app.py

import gradio as gr
# import anthropic # Removed Anthropic import
import PyPDF2
import pandas as pd
import numpy as np
import io
import os
import json
import zipfile
import tempfile
from typing import Dict, List, Tuple, Union, Optional
import re
from pathlib import Path
import openpyxl # Needed for reading .xlsx
from dataclasses import dataclass
from enum import Enum
# No need for docx, reportlab, matplotlib if only text/code output is used and not generating them internally
# import docx
# from docx.shared import Inches, Pt, RGBColor
# from docx.enum.text import WD_ALIGN_PARAGRAPH
# import reportlab
# from reportlab.lib import colors
# from reportlab.lib.pagesizes import letter, A4
# from reportlab.platypus import SimpleDocTemplate, Table, TableStyle, Paragraph, Spacer, PageBreak
# from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
# from reportlab.lib.units import inch
# from reportlab.pdfbase import pdfmetrics
# from reportlab.pdfbase.ttfonts import TTFont
# import matplotlib.pyplot as plt # Moved to the implementation code section
from datetime import datetime

# Import OpenAI for Qwen access
from openai import OpenAI

# Configuración para HuggingFace
os.environ['GRADIO_ANALYTICS_ENABLED'] = 'False'

# Initialize OpenAI client for Qwen
# Read API key from NEBIUS_API_KEY environment variable
NEBIUS_API_KEY = os.environ.get("NEBIUS_API_KEY")
if NEBIUS_API_KEY:
    openai_client = OpenAI(
        base_url="https://api.studio.nebius.com/v1/",
        api_key=NEBIUS_API_KEY
    )
    print("OpenAI client initialized for Nebius Qwen endpoint.")
else:
    openai_client = None
    print("NEBIUS_API_KEY not found. OpenAI client not initialized.")


