import streamlit as st
import requests
from transformers import pipeline
from PIL import Image
import torch
import torchvision.transforms as transforms

# Load Fake News Detection Model from Hugging Face
fake_news_pipeline = pipeline("text-classification", model="mrm8488/bert-tiny-finetuned-fake-news-detection")

def classify_text(news_text):
    result = fake_news_pipeline(news_text)[0]
    label = result['label'].lower()
    score = result['score'] * 100  # Convert to percentage
    return ("Fake" if label == "fake" else "Real"), round(score, 2)

def analyze_image(image):
    transform = transforms.Compose([transforms.Resize((224, 224)), transforms.ToTensor()])
    image_tensor = transform(image).unsqueeze(0)
    return "Image analysis feature coming soon!"

def verify_news(news_text):
    search_url = f"https://www.google.com/search?q={'+'.join(news_text.split())}"
    return search_url

# Streamlit UI
st.set_page_config(page_title="Fake News Detector", layout="wide")
st.title("📰 Fake News Detector")

# Dividing Input Section
col1, col2, col3 = st.columns(3)

# Text Input Section
with col1:
    st.subheader("📄 Enter Text")
    news_text = st.text_area("Enter the news content to check:", height=150)
    analyze_text_clicked = st.button("Analyze News")
    if analyze_text_clicked:
        if not news_text.strip():
            st.warning("Please enter some text.")
        else:
            result, accuracy = classify_text(news_text)
            st.session_state["news_text"] = news_text
            st.session_state["result_text"] = result
            st.session_state["accuracy_text"] = accuracy

# Image Upload Section
with col2:
    st.subheader("🖼️ Upload Image")
    uploaded_image = st.file_uploader("Upload a news image", type=["jpg", "png", "jpeg"])
    analyze_image_clicked = st.button("Analyze Image")
    if uploaded_image and analyze_image_clicked:
        image = Image.open(uploaded_image)
        st.session_state["news_image"] = image
        st.image(image, caption="Uploaded Image", use_column_width=True)
        st.info(analyze_image(image))

# Video Link Input Section
with col3:
    st.subheader("🎥 Enter Video Link")
    video_url = st.text_input("Enter the video link:")
    analyze_video_clicked = st.button("Analyze Video")
    if analyze_video_clicked:
        if not video_url.strip():
            st.warning("Please enter a valid video link.")
        else:
            st.session_state["video_url"] = video_url
            st.video(video_url)
            st.info("Video analysis feature coming soon!")

# Results Section
st.subheader("📊 Analysis Results")
if "result_text" in st.session_state:
    result = st.session_state.get("result_text")
    accuracy = st.session_state.get("accuracy_text")
    verification_link = verify_news(st.session_state.get("news_text"))
    
    if result == "Fake":
        st.error(f"❌ This news is likely **Fake**! (Accuracy: {accuracy}%)", icon="⚠️")
    else:
        st.success(f"✅ This news is likely **Real**! (Accuracy: {accuracy}%)", icon="✅")
    
    st.subheader("🔍 Verification & Trusted Sources")
    sources = [
        "https://www.bbc.com/news",
        "https://www.cnn.com",
        "https://www.reuters.com",
        "https://factcheck.org",
        "https://www.snopes.com",
        "https://www.politifact.com",
        "https://deepfake-o-meter.ai",
        "https://huggingface.co/models?pipeline_tag=text-classification"
    ]
    for link in sources:
        st.markdown(f"[🔗 {link}]({link})")
    st.markdown(f"[🔎 Verify on Google]({verification_link})")