example_usage.py

"""
example_usage.py

Ejemplo de uso de la clase ImageFilterCLIP optimizada.
Demuestra cómo usar la clase con diferentes configuraciones y casos de uso.
"""

import json
from image_filter_class import ImageFilterCLIP


def basic_usage_example():
    """Ejemplo básico de uso de la clase."""
    print("=" * 60)
    print("EJEMPLO BÁSICO")
    print("=" * 60)
    
    # Crear instancia con configuración por defecto
    filter_instance = ImageFilterCLIP()
    
    # URLs de ejemplo (reemplazar con URLs reales para probar)
    image_urls = [
        "https://picsum.photos/400/300?random=1",
        "https://picsum.photos/400/300?random=2", 
        "https://picsum.photos/400/300?random=3"
    ]
    
    # Filtrar imágenes
    results = filter_instance.filter_images(
        description="landscape with trees and mountains",
        image_urls=image_urls
    )
    
    # Mostrar resumen de resultados
    print(f"Descripción: {results['description']}")
    print(f"Modelo: {results['model']}")
    print(f"Tiempo total: {results['processing_time_seconds']}s")
    print(f"Estadísticas: {results['statistics']}")
    print()
    
    # Mostrar resultados detallados
    for result in results['results']:
        state = result['state']
        pos_score = result['score_positive_pct']
        neg_score = result['score_negative_pct']
        print(f"  {result['image'][:50]}... -> {state} (pos: {pos_score}%, neg: {neg_score}%)")
    
    return results


def high_performance_example():
    """Ejemplo con configuración optimizada para alta performance."""
    print("\n" + "=" * 60)
    print("EJEMPLO DE ALTA PERFORMANCE")
    print("=" * 60)
    
    # Configuración optimizada
    high_perf_config = {
        "batch_size": 20,           # Batch más grande
        "max_workers": 12,          # Más workers para descarga
        "positive_threshold": 0.25,  # Umbral más permisivo para speed
        "negative_threshold": 0.30,
        "use_cache": True,          # Activar cache
        "timeout": 5                # Timeout más corto
    }
    
    filter_instance = ImageFilterCLIP(high_perf_config)
    
    # Simular muchas URLs
    image_urls = [
        f"https://picsum.photos/300/200?random={i}" 
        for i in range(10, 20)
    ]
    
    results = filter_instance.filter_images(
        description="beautiful nature scene",
        image_urls=image_urls
    )
    
    print(f"Procesadas {len(results['results'])} imágenes")
    print(f"Tiempo de descarga: {results['download_time_seconds']}s")
    print(f"Tiempo total: {results['processing_time_seconds']}s")
    print(f"Promedio por imagen: {results['processing_time_seconds']/len(image_urls):.2f}s")
    
    return results


def strict_filtering_example():
    """Ejemplo con filtrado estricto para minimizar falsos positivos."""
    print("\n" + "=" * 60)
    print("EJEMPLO DE FILTRADO ESTRICTO")
    print("=" * 60)
    
    # Configuración estricta
    strict_config = {
        "positive_threshold": 0.35,   # Más estricto
        "negative_threshold": 0.20,   # Más sensible a contenido negativo
        "batch_size": 8,              # Batches más pequeños para precisión
        "negative_prompts": [
            "nudity", "naked person", "sexual content", "adult content",
            "violence", "blood", "gore", "weapons", "guns",
            "inappropriate content", "disturbing image",
            "drug use", "alcohol", "smoking",
            "hate symbols", "offensive content"
        ]
    }
    
    filter_instance = ImageFilterCLIP(strict_config)
    
    image_urls = [
        "https://picsum.photos/400/300?random=21",
        "https://picsum.photos/400/300?random=22"
    ]
    
    results = filter_instance.filter_images(
        description="family-friendly content with children playing",
        image_urls=image_urls
    )
    
    print("Configuración estricta aplicada:")
    print(f"  Umbral positivo: {strict_config['positive_threshold']}")
    print(f"  Umbral negativo: {strict_config['negative_threshold']}")
    print(f"  Prompts negativos: {len(strict_config['negative_prompts'])}")
    
    for result in results['results']:
        if result.get('negatives_triggered'):
            print(f"  ⚠️  Contenido problemático detectado en {result['image'][:30]}...")
            for neg in result['negatives_triggered']:
                print(f"     - {neg['prompt']}: {neg['score_pct']}%")
    
    return results


def custom_negative_prompts_example():
    """Ejemplo con prompts negativos personalizados."""
    print("\n" + "=" * 60)
    print("EJEMPLO CON PROMPTS NEGATIVOS PERSONALIZADOS")
    print("=" * 60)
    
    filter_instance = ImageFilterCLIP()
    
