app.py

import numpy as np
import pandas as pd
import statsmodels.formula.api as smf
import statsmodels.api as sm
import plotly.graph_objects as go
from scipy.optimize import minimize
import plotly.express as px
from scipy.stats import t, f
import gradio as gr
import io
import zipfile
import tempfile
from datetime import datetime
import docx
from docx.shared import Inches, Pt
from docx.enum.text import WD_PARAGRAPH_ALIGNMENT
import os

# --- Data definition in global scope ---
data_dict = {
    'Tratamiento': ['T1', 'T2', 'T3', 'T4', 'T5', 'T6', 'T7', 'T8', 'T9', 'T10', 'T11', 'T12', 'T13', 'T14', 'T15'] * 3,
    'Tiempo_fermentacion_h': [16] * 15 + [23] * 15 + [40] * 15,
    'pH': [6.02, 5.39, 6.27, 4.82, 6.25, 4.87, 4.76, 4.68, 4.64, 6.35, 4.67, 6.43, 4.58, 4.60, 6.96,
           5.17, 5.95, 6.90, 5.50, 5.08, 4.95, 5.41, 5.52, 4.98, 7.10, 5.36, 6.91, 5.21, 4.66, 7.10,
           5.42, 5.60, 7.36, 5.36, 4.66, 4.93, 5.18, 5.26, 4.92, 7.28, 5.26, 6.84, 5.19, 4.58, 7.07],
    'Abs_600nm': [1.576, 1.474, 1.293, 1.446, 1.537, 1.415, 1.481, 1.419, 1.321, 1.224, 1.459, 0.345, 1.279, 1.181, 0.662,
                  1.760, 1.690, 1.485, 1.658, 1.728, 1.594, 1.673, 1.607, 1.531, 1.424, 1.595, 0.344, 1.477, 1.257, 0.660,
                  1.932, 1.780, 1.689, 1.876, 1.885, 1.824, 1.913, 1.810, 1.852, 1.694, 1.831, 0.347, 1.752, 1.367, 0.656],
    'Glucosa_g_L': [5,10,0,5,10,5,10,5,10,0,5,0,5,5,0] * 3,
    'Proteina_Pescado_g_L': [1.4,1.4,3.2,3.2,3.2,3.2,3.2,5,5,5,5,0,5,5,0] * 3,
    'Sulfato_Manganeso_g_L': [0.75,0.5,0.75,0.5,0.75,0.5,0.75,0.5,0.25,0.75,0.5,0.25,0.5,0.25,0.5] * 3
}
data = pd.DataFrame(data_dict)
# --- End of data definition in global scope ---


# --- Clase RSM_BoxBehnken ---
class RSM_BoxBehnken:
    def __init__(self, data, x1_name, x2_name, x3_name, y_name, x1_levels, x2_levels, x3_levels):
        """
        Inicializa la clase con los datos del diseño Box-Behnken.
        """
        self.data = data.copy()
        self.model = None
        self.model_simplified = None
        self.model_personalized = None # For personalized model
        self.optimized_results = None
        self.optimal_levels = None
        self.all_figures_full = [] # Separate lists for different model plots
        self.all_figures_simplified = []
        self.all_figures_personalized = []
        self.x1_name = x1_name
        self.x2_name = x2_name
        self.x3_name = x3_name
        self.y_name = y_name

        # Niveles originales de las variables
        self.x1_levels = x1_levels
        self.x2_levels = x2_levels
        self.x3_levels = x3_levels

    def get_levels(self, variable_name):
        """
        Obtiene los niveles para una variable específica.
        """
        if variable_name == self.x1_name:
            return self.x1_levels
        elif variable_name == self.x2_name:
            return self.x2_levels
        elif variable_name == self.x3_name:
            return self.x3_levels
        else:
            raise ValueError(f"Variable desconocida: {variable_name}")

    def fit_model(self):
        """
        Ajusta el modelo de segundo orden completo a los datos.
        """
        formula = f'{self.y_name} ~ {self.x1_name} + {self.x2_name} + {self.x3_name} + ' \
                  f'I({self.x1_name}**2) + I({self.x2_name}**2) + I({self.x3_name}**2) + ' \
                  f'{self.x1_name}:{self.x2_name} + {self.x1_name}:{self.x3_name} + {self.x2_name}:{self.x3_name}'
        self.model = smf.ols(formula, data=self.data).fit()
        print("Modelo Completo:")
        print(self.model.summary())
        return self.model, self.pareto_chart(self.model, "Pareto - Modelo Completo")

