app.py

import gradio as gr
import pandas as pd
from huggingface_hub import hf_hub_download
import json
from datetime import datetime
import os

def load_data():
    """Load the latest news data from the parquet file"""
    try:
        file_path = hf_hub_download(
            repo_id='rajatarya/hf-news-aggregator', 
            filename='data/news_data_summaries.parquet', 
            repo_type='dataset'
        )
        df = pd.read_parquet(file_path)
        # Sort by date_generated descending to get latest first
        df = df.sort_values('date_generated', ascending=False).reset_index(drop=True)
        return df
    except Exception as e:
        print(f"Error loading data: {e}")
        return pd.DataFrame()

def format_metadata(metadata_str):
    """Format metadata as a readable string"""
    try:
        if isinstance(metadata_str, str):
            metadata = json.loads(metadata_str)
        else:
            metadata = metadata_str
        
        formatted = []
        for key, value in metadata.items():
            if key == 'generated_at':
                # Format timestamp
                try:
                    dt = datetime.fromisoformat(value.replace('Z', '+00:00'))
                    formatted.append(f"**Generated:** {dt.strftime('%Y-%m-%d %H:%M UTC')}")
                except:
                    formatted.append(f"**Generated:** {value}")
            else:
                formatted.append(f"**{key.replace('_', ' ').title()}:** {value}")
        
        return "<br>".join(formatted)
    except:
        return str(metadata_str)

def get_news_content(index):
    """Get news content for a specific index"""
    df = load_data()
    
    if df.empty or index >= len(df):
        return (
            "❌ No data available", 
            "",
            f"0 / 0",
            gr.update(interactive=False),
            gr.update(interactive=False)
        )
    
    row = df.iloc[index]
    
    # Create metadata table
    metadata_md = f"""
| Type | {row['summary_type'].title()} |
|-----------------|---------|
| Date Generated | {row['date_generated'].strftime('%Y-%m-%d %H:%M UTC')} | 
| Time Range | {row['time_range']} |
| Articles Analyzed | {row['num_articles_analyzed']} |
| Sources | {("<br>".join([s.strip() for s in row['sources'].split(',')]) if isinstance(row['sources'], str) else str(row['sources']))} |
| Provider | {row['provider']} |
| Metadata | {json.dumps(row['metadata'], indent=2, default=str, sort_keys=True) if row['metadata'] else ""} |"""
    
    # Navigation info
    nav_info = f"{index + 1} / {len(df)}"
    
    # Navigation buttons state
    prev_disabled = index >= len(df) - 1
    next_disabled = index <= 0
    
    return (
        row['summarized_markdown'],
        metadata_md,
        nav_info,
        gr.update(interactive=not prev_disabled),
        gr.update(interactive=not next_disabled)
    )

# Global state for current index
current_index = 0

def next_news():
    global current_index
    current_index = max(0, current_index - 1)  # Newer news (lower index)
    return get_news_content(current_index)

def prev_news():
    global current_index
    df = load_data()
    current_index = min(len(df) - 1, current_index + 1)  # Older news (higher index)
    return get_news_content(current_index)

def refresh_data():
    global current_index
    current_index = 0  # Reset to latest
    return get_news_content(current_index)

# Create the Gradio interface
with gr.Blocks(
    title="HF News (AI-assisted summary)",
    theme=gr.themes.Soft(),
    css="""
    * {
        font-family: Arial, Helvetica, sans-serif !important;
    }
    .news-container {
        max-width: 1200px;
    }
    .nav-button {
        font-size: 16px !important;
        padding: 10px 20px !important;
    }
    .content-shadow {
        box-shadow: 0 4px 8px rgba(0, 0, 0, 0.15) !important;
        padding: 20px !important;
        background: white !important;
        margin: 20px 0 !important;
    }
    """
) as app:
    gr.Markdown("""
        # 🚀 🤗 HF News (AI-assisted summary) 📰
        
        Stay updated with the latest AI and ML news, summarized by AI! ✨
    """)

    with gr.Row(variant="panel", elem_classes="content-shadow"):
        content_display = gr.Markdown(
            label="📄 Content",
            value="Loading news data... 🔄",
        )
    
    # Metadata section (collapsible, collapsed by default)
    with gr.Accordion("📊 Summary Details", open=False):
        metadata_display = gr.Markdown(
            value=""
        )
    
    # Refresh button (moved below metadata)
    with gr.Row():
        refresh_btn = gr.Button("🔄 Refresh Data", variant="secondary")
    
    with gr.Row():
        prev_btn = gr.Button("⬅️ Older News", variant="secondary", elem_classes="nav-button")
        nav_info = gr.Textbox(
            label="📍 Position", 
            interactive=False,
            container=False
        )
        next_btn = gr.Button("➡️ Newer News", variant="primary", elem_classes="nav-button")
    
    gr.HTML("""
    <div style="text-align: center; padding: 20px; color: #666; font-size: 14px;">
        <p>🤖 Powered by AI • 📊 Data from <a href="https://huggingface.co/datasets/rajatarya/hf-news-aggregator" target="_blank">HF News Aggregator</a></p>
    </div>
    """)
    
    # Event handlers
    app.load(
        fn=lambda: get_news_content(0),
        outputs=[content_display, metadata_display, nav_info, prev_btn, next_btn]
    )
    
    next_btn.click(
        fn=next_news,
        outputs=[content_display, metadata_display, nav_info, prev_btn, next_btn]
    )
    
    prev_btn.click(
        fn=prev_news,
        outputs=[content_display, metadata_display, nav_info, prev_btn, next_btn]
    )
    
    refresh_btn.click(
        fn=refresh_data,
        outputs=[content_display, metadata_display, nav_info, prev_btn, next_btn]
    )

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