# Sistema de traducción - Actualizado con nuevas entradas
TRANSLATIONS = {
    'en': {
        'title': '🧬 Comparative Analyzer of Biotechnological Models',
        'subtitle': 'Specialized in comparative analysis of mathematical model fitting results',
        'upload_files': '📁 Upload fitting results (CSV/Excel)',
        'select_model': '🤖 Qwen Model', # Changed label
        'select_language': '🌐 Language',
        'select_theme': '🎨 Theme',
        'detail_level': '📋 Analysis detail level',
        'detailed': 'Detailed',
        'summarized': 'Summarized',
        'analyze_button': '🚀 Analyze and Compare Models',
        'export_format': '📄 Export format',
        'export_button': '💾 Export Report',
        'comparative_analysis': '📊 Comparative Analysis',
        'implementation_code': '💻 Implementation Code',
        'data_format': '📋 Expected data format',
        'examples': '📚 Analysis examples',
        'light': 'Light',
        'dark': 'Dark',
        'best_for': 'Best for',
        'loading': 'Loading...',
        'error_no_api': 'Please configure NEBIUS_API_KEY in HuggingFace Space secrets', # Changed message
        'error_no_files': 'Please upload fitting result files to analyze',
        'report_exported': 'Report exported successfully as',
        'specialized_in': '🎯 Specialized in:',
        'metrics_analyzed': '📊 Analyzed metrics:',
        'what_analyzes': '🔍 What it specifically analyzes:',
        'tips': '💡 Tips for better results:',
        'additional_specs': '📝 Additional specifications for analysis',
        'additional_specs_placeholder': 'Add any specific requirements or focus areas for the analysis...'
    },
    'es': {
        'title': '🧬 Analizador Comparativo de Modelos Biotecnológicos',
        'subtitle': 'Especializado en análisis comparativo de resultados de ajuste de modelos matemáticos',
        'upload_files': '📁 Subir resultados de ajuste (CSV/Excel)',
        'select_model': '🤖 Modelo Qwen', # Changed label
        'select_language': '🌐 Idioma',
        'select_theme': '🎨 Tema',
        'detail_level': '📋 Nivel de detalle del análisis',
        'detailed': 'Detallado',
        'summarized': 'Resumido',
        'analyze_button': '🚀 Analizar y Comparar Modelos',
        'export_format': '📄 Formato de exportación',
        'export_button': '💾 Exportar Reporte',
        'comparative_analysis': '📊 Análisis Comparativo',
        'implementation_code': '💻 Código de Implementación',
        'data_format': '📋 Formato de datos esperado',
        'examples': '📚 Ejemplos de análisis',
        'light': 'Claro',
        'dark': 'Oscuro',
        'best_for': 'Mejor para',
        'loading': 'Cargando...',
        'error_no_api': 'Por favor configura NEBIUS_API_KEY en los secretos del Space', # Changed message
        'error_no_files': 'Por favor sube archivos con resultados de ajuste para analizar',
        'report_exported': 'Reporte exportado exitosamente como',
        'specialized_in': '🎯 Especializado en:',
        'metrics_analyzed': '📊 Métricas analizadas:',
        'what_analyzes': '🔍 Qué analiza específicamente:',
        'tips': '💡 Tips para mejores resultados:',
        'additional_specs': '📝 Especificaciones adicionales para el análisis',
        'additional_specs_placeholder': 'Agregue cualquier requerimiento específico o áreas de enfoque para el análisis...'
    },
    'fr': {
        'title': '🧬 Analyseur Comparatif de Modèles Biotechnologiques',
        'subtitle': 'Spécialisé dans l\'analyse comparative des résultats d\'ajustement',
        'upload_files': '📁 Télécharger les résultats (CSV/Excel)',
        'select_model': '🤖 Modèle Qwen', # Changed label
        'select_language': '🌐 Langue',
        'select_theme': '🎨 Thème',
        'detail_level': '📋 Niveau de détail',
        'detailed': 'Détaillé',
        'summarized': 'Résumé',
        'analyze_button': '🚀 Analyser et Comparer',
        'export_format': '📄 Format d\'export',
        'export_button': '💾 Exporter le Rapport',
        'comparative_analysis': '📊 Analyse Comparative',
        'implementation_code': '💻 Code d\'Implémentation',
        'data_format': '📋 Format de données attendu',
        'examples': '📚 Exemples d\'analyse',
        'light': 'Clair',
        'dark': 'Sombre',
        'best_for': 'Meilleur pour',
        'loading': 'Chargement...',
        'error_no_api': 'Veuillez configurer NEBIUS_API_KEY', # Changed message
        'error_no_files': 'Veuillez télécharger des fichiers à analyser',
        'report_exported': 'Rapport exporté avec succès comme',
        'specialized_in': '🎯 Spécialisé dans:',
        'metrics_analyzed': '📊 Métriques analysées:',
        'what_analyzes': '🔍 Ce qu\'il analyse spécifiquement:',
        'tips': '💡 Conseils pour de meilleurs résultats:',
        'additional_specs': '📝 Spécifications supplémentaires pour l\'analyse',
        'additional_specs_placeholder': 'Ajoutez des exigences spécifiques ou des domaines d\'intérêt pour l\'analyse...'
    },
    'de': {
        'title': '🧬 Vergleichender Analysator für Biotechnologische Modelle',
        'subtitle': 'Spezialisiert auf vergleichende Analyse von Modellanpassungsergebnissen',
        'upload_files': '📁 Ergebnisse hochladen (CSV/Excel)',
        'select_model': '🤖 Qwen Modell', # Changed label
        'select_language': '🌐 Sprache',
        'select_theme': '🎨 Thema',
        'detail_level': '📋 Detailgrad der Analyse',
        'detailed': 'Detailliert',
        'summarized': 'Zusammengefasst',
        'analyze_button': '🚀 Analysieren und Vergleichen',
        'export_format': '📄 Exportformat',
        'export_button': '💾 Bericht Exportieren',
        'comparative_analysis': '📊 Vergleichende Analyse',
        'implementation_code': '💻 Implementierungscode',
        'data_format': '📋 Erwartetes Datenformat',
        'examples': '📚 Analysebeispiele',
        'light': 'Hell',
        'dark': 'Dunkel',
        'best_for': 'Am besten für',
        'loading': 'Laden...',
        'error_no_api': 'Bitte konfigurieren Sie NEBIUS_API_KEY', # Changed message
        'error_no_files': 'Bitte laden Sie Dateien zur Analyse hoch',
        'report_exported': 'Bericht erfolgreich exportiert als',
        'specialized_in': '🎯 Spezialisiert auf:',
        'metrics_analyzed': '📊 Analysierte Metriken:',
        'what_analyzes': '🔍 Was spezifisch analysiert wird:',
        'tips': '💡 Tipps für bessere Ergebnisse:',
        'additional_specs': '📝 Zusätzliche Spezifikationen für die Analyse',
        'additional_specs_placeholder': 'Fügen Sie spezifische Anforderungen oder Schwerpunktbereiche für die Analyse hinzu...'
    },
    'pt': {
        'title': '🧬 Analisador Comparativo de Modelos Biotecnológicos',
        'subtitle': 'Especializado em análise comparativa de resultados de ajuste',
        'upload_files': '📁 Carregar resultados (CSV/Excel)',
        'select_model': '🤖 Modelo Qwen', # Changed label
        'select_language': '🌐 Idioma',
        'select_theme': '🎨 Tema',
        'detail_level': '📋 Nível de detalhe',
        'detailed': 'Detalhado',
        'summarized': 'Resumido',
        'analyze_button': '🚀 Analisar e Comparar',
        'export_format': '📄 Formato de exportação',
        'export_button': '💾 Exportar Relatório',
        'comparative_analysis': '📊 Análise Comparativa',
        'implementation_code': '💻 Código de Implementação',
        'data_format': '📋 Formato de dados esperado',
        'examples': '📚 Exemplos de análise',
        'light': 'Claro',
        'dark': 'Escuro',
        'best_for': 'Melhor para',
        'loading': 'Carregando...',
        'error_no_api': 'Por favor configure NEBIUS_API_KEY', # Changed message
        'error_no_files': 'Por favor carregue arquivos para analisar',
        'report_exported': 'Relatório exportado com sucesso como',
        'specialized_in': '🎯 Especializado em:',
        'metrics_analyzed': '📊 Métricas analisadas:',
        'what_analyzes': '🔍 O que analisa especificamente:',
        'tips': '💡 Dicas para melhores resultados:',
        'additional_specs': '📝 Especificações adicionais para a análise',
        'additional_specs_placeholder': 'Adicione requisitos específicos ou áreas de foco para a análise...'
    }
}

# Temas disponibles
THEMES = {
    'light': gr.themes.Soft(),
    'dark': gr.themes.Base(
        primary_hue="blue",
        secondary_hue="gray",
        neutral_hue="gray",
        font=["Arial", "sans-serif"]
    ).set(
        body_background_fill="dark",
        body_background_fill_dark="*neutral_950",
        button_primary_background_fill="*primary_600",
        button_primary_background_fill_hover="*primary_500",
        button_primary_text_color="white",
        block_background_fill="*neutral_800",
        block_border_color="*neutral_700",
        block_label_text_color="*neutral_200",
        block_title_text_color="*neutral_100",
        checkbox_background_color="*neutral_700",
        checkbox_background_color_selected="*primary_600",
        input_background_fill="*neutral_700",
        input_border_color="*neutral_600",
        input_placeholder_color="*neutral_400"
    )
}

# Enum para tipos de análisis
class AnalysisType(Enum):
    MATHEMATICAL_MODEL = "mathematical_model"
    DATA_FITTING = "data_fitting"
    FITTING_RESULTS = "fitting_results"
    UNKNOWN = "unknown"

# Estructura modular para modelos
@dataclass
class MathematicalModel:
    name: str
    equation: str
    parameters: List[str]
    application: str
    sources: List[str]
    category: str
    biological_meaning: str

