import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
import feedparser
import threading
from datetime import datetime
import json
import os

# Initialize Llama 2 model and tokenizer
model_name = "meta-llama-2-1b"  # Replace with your actual model name
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

def fetch_news_from_rss(interests):
    """Fetch news from RSS feeds based on interests"""
    # Example RSS feeds - you can expand this list
    rss_feeds = {
        "Technology": "https://feeds.feedburner.com/TechCrunch",
        "Business": "https://feeds.feedburner.com/BusinessInsider",
        "Sports": "https://www.espn.com/espn/rss/news",
    }
    
    articles = []
    for interest in interests:
        if interest in rss_feeds:
            feed = feedparser.parse(rss_feeds[interest])
            articles.extend(feed.entries[:3])  # Get top 3 articles per interest
    
    return articles

def generate_summary(text, language="English"):
    """Generate summary using Llama 2"""
    prompt = f"""Please provide a concise summary of the following news article in {language}. 
    Focus on the main points and key details:

    Article: {text}

    Summary:"""
    
    inputs = tokenizer(prompt, return_tensors="pt", max_length=512, truncation=True)
    outputs = model.generate(
        inputs["input_ids"],
        max_length=150,
        min_length=50,
        temperature=0.7,
        num_return_sequences=1
    )
    
    summary = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return summary.split("Summary:")[1].strip()

def save_user_preferences(name, language, interests):
    """Save user preferences to a JSON file"""
    if not name or not language or not interests:
        return "Please fill in all fields!"
    
    preferences = {
        "name": name,
        "language": language,
        "interests": interests,
        "last_updated": datetime.now().isoformat()
    }
    
    with open(f"preferences_{name}.json", "w") as f:
        json.dump(preferences, f)
    
    return f"Preferences saved for {name}!"

def get_personalized_summary(name):
    """Get personalized news summary based on user preferences"""
    try:
        with open(f"preferences_{name}.json", "r") as f:
            preferences = json.load(f)
    except FileNotFoundError:
        return "Please set your preferences first!"

    # Fetch news based on interests
    articles = fetch_news_from_rss(preferences["interests"])
    
    summaries = []
    for article in articles:
        title = article.get("title", "")
        content = article.get("description", "")
        
        # Generate summary using Llama 2
        summary = generate_summary(
            content, 
            language=preferences["language"]
        )
        
        summaries.append(f"📰 {title}\n\n{summary}\n\n---")
    
    return "\n".join(summaries)

# Create Gradio interface
with gr.Blocks(title="Llama 2 News Summarizer") as demo:
    gr.Markdown("# 📰 AI News Summarizer powered by Llama 2")
    
    with gr.Tab("Set Preferences"):
        name_input = gr.Textbox(label="Your Name")
        language_dropdown = gr.Dropdown(
            choices=["English", "Spanish", "French", "Arabic"],
            label="Preferred Language"
        )
        interests_checkboxes = gr.CheckboxGroup(
            choices=["Technology", "Business", "Sports", "Science", "Politics"],
            label="News Interests"
        )
        save_button = gr.Button("Save Preferences")
        preferences_output = gr.Textbox(label="Status")
        
        save_button.click(
            save_user_preferences,
            inputs=[name_input, language_dropdown, interests_checkboxes],
            outputs=[preferences_output]
        )
    
    with gr.Tab("Get News Summary"):
        name_check = gr.Textbox(label="Enter your name to get summary")
        get_summary_button = gr.Button("Get Summary")
        summary_output = gr.Textbox(
            label="Your Personalized News Summary",
            lines=10
        )
        
        get_summary_button.click(
            get_personalized_summary,
            inputs=[name_check],
            outputs=[summary_output]
        )

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