import torch
from transformers import pipeline, set_seed, AutoTokenizer, AutoModelForSequenceClassification
from datasets import load_dataset
import gradio as gr
import requests
from bs4 import BeautifulSoup
from nltk.sentiment import SentimentIntensityAnalyzer
from flair.models import TextClassifier
from flair.data import Sentence
import newspaper3k

# Konfigurasi Model
set_seed(42)
nltk.download('vader_lexicon')

# Model Text Generation
generator = pipeline('text-generation', model='gpt2-xl')  # Model lebih canggih

# Model Sentiment Analysis
sentiment_analyzer = SentimentIntensityAnalyzer()
classifier = TextClassifier.load('en-sentiment')

# Model Klasifikasi Topik
tokenizer_topic = AutoTokenizer.from_pretrained("facebook/bart-large-mnli")
model_topic = AutoModelForSequenceClassification.from_pretrained("facebook/bart-large-mnli")

# Fungsi Helper
def get_trending_topics(platform):
  if platform == "Twitter":
    # Scraping trending topics dari Twitter
    url = "https://twitter.com/i/trends"
    response = requests.get(url)
    soup = BeautifulSoup(response.content, "html.parser")
    trends = [trend.text.strip() for trend in soup.find_all("div", class_="css-901oao r-1awozwy r-18jsvk2 r-6koalj r-370sk r-a023e6 r-b88u0q r-rjixqe r-bcqeeo r-1udh08x r-3s2u2q r-qvutc0")]
    return trends
  elif platform == "TikTok":
    # Scraping trending topics dari TikTok (perlu metode khusus)
    # ...
    return ["Trending TikTok 1", "Trending TikTok 2"]
  else:  # Instagram
    # Scraping trending topics dari Instagram (perlu metode khusus)
    # ...
    return ["Trending Instagram 1", "Trending Instagram 2"]

def find_related_trend(topic, trends):
  # Menggunakan model klasifikasi topik untuk mencari tren yang relevan
  topic_sentence = Sentence(topic)
  classifier.predict(topic_sentence)
  topic_sentiment = topic_sentence.labels[0]

  related_trends = []
  for trend in trends:
    trend_sentence = Sentence(trend)
    classifier.predict(trend_sentence)
    trend_sentiment = trend_sentence.labels[0]
    if topic_sentiment.value == trend_sentiment.value:
      related_trends.append(trend)
  return related_trends

def make_clickbait_title(title):
  # Menggunakan pola clickbait dan analisis sentimen
  title = title.strip()
  sentiment = sentiment_analyzer.polarity_scores(title)

  if sentiment['compound'] > 0.5:
    # Positif -> "Rahasia...", "Terungkap...", dll.
    clickbait_phrases = ["Rahasia ", "Terungkap ", "Hebat! ", "Menakjubkan! "]
  elif sentiment['compound'] < -0.5:
    # Negatif -> "Mengerikan...", "Kontroversial...", dll.
    clickbait_phrases = ["Mengerikan! ", "Kontroversial! ", "Awas! ", "Bahaya! "]
  else:
    # Netral -> "Kamu Tidak Akan Percaya...", "Heboh...", dll.
    clickbait_phrases = ["Kamu Tidak Akan Percaya...", "Heboh! ", "Viral! ", "Trending! "]

  return clickbait_phrases[0] + title

def analyze_content(content):
  # Analisis sentimen dan topik
  sentence = Sentence(content)
  classifier.predict(sentence)
  sentiment = sentence.labels[0]

  # Klasifikasi topik (perlu pengembangan lebih lanjut)
  inputs = tokenizer_topic(content, return_tensors="pt")
  outputs = model_topic(**inputs)
  topic_probs = torch.softmax(outputs.logits, dim=1)
  # ... (Interpretasi hasil klasifikasi topik) ...
  return sentiment, "Topik yang Diprediksi"

def generate_content(topic, format, tone, platform):
  trending_topics = get_trending_topics(platform)
  related_trends = find_related_trend(topic, trending_topics)

  prompt = f"Buat konten {format} tentang {topic} dengan gaya {tone} yang "
  if related_trends:
    prompt += f"berkaitan dengan tren {', '.join(related_trends)}."
  else:
    prompt += f"berpotensi menjadi viral."

  output = generator(prompt, max_length=500, num_return_sequences=1)
  content = output[0]['generated_text']

  if format == "Judul":
    content = make_clickbait_title(content)

  sentiment, predicted_topic = analyze_content(content)
  return content, sentiment.value, predicted_topic

# Antarmuka Gradio
iface = gr.Interface(
    fn=generate_content,
    inputs=[
        gr.Textbox(lines=2, placeholder="Masukkan topik..."),
        gr.Dropdown(["Teks", "Judul", "Tweet", "Artikel"], label="Format Konten"),
        gr.Dropdown(["Netral", "Provokatif", "Humor", "Marah", "Sedih"], label="Tone"),
        gr.Dropdown(["Twitter", "TikTok", "Instagram"], label="Platform")
    ],
    outputs=[
        "text",
        "text",
        "text"
    ],
    title="Pembuat Konten Viral (At Any Cost)",
    description="Hasilkan konten yang dirancang untuk menjadi viral (Eksperimental)."
)

iface.launch(share=True)  # share=True untuk mendapatkan link publik