# Sistema de registro de modelos escalable
class ModelRegistry:
    def __init__(self):
        self.models = {}
        self._initialize_default_models()

    def register_model(self, model: MathematicalModel):
        """Registra un nuevo modelo matemático"""
        if model.category not in self.models:
            self.models[model.category] = {}
        self.models[model.category][model.name] = model

    def get_model(self, category: str, name: str) -> MathematicalModel:
        """Obtiene un modelo específico"""
        return self.models.get(category, {}).get(name)

    def get_all_models(self) -> Dict:
        """Retorna todos los modelos registrados"""
        return self.models

    def _initialize_default_models(self):
        """Inicializa los modelos por defecto"""
        # Modelos de crecimiento
        self.register_model(MathematicalModel(
            name="Monod",
            equation="μ = μmax × (S / (Ks + S))",
            parameters=["μmax (h⁻¹)", "Ks (g/L)"],
            application="Crecimiento limitado por sustrato único",
            sources=["Cambridge", "MIT", "DTU"],
            category="crecimiento_biomasa",
            biological_meaning="Describe cómo la velocidad de crecimiento depende de la concentración de sustrato limitante"
        ))

        self.register_model(MathematicalModel(
            name="Logístico",
            equation="dX/dt = μmax × X × (1 - X/Xmax)",
            parameters=["μmax (h⁻¹)", "Xmax (g/L)"],
            application="Sistemas cerrados batch",
            sources=["Cranfield", "Swansea", "HAL Theses"],
            category="crecimiento_biomasa",
            biological_meaning="Modela crecimiento limitado por capacidad de carga del sistema"
        ))

        self.register_model(MathematicalModel(
            name="Gompertz",
            equation="X(t) = Xmax × exp(-exp((μmax × e / Xmax) × (λ - t) + 1))",
            parameters=["λ (h)", "μmax (h⁻¹)", "Xmax (g/L)"],
            application="Crecimiento con fase lag pronunciada",
            sources=["Lund University", "NC State"],
            category="crecimiento_biomasa",
            biological_meaning="Incluye fase de adaptación (lag) seguida de crecimiento exponencial y estacionario"
        ))

# Instancia global del registro
model_registry = ModelRegistry()

# Available Qwen Models (Updated from Claude)
QWEN_MODELS = {
    "Qwen/Qwen3-14B": { # Using the model specified by the user
        "name": "Qwen 3-14B",
        "description": "A powerful Qwen model suitable for complex analysis.",
        "max_tokens": 8192, # Example context window, adjust based on actual model specs
        "best_for": "Detailed analysis and code generation"
    },
    # Add other Qwen models if available and desired, e.g.:
    # "Qwen/Qwen3-7B": {
    #     "name": "Qwen 3-7B",
    #     "description": "Faster Qwen model",
    #     "max_tokens": 8192,
    #     "best_for": "Quicker analysis"
    # }
}

class FileProcessor:
    """Clase para procesar diferentes tipos de archivos"""

    @staticmethod
    def extract_text_from_pdf(pdf_file) -> str:
        """Extrae texto de un archivo PDF"""
        try:
            pdf_reader = PyPDF2.PdfReader(io.BytesIO(pdf_file))
            text = ""
            for page in pdf_reader.pages:
                text += page.extract_text() + "\n"
            return text
        except Exception as e:
            return f"Error reading PDF: {str(e)}"

    @staticmethod
    def read_csv(csv_file) -> pd.DataFrame:
        """Lee archivo CSV"""
        try:
            return pd.read_csv(io.BytesIO(csv_file))
        except Exception as e:
            print(f"Error reading CSV: {e}")
            return None

    @staticmethod
    def read_excel(excel_file) -> pd.DataFrame:
        """Lee archivo Excel"""
        try:
            return pd.read_excel(io.BytesIO(excel_file))
        except Exception as e:
            print(f"Error reading Excel: {e}")
            return None

    @staticmethod
    def extract_from_zip(zip_file) -> List[Tuple[str, bytes]]:
        """Extrae archivos de un ZIP"""
        files = []
        try:
            with zipfile.ZipFile(io.BytesIO(zip_file), 'r') as zip_ref:
                for file_name in zip_ref.namelist():
                    if not file_name.startswith('__MACOSX'):
                        file_data = zip_ref.read(file_name)
                        files.append((file_name, file_data))
        except Exception as e:
            print(f"Error processing ZIP: {e}")
        return files

class ReportExporter:
    """Clase para exportar reportes a diferentes formatos"""
    # Keep ReportExporter as is, as it processes the analysis text,
    # not the AI interaction itself. It might need docx/reportlab imports
    # re-added if generating those formats. Assuming they are needed for export_to_docx/pdf.
    # Re-adding necessary imports for ReportExporter
    from docx import Document
    from docx.shared import Inches, Pt, RGBColor
    from docx.enum.text import WD_ALIGN_PARAGRAPH
    from reportlab.lib import colors
    from reportlab.lib.pagesizes import letter, A4
    from reportlab.platypus import SimpleDocTemplate, Table, TableStyle, Paragraph, Spacer, PageBreak
    from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
    from reportlab.lib.units import inch
    # pdfmetrics and TTFont might be needed for custom fonts if used, keep them for now.
    from reportlab.pdfbase import pdfmetrics
    from reportlab.pdfbase.ttfonts import TTFont

    @staticmethod
    def export_to_docx(content: str, filename: str, language: str = 'en') -> str:
        """Exporta el contenido a un archivo DOCX"""
        doc = ReportExporter.Document()

        # Configurar estilos
        title_style = doc.styles['Title']
        title_style.font.size = ReportExporter.Pt(24)
        title_style.font.bold = True

        heading_style = doc.styles['Heading 1']
        heading_style.font.size = ReportExporter.Pt(18)
        heading_style.font.bold = True

