app.py

import gradio as gr
from newspaper import Article
from modules.online_search import search_online
from modules.validation import calculate_truthfulness_score
from modules.knowledge_graph import search_kg
from modules.generate_explanation import generate_explanation  # Import the explanation generator
from dotenv import load_dotenv
import os
from concurrent.futures import ThreadPoolExecutor

# Load environment variables from .env file
load_dotenv()

# Constants
KG_INDEX_PATH = "KG/news_category_index.faiss"
KG_DATASET_PATH = "KG/News_Category_Dataset_v3.json"
SEARCH_API_KEY = os.getenv("SEARCH_API_KEY")
SEARCH_BASE_URL = os.getenv("SEARCH_BASE_URL")
SEARCH_MODEL = os.getenv("SEARCH_MODEL")

# Initialize ThreadPoolExecutor
executor = ThreadPoolExecutor(max_workers=3)  # Increased workers to accommodate explanation task

# Function to process input and evaluate truthfulness
def evaluate_news(news_input):
    # Display loading message
    yield "**Processing... Please wait while we analyze the information.** ⏳"

    # Handle URL input
    if news_input.startswith("http"):
        try:
            article = Article(news_input)
            article.download()
            article.parse()
            news_text = article.title + ". " + article.text
        except Exception as e:
            yield f"**Error processing the URL:** {str(e)}"
            return
    else:
        # Direct text input
        news_text = news_input

    try:
        # Run both search functions concurrently using ThreadPoolExecutor
        future_kg = executor.submit(search_kg, news_text, KG_INDEX_PATH, KG_DATASET_PATH)
        future_online = executor.submit(search_online, news_text, SEARCH_API_KEY, SEARCH_BASE_URL, SEARCH_MODEL)

        # Wait for the results of both tasks
        kg_content = future_kg.result()
        online_search_results = future_online.result()

        # Combine context from KG and online search
        context = online_search_results['message_content'] + '\n' + kg_content + '\n' + 'Device set to use cpu'
        # print(context)  # Debug log

        # Calculate truth score
        truth_score = calculate_truthfulness_score(info=news_text, context=context)

        # Determine truthfulness status and recommendation
        if truth_score > 0.7:
            status = "likely true"
            recommendation = "You can reasonably trust this information, but further verification is always recommended for critical decisions."
        elif truth_score > 0.4:
            status = "uncertain"
            recommendation = "This information might be partially true, but additional investigation is required before accepting it as fact."
        else:
            status = "unlikely to be true"
            recommendation = "It is recommended to verify this information through multiple reliable sources before trusting it."

        # Display initial result with score
        result = f"**News**: \"{news_text[:300]}...\"\n\n"
        result += f"**Truthfulness Score**: {truth_score:.2f} (**{status.capitalize()}**)\n\n"
        result += f"**Analysis**: {recommendation}\n\n"
        yield result  # Immediately display score and recommendation

        # Generate explanation asynchronously
        future_explanation = executor.submit(generate_explanation, news_text, context, truth_score)

        # Add explanation and sources once available
        explanation = future_explanation.result()  # Wait for explanation result
        if explanation:
            result += f"**Explanation**: {explanation}\n\n"  # Append explanation

            # Add sources from the online search results (top 5 sources)
            sources = online_search_results.get('sources', [])
            if sources:
                result += "\n**Sources**:\n"
                # Ensure we only show up to 5 sources
                for i, source in enumerate(sources[:5]):
                    result += f"{i + 1}. {source}\n"
                result += "\n*Please make sure to do your own research for more confirmation and to cross-check the information.*"

            yield result  # Update UI with explanation and sources

    except Exception as e:
        yield f"**Error occurred while processing the input:** {str(e)}"


# Gradio Interface
with gr.Blocks() as demo:
    gr.Markdown("# 📰 EchoTruth: Verify News Authenticity in Real-Time")
    
    gr.Markdown("""
    **How to use:**
    1. Enter a news article or URL in the box below.
    2. Click on **Check Truthfulness**.
    3. Receive a **truthfulness score** along with **explanations and sources** to help you assess the authenticity of the content.
    """)
    
    with gr.Row():
        input_box = gr.Textbox(
            placeholder="Enter news text or URL here... (e.g., https://example.com)",
            label="Input News or URL",
            lines=5
        )

    submit_btn = gr.Button("Check Truthfulness")

    output_box = gr.Markdown()

    submit_btn.click(
        fn=evaluate_news,
        inputs=[input_box],
        outputs=[output_box]
    )

    gr.Markdown("### **About EchoTruth**")
    gr.Markdown("EchoTruth uses AI to help users verify news authenticity in real-time. We recommend checking multiple sources before making decisions based on news.")
    
demo.launch()