됩니다.", None

        # Generate audio based on TTS engine
        if tts_engine == "Edge-TTS":
            output_file, _ = await converter.text_to_speech_edge(conversation_json, language)
        elif tts_engine == "Spark-TTS":
            if not SPARK_AVAILABLE:
                return "Spark TTS not available. Please install required dependencies and clone the Spark-TTS repository.", None
            converter.initialize_spark_tts()
            output_file, _ = converter.text_to_speech_spark(conversation_json, language)
        else:  # MeloTTS
            if not MELO_AVAILABLE:
                return "MeloTTS not available. Please install required dependencies.", None
            if language == "Korean":
                return "MeloTTS does not support Korean. Please use Edge-TTS for Korean.", None
            converter.initialize_melo_tts()
            output_file, _ = converter.text_to_speech_melo(conversation_json)

        return "Audio generated successfully!", output_file
        
    except Exception as e:
        return f"Error generating audio: {str(e)}", None


def synthesize_sync(article_url: str, mode: str = "Local", tts_engine: str = "Edge-TTS", language: str = "English"):
    """Synchronous wrapper for async synthesis"""
    return asyncio.run(synthesize(article_url, mode, tts_engine, language))


def regenerate_audio_sync(conversation_text: str, tts_engine: str = "Edge-TTS", language: str = "English"):
    """Synchronous wrapper for async audio regeneration"""
    return asyncio.run(regenerate_audio(conversation_text, tts_engine, language))


def update_tts_engine_for_korean(language):
    """한국어 선택 시 TTS 엔진 옵션 업데이트"""
    if language == "Korean":
        return gr.Radio(
            choices=["Edge-TTS"],
            value="Edge-TTS",
            label="TTS Engine",
            info="한국어는 Edge-TTS만 지원됩니다",
            interactive=False
        )
    else:
        return gr.Radio(
            choices=["Edge-TTS", "Spark-TTS", "MeloTTS"],
            value="Edge-TTS",
            label="TTS Engine",
            info="Edge-TTS: Cloud-based, natural voices | Spark-TTS: Local AI model | MeloTTS: Local, requires GPU",
            interactive=True
        )


# 모델 초기화 (앱 시작 시)
if LLAMA_CPP_AVAILABLE:
    try:
        model_path = hf_hub_download(
            repo_id=converter.config.local_model_repo,
            filename=converter.config.local_model_name,
            local_dir="./models"
        )
        print(f"Model downloaded to: {model_path}")
    except Exception as e:
        print(f"Failed to download model at startup: {e}")


# Gradio Interface
with gr.Blocks(theme='soft', title="URL to Podcast Converter") as demo:
    gr.Markdown("# 🎙️ URL to Podcast Converter")
    gr.Markdown("Convert any article, blog, or news into an engaging podcast conversation!")
    
    # 상단에 로컬 LLM 상태 표시
    with gr.Row():
        gr.Markdown(f"""
        ### 🤖 LLM Configuration:
        - **Primary**: Local LLM ({converter.config.local_model_name}) - Runs on your device
        - **Fallback**: API LLM ({converter.config.api_model_name}) - Used when local fails
        - **Status**: {"✅ Llama CPP Available" if LLAMA_CPP_AVAILABLE else "❌ Llama CPP Not Available - Install llama-cpp-python"}
        - **한국어 지원**: 영어 대화 생성 후 자연스러운 한국어 구어체로 자동 번역
        """)
    
    with gr.Row():
        with gr.Column(scale=3):
            url_input = gr.Textbox(
                label="Article URL", 
                placeholder="Enter the article URL here...",
                value=""
            )
        with gr.Column(scale=1):
            # 언어 선택 추가
            language_selector = gr.Radio(
                choices=["English", "Korean"],
                value="English",
                label="Language / 언어",
                info="Select output language / 출력 언어를 선택하세요 (한국어 선택 시 자동 번역)"
            )
            
            mode_selector = gr.Radio(
                choices=["Local", "API"],
                value="Local",
                label="Processing Mode",
                info="Local: Runs on device (Primary) | API: Cloud-based (Fallback)"
            )
            
            # TTS 엔진 선택
            with gr.Group():
                gr.Markdown("### TTS Engine Selection")
                tts_selector = gr.Radio(
                    choices=["Edge-TTS", "Spark-TTS", "MeloTTS"],
                    value="Edge-TTS",
                    label="TTS Engine",
                    info="Edge-TTS: Cloud-based, natural voices | Spark-TTS: Local AI model | MeloTTS: Local, requires GPU"
                )
                
                gr.Markdown("""
                **Recommended:**
                - 🌟 **Edge-TTS**: Best quality, cloud-based, instant setup
                - 🤖 **Spark-TTS**: Local AI model (0.5B), zero-shot voice cloning
                