        # Título
        title_text = {
            'en': 'Comparative Analysis Report - Biotechnological Models',
            'es': 'Informe de Análisis Comparativo - Modelos Biotecnológicos',
            'fr': 'Rapport d\'Analyse Comparative - Modèles Biotechnologiques',
            'de': 'Vergleichsanalysebericht - Biotechnologische Modelle',
            'pt': 'Relatório de Análise Comparativa - Modelos Biotecnológicos'
        }

        doc.add_heading(title_text.get(language, title_text['en']), 0)

        # Fecha
        date_text = {
            'en': 'Generated on',
            'es': 'Generado el',
            'fr': 'Généré le',
            'de': 'Erstellt am',
            'pt': 'Gerado em'
        }
        doc.add_paragraph(f"{date_text.get(language, date_text['en'])}: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")
        doc.add_paragraph()

        # Procesar contenido
        lines = content.split('\n')
        current_paragraph = None

        for line in lines:
            line = line.strip()

            if line.startswith('###'):
                doc.add_heading(line.replace('###', '').strip(), level=2)
            elif line.startswith('##'):
                doc.add_heading(line.replace('##', '').strip(), level=1)
            elif line.startswith('#'):
                doc.add_heading(line.replace('#', '').strip(), level=0)
            elif line.startswith('**') and line.endswith('**'):
                # Texto en negrita
                p = doc.add_paragraph()
                run = p.add_run(line.replace('**', ''))
                run.bold = True
            elif line.startswith('- ') or line.startswith('* '):
                # Lista
                doc.add_paragraph(line[2:], style='List Bullet')
            elif line.startswith(tuple('0123456789')):
                # Lista numerada
                doc.add_paragraph(line, style='List Number')
            elif line == '---' or line.startswith('==='):
                # Separador
                doc.add_paragraph('_' * 50)
            elif line:
                # Párrafo normal
                doc.add_paragraph(line)

        # Guardar documento
        doc.save(filename)
        return filename

    @staticmethod
    def export_to_pdf(content: str, filename: str, language: str = 'en') -> str:
        """Exporta el contenido a un archivo PDF"""
        # Crear documento PDF
        doc = ReportExporter.SimpleDocTemplate(filename, pagesize=ReportExporter.letter)
        story = []
        styles = ReportExporter.getSampleStyleSheet()

        # Estilos personalizados
        title_style = ReportExporter.ParagraphStyle(
            'CustomTitle',
            parent=styles['Title'],
            fontSize=24,
            textColor=ReportExporter.colors.HexColor('#1f4788'),
            spaceAfter=30
        )

        heading_style = ReportExporter.ParagraphStyle(
            'CustomHeading',
            parent=styles['Heading1'],
            fontSize=16,
            textColor=ReportExporter.colors.HexColor('#2e5090'),
            spaceAfter=12
        )

        # Título
        title_text = {
            'en': 'Comparative Analysis Report - Biotechnological Models',
            'es': 'Informe de Análisis Comparativo - Modelos Biotecnológicos',
            'fr': 'Rapport d\'Analyse Comparative - Modèles Biotechnologiques',
            'de': 'Vergleichsanalysebericht - Biotechnologische Modelle',
            'pt': 'Relatório de Análise Comparativa - Modelos Biotecnológicos'
        }

        story.append(ReportExporter.Paragraph(title_text.get(language, title_text['en']), title_style))

        # Fecha
        date_text = {
            'en': 'Generated on',
            'es': 'Generado el',
            'fr': 'Généré le',
            'de': 'Erstellt am',
            'pt': 'Gerado em'
        }
        story.append(ReportExporter.Paragraph(f"{date_text.get(language, date_text['en'])}: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}", styles['Normal']))
        story.append(ReportExporter.Spacer(1, 0.5*ReportExporter.inch))

        # Procesar contenido
        lines = content.split('\n')

        for line in lines:
            line = line.strip()

            if not line:
                story.append(ReportExporter.Spacer(1, 0.2*ReportExporter.inch))
            elif line.startswith('###'):
                story.append(ReportExporter.Paragraph(line.replace('###', '').strip(), styles['Heading3']))
            elif line.startswith('##'):
                story.append(ReportExporter.Paragraph(line.replace('##', '').strip(), styles['Heading2']))
            elif line.startswith('#'):
                story.append(ReportExporter.Paragraph(line.replace('#', '').strip(), heading_style))
            elif line.startswith('**') and line.endswith('**'):
                text = line.replace('**', '')
                story.append(ReportExporter.Paragraph(f"<b>{text}</b>", styles['Normal']))
            elif line.startswith('- ') or line.startswith('* '):
                story.append(ReportExporter.Paragraph(f"• {line[2:]}", styles['Normal']))
            elif line == '---' or line.startswith('==='):
                story.append(ReportExporter.Spacer(1, 0.3*ReportExporter.inch))
                story.append(ReportExporter.Paragraph("_" * 70, styles['Normal']))
                story.append(ReportExporter.Spacer(1, 0.3*ReportExporter.inch))
            else:
                # Limpiar caracteres especiales para PDF
                clean_line = line.replace('📊', '[GRAPH]').replace('🎯', '[TARGET]').replace('🔍', '[SEARCH]').replace('💡', '[TIP]')
                story.append(ReportExporter.Paragraph(clean_line, styles['Normal']))