    def fit_simplified_model(self):
        """
        Ajusta el modelo de segundo orden a los datos, eliminando términos no significativos.
        """
        formula = f'{self.y_name} ~ {self.x1_name} + {self.x2_name} + ' \
                  f'I({self.x1_name}**2) + I({self.x2_name}**2) + I({self.x3_name}**2)' # Adjusted formula to include x3^2
        self.model_simplified = smf.ols(formula, data=self.data).fit()
        print("\nModelo Simplificado:")
        print(self.model_simplified.summary())
        return self.model_simplified, self.pareto_chart(self.model_simplified, "Pareto - Modelo Simplificado")

    def optimize(self, method='Nelder-Mead'):
        """
        Encuentra los niveles óptimos de los factores para maximizar la respuesta usando el modelo simplificado.
        """
        if self.model_simplified is None:
            print("Error: Ajusta el modelo simplificado primero.")
            return

        def objective_function(x):
            return -self.model_simplified.predict(pd.DataFrame({
                self.x1_name: [x[0]],
                self.x2_name: [x[1]],
                self.x3_name: [x[2]]
            })).values[0]

        bounds = [(-1, 1), (-1, 1), (-1, 1)]
        x0 = [0, 0, 0]

        self.optimized_results = minimize(objective_function, x0, method=method, bounds=bounds)
        self.optimal_levels = self.optimized_results.x

        # Convertir niveles óptimos de codificados a naturales
        optimal_levels_natural = [
            self.coded_to_natural(self.optimal_levels[0], self.x1_name),
            self.coded_to_natural(self.optimal_levels[1], self.x2_name),
            self.coded_to_natural(self.optimal_levels[2], self.x3_name)
        ]
        # Crear la tabla de optimización
        optimization_table = pd.DataFrame({
            'Variable': [self.x1_name, self.x2_name, self.x3_name],
            'Nivel Óptimo (Natural)': optimal_levels_natural,
            'Nivel Óptimo (Codificado)': self.optimal_levels
        })

        return optimization_table.round(3)  # Redondear a 3 decimales

    def fit_personalized_model(self, formula):
        """
        Ajusta un modelo personalizado de segundo orden a los datos, usando la formula dada.
        """
        self.model_personalized = smf.ols(formula, data=self.data).fit()
        print("\nModelo Personalizado:")
        print(self.model_personalized.summary())
        return self.model_personalized, self.pareto_chart(self.model_personalized, "Pareto - Modelo Personalizado")

    def generate_all_plots(self):
        """
        Genera todas las gráficas de RSM para todos los modelos.
        """
        if self.model_simplified is None:
            print("Error: Ajusta el modelo simplificado primero.")
            return

        self.all_figures_full = [] # Reset lists for each model type
        self.all_figures_simplified = []
        self.all_figures_personalized = []

        levels_to_plot_natural = { # Levels from data, as before
            self.x1_name: sorted(list(set(self.data[self.x1_name]))),
            self.x2_name: sorted(list(set(self.data[self.x2_name]))),
            self.x3_name: sorted(list(set(self.data[self.x3_name])))
        }

        for fixed_variable in [self.x1_name, self.x2_name, self.x3_name]:
            for level in levels_to_plot_natural[fixed_variable]:
                fig_full = self.plot_rsm_individual(fixed_variable, level, model_type='full') # Pass model_type
                if fig_full is not None:
                    self.all_figures_full.append(fig_full)
                fig_simplified = self.plot_rsm_individual(fixed_variable, level, model_type='simplified') # Pass model_type
                if fig_simplified is not None:
                    self.all_figures_simplified.append(fig_simplified)
                if self.model_personalized is not None: # Generate personalized plots only if model exists
                    fig_personalized = self.plot_rsm_individual(fixed_variable, level, model_type='personalized') # Pass model_type
                    if fig_personalized is not None:
                        self.all_figures_personalized.append(fig_personalized)

