app.py

import requests
from bs4 import BeautifulSoup
import pandas as pd
import gradio as gr
import plotly.express as px

BASE_URL = "https://scale.com/leaderboard"

LEADERBOARDS = {
    "Coding": "/coding",
    "Adversarial Robustness": "/adversarial_robustness",
    "Instruction Following": "/instruction_following",
    "Math": "/math"
}

def scrape_leaderboard(leaderboard):
    url = BASE_URL + LEADERBOARDS[leaderboard]
    
    response = requests.get(url)
    soup = BeautifulSoup(response.content, 'html.parser')
    
    leaderboard_div = soup.find('div', class_='flex flex-col gap-4 sticky top-20')
    
    if not leaderboard_div:
        raise ValueError("Leaderboard div not found. The page structure might have changed.")
    
    table = leaderboard_div.find('table', class_='w-full caption-bottom text-sm')
    
    if not table:
        raise ValueError("Leaderboard table not found within the div.")
    
    data = []
    for row in table.find('tbody').find_all('tr'):
        cols = row.find_all('td')
        rank = cols[0].find('div', class_='flex').text.strip().split()[0]
        model = cols[0].find('a').text.strip()
        score = cols[1].text.strip()
        confidence = cols[2].text.strip()
        data.append([rank, model, score, confidence])
    
    df = pd.DataFrame(data, columns=['Rank', 'Model', 'Score', '95% Confidence'])
    df['Score'] = pd.to_numeric(df['Score'])  # Convert Score to numeric
    return df

def create_chart(df):
    fig = px.bar(df, x='Model', y='Score', title='Model Scores Comparison',
                 labels={'Score': 'Overall Score', 'Model': 'Model Name'},
                 color='Score', color_continuous_scale='viridis')
    fig.update_layout(xaxis_tickangle=-45, xaxis_title=None)
    return fig

def update_leaderboard(leaderboard):
    try:
        df = scrape_leaderboard(leaderboard)
        chart = create_chart(df)
        return df, df.to_html(index=False), chart
    except Exception as e:
        return None, f"An error occurred: {str(e)}", None

# Create Gradio interface
with gr.Blocks() as iface:
    gr.Markdown("# Scale AI Leaderboard Viewer")
    dropdown = gr.Dropdown(choices=list(LEADERBOARDS.keys()), label="Select Leaderboard", value="Coding")
    
    table_output = gr.HTML()
    chart_output = gr.Plot()
    
    def on_load():
        df, html, chart = update_leaderboard("Coding")
        return html, chart
    
    dropdown.change(update_leaderboard, inputs=[dropdown], outputs=[table_output, table_output, chart_output])
    
    iface.load(on_load, outputs=[table_output, chart_output])

# Launch the app
iface.launch()