The function below, pipes the input followed by the generator, followed by the discriminator to form the DCGAN architecture.

def dcgan(discriminator, generator, input_shape):
    # Set discriminator as non trainable before compiling GAN
    discriminator.trainable = False

    # Accepts the noised input
    gan_input = Input(shape=input_shape)
    
    # Generates image by passing the above received input to the generator
    gen_img = generator(gan_input)
    
    # Feeds the generated image to the discriminator
    gan_output = discriminator(gen_img)
    
    # Compile everything as a model with binary crossentropy loss
    gan = Model(inputs=gan_input, outputs=gan_output)
    return gan

If you have not seen how to use the Model function API before, please visit the detailed documentation by Keras on using the Model function API and compiling it here https://keras.io/models/model/