    def plot_rsm_individual(self, fixed_variable, fixed_level, model_type='simplified'): # Added model_type parameter
        """
        Genera un gráfico de superficie de respuesta (RSM) individual para una configuración específica y modelo.
        """
        model_to_use = self.model_simplified # Default to simplified model
        model_title_suffix = "(Modelo Simplificado)"
        if model_type == 'full':
            model_to_use = self.model
            model_title_suffix = "(Modelo Completo)"
        elif model_type == 'personalized':
            if self.model_personalized is None:
                print("Error: Modelo personalizado no ajustado.")
                return None
            model_to_use = self.model_personalized
            model_title_suffix = "(Modelo Personalizado)"

        if model_to_use is None: # Use model_to_use instead of self.model_simplified
            print(f"Error: Ajusta el modelo {model_type} primero.") # More informative error message
            return None

        # Determinar las variables que varían y sus niveles naturales
        varying_variables = [var for var in [self.x1_name, self.x2_name, self.x3_name] if var != fixed_variable]

        # Establecer los niveles naturales para las variables que varían
        x_natural_levels = self.get_levels(varying_variables[0])
        y_natural_levels = self.get_levels(varying_variables[1])

        # Crear una malla de puntos para las variables que varían (en unidades naturales)
        x_range_natural = np.linspace(x_natural_levels[0], x_natural_levels[-1], 100)
        y_range_natural = np.linspace(y_natural_levels[0], y_natural_levels[-1], 100)
        x_grid_natural, y_grid_natural = np.meshgrid(x_range_natural, y_range_natural)

        # Convertir la malla de variables naturales a codificadas
        x_grid_coded = self.natural_to_coded(x_grid_natural, varying_variables[0])
        y_grid_coded = self.natural_to_coded(y_range_natural, varying_variables[1])

        # Crear un DataFrame para la predicción con variables codificadas
        prediction_data = pd.DataFrame({
            varying_variables[0]: x_grid_coded.flatten(),
            varying_variables[1]: y_grid_coded.flatten(),
        })
        prediction_data[fixed_variable] = self.natural_to_coded(fixed_level, fixed_variable)

        # Fijar la variable fija en el DataFrame de predicción
        fixed_var_levels = self.get_levels(fixed_variable)
        if len(fixed_var_levels) == 3:  # Box-Behnken design levels
            prediction_data[fixed_variable] = self.natural_to_coded(fixed_level, fixed_variable)
        elif len(fixed_var_levels) > 0: # Use the closest level if not Box-Behnken
            closest_level_coded = self.natural_to_coded(min(fixed_var_levels, key=lambda x:abs(x-fixed_level)), fixed_variable)
            prediction_data[fixed_variable] = closest_level_coded


        # Calcular los valores predichos
        z_pred = model_to_use.predict(prediction_data).values.reshape(x_grid_coded.shape) # Use model_to_use here

        # Filtrar por el nivel de la variable fija (en codificado)
        fixed_level_coded = self.natural_to_coded(fixed_level, fixed_variable)
        subset_data = self.data[np.isclose(self.data[fixed_variable], fixed_level_coded)]

        # Filtrar por niveles válidos en las variables que varían
        valid_levels = [-1, 0, 1]
        experiments_data = subset_data[
            subset_data[varying_variables[0]].isin(valid_levels) &
            subset_data[varying_variables[1]].isin(valid_levels)
        ]

        # Convertir coordenadas de experimentos a naturales
        experiments_x_natural = experiments_data[varying_variables[0]].apply(lambda x: self.coded_to_natural(x, varying_variables[0]))
        experiments_y_natural = experiments_data[varying_variables[1]].apply(lambda x: self.coded_to_natural(x, varying_variables[1]))

        # Crear el gráfico de superficie con variables naturales en los ejes y transparencia
        fig = go.Figure(data=[go.Surface(z=z_pred, x=x_grid_natural, y=y_grid_natural, colorscale='Viridis', opacity=0.7, showscale=True)])