        # Construir PDF
        doc.build(story)
        return filename


class AIAnalyzer:
    """Clase para análisis con IA (usando OpenAI for Qwen)"""

    def __init__(self, client, model_registry):
        # client is now an OpenAI client instance
        self.client = client
        self.model_registry = model_registry
        # Qwen specific parameters from user example
        self.temperature = 0.6
        self.top_p = 0.95

    def detect_analysis_type(self, content: Union[str, pd.DataFrame]) -> AnalysisType:
        """Detecta el tipo de análisis necesario"""
        if isinstance(content, pd.DataFrame):
            columns = [col.lower() for col in content.columns]

            fitting_indicators = [
                'r2', 'r_squared', 'rmse', 'mse', 'aic', 'bic',
                'parameter', 'param', 'coefficient', 'fit',
                'model', 'equation', 'goodness', 'chi_square',
                'p_value', 'confidence', 'standard_error', 'se'
            ]

            has_fitting_results = any(indicator in ' '.join(columns) for indicator in fitting_indicators)

            if has_fitting_results:
                return AnalysisType.FITTING_RESULTS
            else:
                # Assuming any dataframe without clear fitting metrics is raw data
                return AnalysisType.DATA_FITTING

        # Use a quick Qwen model for type detection
        # Using the same model as the main analysis for simplicity, could use a smaller one if available
        model_for_detection = list(QWEN_MODELS.keys())[0] # Use the first available Qwen model

        prompt = """
        Analyze this content and determine if it is:
        1. A scientific article describing biotechnological mathematical models
        2. Experimental data for parameter fitting
        3. Model fitting results (with parameters, R², RMSE, etc.)

        Reply only with: "MODEL", "DATA" or "RESULTS". Be very concise.
        """

        try:
            response = self.client.chat.completions.create(
                model=model_for_detection,
                temperature=0.1, # Lower temp for deterministic output
                max_tokens=10,
                messages=[{"role": "user", "content": f"{prompt}\n\n{content[:1000]}"}]
            )

            # Extract text from OpenAI response
            result = response.choices[0].message.content.strip().upper()

            if "MODEL" in result:
                return AnalysisType.MATHEMATICAL_MODEL
            elif "RESULTS" in result:
                return AnalysisType.FITTING_RESULTS
            elif "DATA" in result:
                return AnalysisType.DATA_FITTING
            else:
                return AnalysisType.UNKNOWN

        except Exception as e:
            print(f"Error during analysis type detection: {e}")
            return AnalysisType.UNKNOWN

    def get_language_prompt_prefix(self, language: str) -> str:
        """Obtiene el prefijo del prompt según el idioma"""
        prefixes = {
            'en': "Please respond exclusively in English. ",
            'es': "Por favor responde exclusivamente en español. ",
            'fr': "Veuillez répondre exclusivement en français. ",
            'de': "Bitte antworten Sie ausschließlich auf Deutsch. ",
            'pt': "Por favor responda exclusivamente em português. "
        }
        return prefixes.get(language, prefixes['en'])

    def analyze_fitting_results(self, data: pd.DataFrame, qwen_model: str, detail_level: str = "detailed",
                              language: str = "en", additional_specs: str = "") -> Dict:
        """Analiza resultados de ajuste de modelos con soporte multiidioma y especificaciones adicionales"""

        # Prepare comprehensive data summary for the model
        data_summary = f"""
        FITTING RESULTS DATA (as JSON records for parsing):
        {json.dumps(data.to_dict('records'), indent=2)}

        DATA OVERVIEW:
        - Columns: {list(data.columns)}
        - Number of models evaluated: {len(data)}
        """

        # Get language prefix
        lang_prefix = self.get_language_prompt_prefix(language)

        # Add user additional specifications if they exist
        user_specs_section = f"""

        USER ADDITIONAL SPECIFICATIONS / FOCUS AREAS:
        {additional_specs}

        Please ensure your analysis incorporates these specific requirements and focus areas.
        """ if additional_specs else ""

        # Prompt enhanced with specific instructions for each level
        # Added system message for better role adherence
        if detail_level == "detailed":
            messages = [
                {"role": "system", "content": f"{lang_prefix} You are an expert in biotechnology and mathematical modeling, specializing in the comparative analysis of model fitting results. Your task is to provide a comprehensive, structured analysis of the provided data, focusing on the comparative performance of models across different experimental conditions. Include specific numerical values from the data in your analysis. Use Markdown formatting."}
            ]
            prompt_content = f"""
            Analyze these kinetic/biotechnological model fitting results.

            {user_specs_section}

            DETAIL LEVEL: DETAILED - Provide comprehensive analysis structured BY EXPERIMENT/CONDITION.

            PERFORM A COMPREHENSIVE COMPARATIVE ANALYSIS PER EXPERIMENT/CONDITION:

            1.  **IDENTIFY ALL EXPERIMENTS/CONDITIONS:** List and describe each unique experimental condition present in the data (e.g., pH levels, temperatures, media compositions).
            2.  **MODELS TESTED PER EXPERIMENT:** For EACH experiment, list ALL fitted mathematical models tested. Classify them (Biomass, Substrate, Product, etc.) if a 'Type' column exists.
            3.  **DETAILED COMPARISON PER EXPERIMENT:** Create a dedicated section for *each* experiment. Within each experiment section:
                *   Report the experimental condition.
                *   For each Variable Type (Biomass, Substrate, Product) analyzed in this experiment (if applicable):
                    *   Identify the **Best Model** based on R² (primary metric) and RMSE (secondary metric). State its name and the exact R² and RMSE values for this experiment.
                    *   List the values of the main parameters obtained for the best model in this specific experiment.
                    *   Provide a ranked list of *all* models tested for this variable type in this experiment, showing Model Name, R², and RMSE.
            4.  **COMPARATIVE TABLES (Across Experiments):**
                *   Create a summary table showing the Best Model, R², and RMSE for EACH Variable Type within EACH Experiment.
                *   Create a table summarizing the performance (Average R², Average RMSE, Number of experiments tested) of key models across *all* experiments where they were applied.
            5.  **PARAMETER ANALYSIS ACROSS EXPERIMENTS:** Analyze how the key parameters (e.g., μmax, Ks, Xmax) for frequently used or important models change from one experimental condition to another. Identify trends or sensitivities to experimental conditions.
            6.  **BIOLOGICAL INTERPRETATION & EXPERIMENTAL INSIGHTS:** For each experiment, provide a brief biological interpretation based on the fitting results and parameter values. Discuss whether the parameter values are biologically reasonable for the given conditions. Highlight key differences or findings between experiments.
            7.  **OVERALL BEST MODELS:** Based on performance across *all* experiments, identify the overall best model(s) for Biomass, Substrate, and Product (if applicable). Justify your selection with average metrics and consistency across conditions, citing numerical evidence.
            8.  **CONCLUSIONS AND RECOMMENDATIONS:** Summarize the main findings. Recommend which models are most robust or suitable for different types of analysis or specific experimental conditions. Discuss practical implications, confidence levels, and potential considerations for scale-up or further research based on the analysis.