                **Additional Option:**
                - ⚡ **MeloTTS**: Local processing, GPU recommended
                
                **한국어 지원:**
                - 🇰🇷 한국어 선택 시 Edge-TTS만 사용 가능합니다
                - 📝 영어로 대화 생성 후 자연스러운 한국어 구어체로 자동 번역됩니다
                """)
    
    convert_btn = gr.Button("🎯 Generate Conversation / 대화 생성", variant="primary", size="lg")
    
    with gr.Row():
        with gr.Column():
            conversation_output = gr.Textbox(
                label="Generated Conversation (Editable) / 생성된 대화 (편집 가능)",
                lines=15,
                max_lines=30,
                interactive=True,
                placeholder="Generated conversation will appear here. You can edit it before generating audio.\n생성된 대화가 여기에 표시됩니다. 오디오 생성 전에 편집할 수 있습니다.",
                info="Edit the conversation as needed. Format: 'Speaker Name: Text' / 필요에 따라 대화를 편집하세요. 형식: '화자 이름: 텍스트'"
            )
            
            # 오디오 생성 버튼 추가
            with gr.Row():
                generate_audio_btn = gr.Button("🎙️ Generate Audio from Text / 텍스트에서 오디오 생성", variant="secondary", size="lg")
                gr.Markdown("*Edit the conversation above, then click to generate audio / 위의 대화를 편집한 후 클릭하여 오디오를 생성하세요*")
            
        with gr.Column():
            audio_output = gr.Audio(
                label="Podcast Audio / 팟캐스트 오디오",
                type="filepath",
                interactive=False
            )
            
            # 상태 메시지 추가
            status_output = gr.Textbox(
                label="Status / 상태",
                interactive=False,
                visible=True
            )
    
    # TTS 엔진별 설명 및 설치 안내 추가
    with gr.Row():
        gr.Markdown("""
        ### TTS Engine Details / TTS 엔진 상세정보:
        
        - **Edge-TTS**: Microsoft's cloud TTS service with high-quality natural voices. Requires internet connection.
          - 🇰🇷 **한국어 지원**: 자연스러운 한국어 음성 (여성: SunHi, 남성: Hyunsu)
        - **Spark-TTS**: SparkAudio's local AI model (0.5B parameters) with zero-shot voice cloning capability. 
          - **Setup required**: Clone [Spark-TTS repository](https://github.com/SparkAudio/Spark-TTS) in current directory
          - Features: Bilingual support (Chinese/English), controllable speech generation
          - License: CC BY-NC-SA (Non-commercial use only)
          - ⚠️ **한국어 미지원**
        - **MeloTTS**: Local TTS with multiple voice options. GPU recommended for better performance.
          - ⚠️ **한국어 미지원**
        
        ### Local LLM Setup / 로컬 LLM 설정:
        The system now uses **Private-BitSix-Mistral-Small-3.1-24B-Instruct** as the primary LLM, which runs locally on your device for privacy and independence. API fallback is available when needed.
        
        로컬 디바이스에서 개인정보 보호와 독립성을 위해 **Private-BitSix-Mistral-Small-3.1-24B-Instruct**를 기본 LLM으로 사용합니다. 필요시 API 폴백이 제공됩니다.
        
        ### 한국어 번역 기능 / Korean Translation Feature:
        - 한국어를 선택하면 영어로 대화를 생성한 후 자연스러운 한국어 구어체로 자동 번역됩니다
        - 한국적 표현과 친근한 말투를 사용하여 자연스러운 팟캐스트 대화를 만듭니다
        - When Korean is selected, conversations are generated in English first, then automatically translated to natural Korean colloquial style
        
        ### Spark-TTS Setup Instructions:
        ```bash
        git clone https://github.com/SparkAudio/Spark-TTS.git
        cd Spark-TTS
        pip install -r requirements.txt
        ```
        """)
    
    gr.Examples(
        examples=[
            ["https://huggingface.co/blog/openfree/cycle-navigator", "Local", "Edge-TTS", "English"],
            ["https://arxiv.org/html/2505.16938v1", "Local", "Edge-TTS", "English"],            
            ["https://domeggook.com/55204552?from=popular100", "Local", "Edge-TTS", "Korean"],
            ["https://www.wsj.com/articles/nvidia-pushes-further-into-cloud-with-gpu-marketplace-4fba6bdd","Local", "Edge-TTS", "English"],
        ],
        inputs=[url_input, mode_selector, tts_selector, language_selector],
        outputs=[conversation_output, status_output],
        fn=synthesize_sync,
        cache_examples=False,
    )
    
    # 언어 변경 시 TTS 엔진 옵션 업데이트
    language_selector.change(
        fn=update_tts_engine_for_korean,
        inputs=[language_selector],
        outputs=[tts_selector]
    )
    
    # 이벤트 연결
    convert_btn.click(
        fn=synthesize_sync,
        inputs=[url_input, mode_selector, tts_selector, language_selector],
        outputs=[conversation_output, status_output]
    )
    
    generate_audio_btn.click(
        fn=regenerate_audio_sync,
        inputs=[conversation_output, tts_selector, language_selector],
        outputs=[status_output, audio_output]
    )


# Launch the app
if __name__ == "__main__":
    demo.queue(api_open=True, default_concurrency_limit=10).launch(
        show_api=True,
        share=False,
        server_name="0.0.0.0",
        server_port=7860
    )