        # --- Añadir cuadrícula a la superficie ---
        # Líneas en la dirección x
        for i in range(x_grid_natural.shape[0]):
            fig.add_trace(go.Scatter3d(
                x=x_grid_natural[i, :],
                y=y_grid_natural[i, :],
                z=z_pred[i, :],
                mode='lines',
                line=dict(color='gray', width=2),
                showlegend=False,
                hoverinfo='skip'
            ))
        # Líneas en la dirección y
        for j in range(x_grid_natural.shape[1]):
            fig.add_trace(go.Scatter3d(
                x=x_grid_natural[:, j],
                y=y_grid_natural[:, j],
                z=z_pred[:, j],
                mode='lines',
                line=dict(color='gray', width=2),
                showlegend=False,
                hoverinfo='skip'
            ))

        # --- Fin de la adición de la cuadrícula ---

        # Añadir los puntos de los experimentos en la superficie de respuesta con diferentes colores y etiquetas
        colors = px.colors.qualitative.Safe
        point_labels = [f"{row[self.y_name]:.3f}" for _, row in experiments_data.iterrows()]

        fig.add_trace(go.Scatter3d(
            x=experiments_x_natural,
            y=experiments_y_natural,
            z=experiments_data[self.y_name].round(3),
            mode='markers+text',
            marker=dict(size=4, color=colors[:len(experiments_x_natural)]),
            text=point_labels,
            textposition='top center',
            name='Experimentos'
        ))

        # Añadir etiquetas y título con variables naturales
        fig.update_layout(
            scene=dict(
                xaxis_title=f"{varying_variables[0]} ({self.get_units(varying_variables[0])})",
                yaxis_title=f"{varying_variables[1]} ({self.get_units(varying_variables[1])})",
                zaxis_title=self.y_name,
            ),
            title=f"{self.y_name} vs {varying_variables[0]} y {varying_variables[1]}<br><sup>{fixed_variable} fijo en {fixed_level:.3f} ({self.get_units(fixed_variable)}) {model_title_suffix}</sup>", # Updated title
            height=800,
            width=1000,
            showlegend=True
        )
        return fig


# --- Funciones para la Interfaz de Gradio ---

model_completo_output = gr.HTML()
pareto_completo_output = gr.Plot()
model_simplificado_output = gr.HTML()
pareto_simplificado_output = gr.Plot()
equation_output = gr.HTML()
optimization_table_output = gr.Dataframe(label="Tabla de Optimización", interactive=False)
prediction_table_output = gr.Dataframe(label="Tabla de Predicciones", interactive=False)
contribution_table_output = gr.Dataframe(label="Tabla de % de Contribución", interactive=False)
anova_table_output = gr.Dataframe(label="Tabla ANOVA Detallada", interactive=False)
download_all_plots_button = gr.DownloadButton("Descargar Todos los Gráficos (ZIP)")
download_excel_button = gr.DownloadButton("Descargar Tablas en Excel")
rsm_plot_output = gr.Plot()
plot_info = gr.Textbox(label="Información del Gráfico", value="Gráfico 1 de 9", interactive=False)
current_index_state = gr.State(0)
all_figures_state = gr.State([])
current_model_type_state = gr.State('simplified')
model_personalized_output = gr.HTML() # Output for personalized model summary
pareto_personalized_output = gr.Plot() # Pareto for personalized model
factor_checkboxes = gr.CheckboxGroup(["factors", "x1_sq", "x2_sq", "x3_sq"], label="Términos de Factores", value=["factors", "x1_sq", "x2_sq", "x3_sq"]) # Factor Checkboxes
interaction_checkboxes = gr.CheckboxGroup(["x1x2", "x1x3", "x2x3"], label="Términos de Interacción") # Interaction Checkboxes


def create_gradio_interface():
    global model_completo_output, pareto_completo_output, model_simplificado_output, pareto_simplificado_output, equation_output, optimization_table_output, prediction_table_output, contribution_table_output, anova_table_output, download_all_plots_button, download_excel_button, rsm_plot_output, plot_info, current_index_state, all_figures_state, current_model_type_state, model_personalized_output, pareto_personalized_output, factor_checkboxes, interaction_checkboxes

    with gr.Blocks() as demo:
        gr.Markdown("# Optimización de la Absorbancia usando RSM")