            Use clear Markdown headings (`#`, `##`, `###`), bold text (`**text**`), and lists (`- ` or `1. `). Include ALL relevant numerical values from the provided data.
            """
            messages.append({"role": "user", "content": f"{prompt_content}\n\n{data_summary}"})

        else:  # summarized
             messages = [
                {"role": "system", "content": f"{lang_prefix} You are an expert in biotechnology, providing a concise comparative analysis of mathematical model fitting results across different experiments. Focus on identifying the best models per experiment and overall winners. Include essential numerical information. Use Markdown formatting."}
            ]
             prompt_content = f"""
            Analyze these kinetic/biotechnological model fitting results CONCISELY but completely, structured BY EXPERIMENT/CONDITION.

            {user_specs_section}

            DETAIL LEVEL: SUMMARIZED - Be concise but include all experiments and essential information.

            PROVIDE A FOCUSED COMPARATIVE ANALYSIS:

            1.  **EXPERIMENTS OVERVIEW:** Briefly state the total number of experiments/conditions analyzed and list the types of experimental conditions covered. Mention the variables measured (Biomass, Substrate, Product).
            2.  **BEST MODELS QUICK SUMMARY BY EXPERIMENT:** For *each* experiment/condition, clearly state:
                *   The experimental condition name.
                *   The Best Model found for Biomass (with its R² value).
                *   The Best Model found for Substrate (with its R² value).
                *   The Best Model found for Product (with its R² value).
                (Only include variable types present in the experiment).
            3.  **OVERALL BEST MODELS ACROSS ALL EXPERIMENTS:** Identify the single best model overall for Biomass, Substrate, and Product based on average performance or frequency of being the best model across experiments. State their average R² (if applicable) and mention how many experiments they were tested in.
            4.  **SUMMARY TABLE:** Provide a concise table summarizing the Best Model and its R²/RMSE for each Experiment and Variable Type combination.
            5.  **KEY FINDINGS & PARAMETER RANGES:** Highlight the most important findings. Briefly mention the observed range or average values for key parameters (e.g., μmax, Ks) across the experiments.
            6.  **PRACTICAL RECOMMENDATIONS:** Offer concise recommendations on which models are most suitable for which variables or conditions based on the analysis.

            Keep it concise but include ALL experiments, model names, and their key R² or RMSE metrics. Use Markdown.
            """
             messages.append({"role": "user", "content": f"{prompt_content}\n\n{data_summary}"})


        try:
            # Main analysis call
            response = self.client.chat.completions.create(
                model=qwen_model,
                messages=messages,
                temperature=self.temperature, # Use defined temperature
                top_p=self.top_p,           # Use defined top_p
                max_tokens=QWEN_MODELS.get(qwen_model, {}).get("max_tokens", 4000) # Use model max tokens, default 4000
            )

            # Extract analysis text
            analysis_text = response.choices[0].message.content

            # Generate implementation code - This prompt is adjusted to match the analysis structure
            # Also using a system message for the code generation role
            code_messages = [
                 {"role": "system", "content": f"{lang_prefix} You are an expert Python programmer specializing in biotechnological modeling and data analysis. Your task is to generate executable Python code based on the provided data and analysis. The code should implement the comparison of models by experiment and variable type, identify best models, and include basic plotting functions. Ensure actual numerical values from the data are used where appropriate for demonstration or analysis within the code."}
            ]
            code_prompt_content = f"""
            Generate complete, executable Python code to analyze and visualize the biotechnological model fitting results provided earlier.

            Use the actual data, which looks like this (as JSON records):
            {json.dumps(data.to_dict('records'), indent=2)}

            The code should:
            1.  Load this specific dataset.
            2.  Implement a class or functions to analyze model fitting results.
            3.  Perform analysis BY EXPERIMENT AND VARIABLE TYPE (Biomass, Substrate, Product), identifying the best model for each combination based on R² and RMSE.
            4.  Identify overall best models across all experiments for each variable type.
            5.  Include functions to generate visualizations comparing model performance (e.g., R² values) across experiments and variable types.
            6.  Include comments explaining the logic and findings, especially which model was best for which category/experiment and why.
            7.  Provide example usage of the code with the embedded data.

            Make the code robust and well-commented. Focus on clear data handling, analysis, and visualization.
            """
            code_messages.append({"role": "user", "content": code_prompt_content})


            code_response = self.client.chat.completions.create(
                model=qwen_model, # Use the same Qwen model for consistency
                messages=code_messages,
                temperature=0.5, # Slightly lower temp for more structured code
                top_p=0.9,
                max_tokens=3000 # Code might be shorter than analysis
            )

