app.py

import pandas as pd
import gradio as gr

# Load the CSV file
model_test_results = pd.read_csv('test_results_by_type.csv')
# Get models with failed tests and their failure counts
failed_models_counts = model_test_results[
    (model_test_results['test_type'] == 'failed') & 
    (model_test_results['number_of_tests'] > 0)
].groupby('model')['number_of_tests'].first().to_dict()

# Add ❌ and failure count to model names that have failures, ✅ for passing models
model_test_results['model'] = model_test_results.apply(
    lambda row: f"{row['model']} ❌ ({failed_models_counts[row['model']]})" if row['model'] in failed_models_counts else f"{row['model']} ✅",
    axis=1
)

# Separate failed tests and other tests
failed_tests = model_test_results[model_test_results['test_type'] == 'failed'].sort_values('number_of_tests', ascending=False)
other_tests = model_test_results[model_test_results['test_type'] != 'failed']

# Concatenate the dataframes
model_test_results = pd.concat([failed_tests, other_tests])


# Create the Gradio interface
with gr.Blocks() as test_results_viz:
    gr.Markdown("# Test Results by Model")
    
    # Sort models by success/failure and number of failed tests
    model_order = model_test_results.sort_values(
        by=['conclusion', 'test_type', 'number_of_tests'],
        ascending=[True, False, False]
    )['model'].unique().tolist()
    
    # Create the stacked bar plot using Gradio's BarPlot
    test_results_plot = gr.BarPlot(
        model_test_results,
        x="model",
        y="number_of_tests",  # Base layer
        color="test_type",  # Color by pass/fail status
        color_map={"passed": "#008550", "skipped": "#F0B702", "failed": "#8B1710"},
        title="Test Results by Model",
        x_title="Model", 
        y_title="Number of Tests",
        height=600,
        width=1000,
        x_label_angle=45,  # Rotate x-axis labels by 45 degrees
        x_order=model_order  # Set custom order of x-axis
    )

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