app.py

import os
import json
import datetime
import gradio as gr
import pandas as pd
import numpy as np

from src.about import (
    CITATION_BUTTON_LABEL,
    CITATION_BUTTON_TEXT,
    EVALUATION_QUEUE_TEXT,
    INTRODUCTION_TEXT,
    LLM_BENCHMARKS_TEXT,
    TITLE,
)
from src.display.css_html_js import custom_css
from src.display.utils import (
    BENCHMARK_COLS,
    COLS,
    EVAL_COLS,
    EVAL_TYPES,
    AutoEvalColumn,
    ModelType,
    fields,
    WeightType,
    Precision
)

# SAGE specific imports
try:
    from src.leaderboard.sage_eval import load_initial_sage_results, SAGEResult
    from src.submission.sage_submit import process_sage_submission_simple
    from src.populate import get_sage_leaderboard_df
    SAGE_MODULES_AVAILABLE = True
except ImportError as e:
    print(f"Warning: SAGE modules not available: {e}")
    SAGE_MODULES_AVAILABLE = False


# Configuration
TOKEN = os.environ.get("HF_TOKEN", None)
OWNER = "opencompass"

def format_error(msg):
    return f"<p style='color: red; font-size: 20px; text-align: center;'>{msg}</p>"

def format_warning(msg):
    return f"<p style='color: orange; font-size: 20px; text-align: center;'>{msg}</p>"

def format_log(msg):
    return f"<p style='color: green; font-size: 20px; text-align: center;'>{msg}</p>"

def model_hyperlink(link, model_name):
    if link and link.startswith("http"):
        return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'
    return model_name

def get_leaderboard_dataframe():
    """Generate leaderboard dataframe from SAGE results"""
    if not SAGE_MODULES_AVAILABLE:
        return pd.DataFrame()
        
    sage_results = load_initial_sage_results()
    
    if not sage_results:
        return pd.DataFrame()
    
    # Convert to leaderboard format
    leaderboard_data = []
    for result in sage_results:
        # Extract model name from submission_id
        if result.submission_id.startswith("initial_"):
            model_name = result.submission_id.split("_", 2)[-1].replace("_", " ")
        else:
            model_name = result.submission_id
        
        # Create model hyperlink (for now just display name)
        model_display = f"**{model_name}**"
        
        row = {
            "Model": model_display,
            "Organization": result.organization,
            "Overall (%)": result.results.get("sage_overall", 0),
            "Mathematics (%)": result.results.get("sage_math", 0),
            "Physics (%)": result.results.get("sage_physics", 0),
            "Chemistry (%)": result.results.get("sage_chemistry", 0),
            "Biology (%)": result.results.get("sage_biology", 0),
            "Earth Science (%)": result.results.get("sage_earth_science", 0),
            "Astronomy (%)": result.results.get("sage_astronomy", 0),
            "Submission Date": result.submitted_time
        }
        leaderboard_data.append(row)
    
    df = pd.DataFrame(leaderboard_data)
    if not df.empty:
        df = df.sort_values(by=["Overall (%)"], ascending=False)
    
    return df

def refresh_leaderboard():
    """Refresh the leaderboard data"""
    print("🔄 Refreshing leaderboard data...")
    return get_leaderboard_dataframe()

# Initialize data
leaderboard_df = get_leaderboard_dataframe()

# Define column types for the dataframe
COLUMN_TYPES = ["markdown", "str", "number", "number", "number", "number", "number", "number", "number", "str"]


# Create Gradio interface
demo = gr.Blocks(css="""
.markdown-text {
    font-size: 16px !important;
}
#citation-button {
    font-family: monospace;
}
""")

with demo:
    gr.HTML(TITLE)
    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")

    with gr.Row():
        with gr.Accordion("📙 Citation", open=False):
            citation_button = gr.Textbox(
                value=CITATION_BUTTON_TEXT,
                label=CITATION_BUTTON_LABEL,
                elem_id="citation-button",
                lines=10,
                max_lines=10,
                interactive=False
            )

    # Main leaderboard table
    gr.Markdown("## 🏆 SAGE Benchmark Results", elem_classes="markdown-text")
    leaderboard_table = gr.Dataframe(
        value=leaderboard_df,
        datatype=COLUMN_TYPES,
        interactive=False,
        wrap=True,
        column_widths=["25%", "15%", "8%", "8%", "8%", "8%", "8%", "8%", "8%", "12%"]
    )

    # Refresh button
    refresh_button = gr.Button("🔄 Refresh Leaderboard")
    refresh_button.click(
        refresh_leaderboard,
        inputs=[],
        outputs=[leaderboard_table]
    )

    # Submission section
    with gr.Accordion("📊 Submit Your SAGE Results", open=False):
        gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
        
        with gr.Row():
            with gr.Column():
                org_textbox = gr.Textbox(label="Organization Name", placeholder="Your Organization")
                email_textbox = gr.Textbox(label="Contact Email", placeholder="contact@example.com")
            with gr.Column():
                file_upload = gr.File(
                    label="Upload SAGE Results (JSON)",
                    file_types=[".json"],
                    type="filepath"
                )

        submit_button = gr.Button("Submit Results", variant="primary")
        submission_result = gr.HTML()
        
        if SAGE_MODULES_AVAILABLE:
            submit_button.click(
                process_sage_submission_simple,
                inputs=[file_upload, org_textbox, email_textbox],
                outputs=[submission_result]
            ).then(
                refresh_leaderboard,  # Auto-refresh after submission
                inputs=[],
                outputs=[leaderboard_table]
            )
        else:
            submit_button.click(
                lambda: format_error("SAGE submission system not available"),
                inputs=[],
                outputs=[submission_result]
            )

# Launch the app
if __name__ == "__main__":
    demo.launch(debug=True)