        with gr.Row():
            with gr.Column():
                gr.Markdown("## Configuración del Análisis")
                data_input = gr.Textbox(label="Datos del Experimento (formato CSV - Ignored, Data is Hardcoded)", lines=5, interactive=False, value="Data is pre-loaded, ignore input.")
                load_button = gr.Button("Cargar Datos")
                data_dropdown = gr.Dropdown(["All Data"], value="All Data", label="Seleccionar Datos") # Data Selection Dropdown - currently only 'All Data'

            with gr.Column():
                gr.Markdown("## Datos Cargados")
                data_output = gr.Dataframe(label="Tabla de Datos", interactive=False)

        with gr.Row(visible=False) as analysis_row:
            with gr.Column():
                fit_button = gr.Button("Ajustar Modelo Simplificado y Completo") # Button label changed
                gr.Markdown("**Modelo Completo**")
                model_completo_output = model_completo_output # gr.HTML() # output_components are now global
                pareto_completo_output = pareto_completo_output # gr.Plot()
                gr.Markdown("**Modelo Simplificado**")
                model_simplificado_output = model_simplificado_output # gr.HTML()
                pareto_simplificado_output = pareto_simplificado_output # gr.Plot()

                gr.Markdown("## Modelo Personalizado") # Personalized Model Section
                factor_checkboxes_comp = factor_checkboxes # gr.CheckboxGroup(["factors", "x1_sq", "x2_sq", "x3_sq"], label="Términos de Factores", value=["factors", "x1_sq", "x2_sq", "x3_sq"]) # Factor Checkboxes
                interaction_checkboxes_comp = interaction_checkboxes # gr.CheckboxGroup(["x1x2", "x1x3", "x2x3"], label="Términos de Interacción") # Interaction Checkboxes
                custom_model_button = gr.Button("Ajustar Modelo Personalizado") # Fit Custom Model Button
                model_personalized_output_comp = model_personalized_output # gr.HTML() # Output for personalized model summary
                pareto_personalized_output_comp = pareto_personalized_output # gr.Plot() # Pareto for personalized model

                gr.Markdown("**Ecuación del Modelo Simplificado**")
                equation_output = equation_output # gr.HTML()
                optimization_table_output = optimization_table_output # gr.Dataframe(label="Tabla de Optimización", interactive=False)
                prediction_table_output = prediction_table_output # gr.Dataframe(label="Tabla de Predicciones", interactive=False)
                contribution_table_output = contribution_table_output # gr.Dataframe(label="Tabla de % de Contribución", interactive=False)
                anova_table_output = anova_table_output # gr.Dataframe(label="Tabla ANOVA Detallada", interactive=False)

                gr.Markdown("## Descargar Todas las Tablas")
                download_excel_button_comp = download_excel_button # gr.DownloadButton("Descargar Tablas en Excel")
                download_word_button = gr.DownloadButton("Descargar Tablas en Word")

            with gr.Column():
                gr.Markdown("## Gráficos de Superficie de Respuesta")
                model_type_radio = gr.Radio(["simplified", "full", "personalized"], value="simplified", label="Tipo de Modelo para Gráficos") # Model Type Radio
                fixed_variable_input = gr.Dropdown(label="Variable Fija", choices=["Glucosa_g_L", "Proteina_Pescado_g_L", "Sulfato_Manganeso_g_L"], value="Glucosa_g_L")
                fixed_level_input = gr.Slider(label="Nivel de Variable Fija (Natural Units)", minimum=min(data['Glucosa_g_L']), maximum=max(data['Glucosa_g_L']), step=0.1, value=5.0)
                plot_button = gr.Button("Generar Gráficos")
                with gr.Row():
                    left_button = gr.Button("<")
                    right_button = gr.Button(">")
                rsm_plot_output = rsm_plot_output # gr.Plot()
                plot_info = plot_info # gr.Textbox(label="Información del Gráfico", value="Gráfico 1 de 9", interactive=False)
                with gr.Row():
                    download_plot_button_comp = download_plot_button # gr.DownloadButton("Descargar Gráfico Actual (PNG)")
                    download_all_plots_button_comp = download_all_plots_button # gr.DownloadButton("Descargar Todos los Gráficos (ZIP)")
                current_index_state = current_index_state # gr.State(0)
                all_figures_state = all_figures_state # gr.State([])
                current_model_type_state = current_model_type_state # gr.State('simplified')