            # Extract code text, handle potential code block markdown
            code_text_raw = code_response.choices[0].message.content
            # Remove markdown code block fences if present
            if code_text_raw.startswith("```python"):
                 code_text = code_text_raw.strip().replace("```python\n", "", 1).strip("```")
            elif code_text_raw.startswith("```"):
                 # Handle generic code blocks
                 code_text = code_text_raw.strip().replace("```\n", "", 1).strip("```")
            else:
                 code_text = code_text_raw


            return {
                "tipo": "Comparative Analysis of Mathematical Models",
                "analisis_completo": analysis_text,
                "codigo_implementacion": code_text,
                "resumen_datos": {
                    "n_modelos": len(data),
                    "columnas": list(data.columns),
                    "metricas_disponibles": [col for col in data.columns if any(metric in col.lower()
                                           for metric in ['r2', 'rmse', 'aic', 'bic', 'mse'])],
                    # Safely get best R2 and model name if columns exist
                    "mejor_r2": data['R2'].max() if 'R2' in data.columns else None,
                    "mejor_modelo_r2": data.loc[data['R2'].idxmax()]['Model'] if 'R2' in data.columns and 'Model' in data.columns else None,
                    "datos_completos": data.to_dict('records') # Include all data for code
                }
            }

        except Exception as e:
            print(f"Error during AI analysis: {e}")
            return {"error": str(e)}

def process_files(files, qwen_model: str, detail_level: str = "detailed",
                 language: str = "en", additional_specs: str = "") -> Tuple[str, str]:
    """Procesa múltiples archivos con soporte de idioma y especificaciones adicionales"""
    # Check if the OpenAI client was successfully initialized
    if openai_client is None:
         error_msg = TRANSLATIONS.get(language, TRANSLATIONS['en'])['error_no_api']
         return error_msg, generate_implementation_code(error_msg) # Return error message and fallback code

    processor = FileProcessor()
    analyzer = AIAnalyzer(openai_client, model_registry) # Pass the initialized openai_client
    results = []
    all_code = []

    for file in files:
        if file is None:
            continue

        file_name = file.name if hasattr(file, 'name') else "archivo"
        file_ext = Path(file_name).suffix.lower()

        try:
            # Use tempfile to get the actual file path provided by Gradio
            if isinstance(file, str):
                 # Gradio >= 4.0 might pass strings (file paths)
                 file_path = file
            else:
                 # Handle older Gradio or other file-like objects if necessary
                 # For now, assume Gradio provides path string
                 raise TypeError("Unexpected file input type")

            file_content = None # Process using path or read bytes as needed
            df = None

            if file_ext in ['.csv', '.xlsx', '.xls']:
                if language == 'es':
                    results.append(f"## 📊 Análisis de Resultados: {file_name}")
                else:
                    results.append(f"## 📊 Results Analysis: {file_name}")

                # Read dataframe directly from path
                if file_ext == '.csv':
                    df = pd.read_csv(file_path)
                else:
                    df = pd.read_excel(file_path)

                if df is not None and not df.empty:
                    analysis_type = analyzer.detect_analysis_type(df)

                    if analysis_type == AnalysisType.FITTING_RESULTS:
                        result = analyzer.analyze_fitting_results(
                            df, qwen_model, detail_level, language, additional_specs
                        )

                        if "error" in result:
                             results.append(f"Error during analysis of {file_name}: {result['error']}")
                        else:
                            if language == 'es':
                                results.append("### 🎯 ANÁLISIS COMPARATIVO DE MODELOS MATEMÁTICOS")
                            else:
                                results.append("### 🎯 COMPARATIVE ANALYSIS OF MATHEMATICAL MODELS")
                            results.append(result.get("analisis_completo", ""))
                            if "codigo_implementacion" in result:
                                all_code.append(result["codigo_implementacion"])
                    elif analysis_type == AnalysisType.DATA_FITTING:
                         # Handle raw data - Could add a prompt here for Qwen
                         if language == 'es':
                              results.append(f"### 📈 Datos Experimentales Detectados: {file_name}")
                              results.append("Se detectaron datos experimentales. Esta herramienta se especializa en *resultados de ajuste*.")
                         else:
                             results.append(f"### 📈 Experimental Data Detected: {file_name}")
                             results.append("Experimental data was detected. This tool specializes in *fitting results*.")
                         # Optionally call Qwen to suggest fitting approach
                    elif analysis_type == AnalysisType.MATHEMATICAL_MODEL:
                         if language == 'es':
                             results.append(f"### 🔬 Descripción de Modelo Detectada: {file_name}")
                             results.append("Se detectó una descripción de modelo matemático. Esta herramienta se especializa en análisis comparativos de *resultados de ajuste*.")
                         else:
                              results.append(f"### 🔬 Mathematical Model Description Detected: {file_name}")
                              results.append("A mathematical model description was detected. This tool specializes in comparative analysis of *fitting results*.")
                    else: # Unknown
                         if language == 'es':
                              results.append(f"### 🤔 Tipo de Contenido Desconocido: {file_name}")
                              results.append("El tipo de contenido en este archivo no pudo ser determinado. Por favor, sube archivos con resultados de ajuste de modelos (con columnas como 'R2', 'RMSE', 'Model', etc.).")
                         else:
                             results.append(f"### 🤔 Unknown Content Type: {file_name}")
                             results.append("The type of content in this file could not be determined. Please upload files containing model fitting results (with columns like 'R2', 'RMSE', 'Model', etc.).")

                else:
                     if language == 'es':
                          results.append(f"### ⚠️ Error al leer o archivo vacío: {file_name}")
                     else:
                          results.append(f"### ⚠️ Error reading or empty file: {file_name}")

            # Add handling for PDF, ZIP if necessary, though the core tool is for CSV/Excel
            # elif file_ext == '.pdf':
            #     # Process PDF text if needed for model description analysis
            #     text = processor.extract_text_from_pdf(file_content)
            #     # Could call Qwen to analyze text here if needed
            #     results.append(f"Processed PDF {file_name}. Text extracted.")
            # elif file_ext == '.zip':
            #     extracted_files = processor.extract_from_zip(file_content)
            #     # Process extracted files recursively or as needed
            #     results.append(f"Processed ZIP {file_name}. Found {len(extracted_files)} files.")