    # Prompts negativos específicos para un contexto (ej: sitio web de cocina)
    custom_negatives = [
        "raw meat", "blood", "knives being used violently",
        "dirty kitchen", "spoiled food", "insects",
        "fire hazard", "dangerous cooking"
    ]
    
    image_urls = [
        "https://picsum.photos/400/300?random=31",
        "https://picsum.photos/400/300?random=32"
    ]
    
    results = filter_instance.filter_images(
        description="delicious cooked meal on a plate",
        image_urls=image_urls,
        negative_prompts=custom_negatives
    )
    
    print("Prompts negativos personalizados para contexto culinario:")
    for prompt in custom_negatives:
        print(f"  - {prompt}")
    
    return results


def configuration_update_example():
    """Ejemplo de actualización de configuración en tiempo real."""
    print("\n" + "=" * 60)
    print("EJEMPLO DE ACTUALIZACIÓN DE CONFIGURACIÓN")
    print("=" * 60)
    
    filter_instance = ImageFilterCLIP()
    
    # Mostrar configuración inicial
    print("Configuración inicial:")
    config = filter_instance.get_config()
    print(f"  Umbral positivo: {config['positive_threshold']}")
    print(f"  Batch size: {config['batch_size']}")
    
    # Actualizar configuración
    filter_instance.update_config(
        positive_threshold=0.40,
        batch_size=16,
        max_workers=6
    )
    
    print("\nConfiguración actualizada:")
    config = filter_instance.get_config()
    print(f"  Umbral positivo: {config['positive_threshold']}")
    print(f"  Batch size: {config['batch_size']}")
    print(f"  Max workers: {config['max_workers']}")
    
    # Limpiar cache si es necesario
    filter_instance.clear_cache()
    print("\nCache de embeddings limpiado")


def error_handling_example():
    """Ejemplo de manejo de errores con URLs inválidas."""
    print("\n" + "=" * 60)
    print("EJEMPLO DE MANEJO DE ERRORES")
    print("=" * 60)
    
    filter_instance = ImageFilterCLIP()
    
    # URLs con algunos errores intencionados
    image_urls = [
        "https://picsum.photos/400/300?random=41",  # Válida
        "https://invalid-url-that-does-not-exist.com/image.jpg",  # Error 404
        "https://httpbin.org/status/500",  # Error 500
        "https://picsum.photos/400/300?random=42",  # Válida
        "not-a-url-at-all",  # URL inválida
    ]
    
    results = filter_instance.filter_images(
        description="test image",
        image_urls=image_urls
    )
    
    print("Resultados con manejo de errores:")
    for result in results['results']:
        if 'error' in result:
            print(f"  ❌ {result['image'][:40]}... -> ERROR: {result['error']}")
        else:
            print(f"  ✅ {result['image'][:40]}... -> {result['state']}")
    
    print(f"\nEstadísticas finales: {results['statistics']}")


def benchmark_example():
    """Ejemplo de benchmark de rendimiento."""
    print("\n" + "=" * 60)
    print("EJEMPLO DE BENCHMARK")
    print("=" * 60)
    
    import time
    
    configurations = [
        {"name": "Default", "config": {}},
        {"name": "High Performance", "config": {"batch_size": 24, "max_workers": 16}},
        {"name": "Conservative", "config": {"batch_size": 4, "max_workers": 4}}
    ]
    
    test_urls = [f"https://picsum.photos/300/200?random={i}" for i in range(50, 55)]
    
    for conf in configurations:
        print(f"\nTesting {conf['name']} configuration:")
        
        filter_instance = ImageFilterCLIP(conf['config'])
        
        start_time = time.time()
        results = filter_instance.filter_images(
            description="random test image",
            image_urls=test_urls
        )
        end_time = time.time()
        
        print(f"  Total time: {end_time - start_time:.2f}s")
        print(f"  Download time: {results['download_time_seconds']:.2f}s")
        print(f"  Processing time: {results['processing_time_seconds']:.2f}s")
        print(f"  Avg per image: {results['processing_time_seconds']/len(test_urls):.2f}s")
        print(f"  Success rate: {(results['statistics']['total'] - results['statistics']['errors'])/results['statistics']['total']*100:.1f}%")


def main():
    """Ejecuta todos los ejemplos."""
    print("🖼️  EJEMPLOS DE USO - ImageFilterCLIP")
    print("Clase optimizada para filtrado de imágenes con CLIP")
    
    try:
        # Ejecutar ejemplos
        basic_usage_example()
        high_performance_example()
        strict_filtering_example()
        custom_negative_prompts_example()
        configuration_update_example()
        error_handling_example()
        
        # Uncomment to run benchmark (takes longer)
        # benchmark_example()
        
        print("\n" + "=" * 60)
        print("✅ TODOS LOS EJEMPLOS COMPLETADOS")
        print("=" * 60)
        
    except Exception as e:
        print(f"\n❌ Error en los ejemplos: {str(e)}")
        import traceback
        traceback.print_exc()


if __name__ == "__main__":
    main()