        # Cargar datos
        load_button.click(
            load_data,
            inputs=[data_input],
            outputs=[data_output, analysis_row]
        )

        # Ajustar modelo y optimizar (Simplified and Full)
        fit_button.click(
            fit_and_optimize_model,
            inputs=[],
            outputs=[
                model_completo_output,
                pareto_completo_output,
                model_simplificado_output,
                pareto_simplificado_output,
                equation_output,
                optimization_table_output,
                prediction_table_output,
                contribution_table_output,
                anova_table_output,
                download_all_plots_button,
                download_excel_button
            ]
        )

        # Ajustar modelo personalizado
        custom_model_button.click( # New event for custom model fitting
            fit_custom_model,
            inputs=[factor_checkboxes_comp, interaction_checkboxes_comp],
            outputs=[model_personalized_output_comp, pareto_personalized_output_comp] # pass output components
        )


        # Generar y mostrar los gráficos
        plot_button.click(
            lambda fixed_var, fixed_lvl, model_type: show_plot(0, [], model_type) if not hasattr(rsm, 'all_figures_full') or not rsm.all_figures_full else show_plot(0, [], model_type) if model_type == 'full' and not rsm.all_figures_full else show_plot(0, [], model_type) if model_type == 'simplified' and not rsm.all_figures_simplified else show_plot(0, [], model_type) if model_type == 'personalized' and not rsm.all_figures_personalized else show_plot(0, rsm.all_figures_full if model_type == 'full' else rsm.all_figures_simplified if model_type == 'simplified' else rsm.all_figures_personalized, model_type),
            inputs=[fixed_variable_input, fixed_level_input, model_type_radio],
            outputs=[rsm_plot_output, plot_info, current_index_state, current_model_type_state]
        )


        # Navegación de gráficos
        left_button.click(
            lambda current_index, all_figures, model_type: navigate_plot('left', current_index, all_figures, model_type),
            inputs=[current_index_state, all_figures_state, current_model_type_state],
            outputs=[rsm_plot_output, plot_info, current_index_state]
        )
        right_button.click(
            lambda current_index, all_figures, model_type: navigate_plot('right', current_index, all_figures, model_type),
            inputs=[current_index_state, all_figures_state, current_model_type_state],
            outputs=[rsm_plot_output, plot_info, current_index_state]
        )

        # Descargar gráfico actual
        download_plot_button.click(
            download_current_plot,
            inputs=[all_figures_state, current_index_state, current_model_type_state],
            outputs=download_plot_button
        )

        # Descargar todos los gráficos en ZIP
        download_all_plots_button.click(
            lambda model_type: download_all_plots_zip(model_type),
            inputs=[current_model_type_state],
            outputs=download_all_plots_button
        )

        # Descargar todas las tablas en Excel y Word
        download_excel_button.click(
            fn=lambda: download_all_tables_excel(),
            inputs=[],
            outputs=download_excel_button
        )

        download_word_button.click(
            fn=lambda: exportar_word(rsm, rsm.get_all_tables()),
            inputs=[],
            outputs=download_word_button
        )


        # Ejemplo de uso
        gr.Markdown("## Instrucciones:")
        gr.Markdown("""
        1. Click 'Cargar Datos' para usar los datos precargados.
        2. Click 'Ajustar Modelo Simplificado y Completo'.
        3. Opcional: Define un Modelo Personalizado seleccionando términos y haz clic en 'Ajustar Modelo Personalizado'.
        4. Selecciona el 'Tipo de Modelo para Gráficos' (Simplificado, Completo o Personalizado).
        5. Select 'Variable Fija' and 'Nivel de Variable Fija'.
        6. Click 'Generar Gráficos'.
        7. Navega entre los gráficos usando los botones '<' y '>'.
        8. Descarga el gráfico actual en PNG o descarga todos los gráficos en un ZIP.
        9. Descarga todas las tablas en un archivo Excel o Word con los botones correspondientes.
        """)

    return demo

# --- Función Principal ---

def main():
    interface = create_gradio_interface()
    interface.launch(share=True)

if __name__ == "__main__":
    main()