            else:
                 if language == 'es':
                     results.append(f"### ⚠️ Formato de archivo no soportado: {file_name}")
                 else:
                     results.append(f"### ⚠️ Unsupported file format: {file_name}")

        except Exception as e:
             # Catch any unexpected errors during file processing
             if language == 'es':
                 results.append(f"### ❌ Error inesperado al procesar {file_name}: {str(e)}")
             else:
                 results.append(f"### ❌ Unexpected error processing {file_name}: {str(e)}")


        results.append("\n---\n") # Separator between files

    analysis_text = "\n".join(results)
    # Combine all generated code snippets
    # The fallback code generator is less critical now that the API generates code
    # But keep it as a safeguard or example if API fails
    code_text = "\n\n# === Combined Implementation Code ===\n\n" + "\n\n".join(all_code) if all_code else generate_implementation_code(analysis_text)

    return analysis_text, code_text


def generate_implementation_code(analysis_results: str) -> str:
    """Generates a default or fallback implementation code structure."""
    # This function is less critical if the AI generates code, but kept as a fallback.
    # The generated code structure from the AI is preferred.
    # This fallback provides a basic template if AI fails to produce code.

    code = """
# Fallback Implementation Code (Generated if AI code generation fails)
# This code provides a basic structure for analyzing fitting results.
# Replace placeholder data with your actual results dataframe.

import numpy as np
import pandas as pd
# Matplotlib and Seaborn imports moved here as they are for the generated code
import matplotlib.pyplot as plt
import seaborn as sns

# Visualization configuration
plt.style.use('seaborn-v0_8-darkgrid')
sns.set_palette("husl")

class ExperimentalModelAnalyzer:
    \"\"\"
    Basic class for comparative analysis of biotechnological models across multiple experiments.
    This is a fallback implementation.
    \"\"\"

    def __init__(self, results_df: pd.DataFrame = None):
        self.results_df = results_df
        if self.results_df is not None and 'Experiment' not in self.results_df.columns:
            # Add a default experiment if none exists
            self.results_df['Experiment'] = 'Default_Experiment'

    def load_results(self, file_path: str = None, data_dict: dict = None):
        \"\"\"Load fitting results from CSV/Excel file or dictionary\"\"\"
        if data_dict:
            self.results_df = pd.DataFrame(data_dict)
        elif file_path:
            if file_path.endswith('.csv'):
                self.results_df = pd.read_csv(file_path)
            else:
                self.results_df = pd.read_excel(file_path)

        if self.results_df is not None and 'Experiment' not in self.results_df.columns:
            self.results_df['Experiment'] = 'Default_Experiment'

        if self.results_df is not None:
            print(f"✅ Data loaded: {len(self.results_df)} models")
            print(f"📊 Available columns: {list(self.results_df.columns)}")
            if 'Experiment' in self.results_df.columns:
                print(f"🧪 Experiments found: {self.results_df['Experiment'].unique()}")

    def analyze_by_experiment(self,
                            experiment_col: str = 'Experiment',
                            model_col: str = 'Model',
                            type_col: str = 'Type',
                            r2_col: str = 'R2',
                            rmse_col: str = 'RMSE') -> Dict:
        \"\"\"
        Analyze models by experiment and variable type.
        Identifies best models for biomass, substrate, and product in each experiment.
        \"\"\"
        if self.results_df is None or self.results_df.empty:
            print("⚠️ No data loaded for analysis.")
            return {}

        results_by_exp = {}
        experiments = self.results_df[experiment_col].unique()

        print("\\n" + "="*80)
        print("📊 ANALYSIS BY EXPERIMENT AND VARIABLE TYPE")
        print("="*80)

        for exp in experiments:
            print(f"\\n🧪 EXPERIMENT: {exp}")
            print("-"*50)

            exp_data = self.results_df[self.results_df[experiment_col] == exp].copy() # Use copy to avoid SettingWithCopyWarning
            results_by_exp[exp] = {}

            var_types = exp_data[type_col].unique() if type_col in exp_data.columns else ['All_Types']

            for var_type in var_types:
                if type_col in exp_data.columns:
                    var_data = exp_data[exp_data[type_col] == var_type]
                else:
                    var_data = exp_data # Analyze all together if no type column

                if not var_data.empty and r2_col in var_data.columns:
                    # Find best model for this variable type (or all) based on R2
                    best_idx = var_data[r2_col].idxmax()
                    best_model = var_data.loc[best_idx]

                    results_by_exp[exp][var_type] = {
                        'best_model': best_model.get(model_col, 'N/A'),
                        'r2': best_model.get(r2_col, np.nan),
                        'rmse': best_model.get(rmse_col, np.nan),
                        'all_models': var_data[[model_col, r2_col, rmse_col]].to_dict('records') if {model_col, r2_col, rmse_col}.issubset(var_data.columns) else var_data.to_dict('records')
                    }

                    print(f"\\n  📈 {var_type.upper()}:")
                    print(f"    Best Model: {results_by_exp[exp][var_type]['best_model']}")
                    print(f"    R² = {results_by_exp[exp][var_type]['r2']:.4f}" if not np.isnan(results_by_exp[exp][var_type]['r2']) else "R² = N/A")
                    print(f"    RMSE = {results_by_exp[exp][var_type]['rmse']:.4f}" if not np.isnan(results_by_exp[exp][var_type]['rmse']) else "RMSE = N/A")

                    # Show all models for this variable
                    if 'all_models' in results_by_exp[exp][var_type]:
                        print(f"\\n    All {var_type} models tested:")
                        for model_entry in results_by_exp[exp][var_type]['all_models']:
                             r2_val = model_entry.get(r2_col, np.nan)
                             rmse_val = model_entry.get(rmse_col, np.nan)
                             model_name = model_entry.get(model_col, 'N/A')
                             print(f"      - {model_name}: R²={r2_val:.4f}" if not np.isnan(r2_val) else f"      - {model_name}: R²=N/A", end="")
                             print(f", RMSE={rmse_val:.4f}" if not np.isnan(rmse_val) else ", RMSE=N/A")
                elif not var_data.empty:
                     print(f"\\n  📈 {var_type.upper()}:")
                     print(f"    No '{r2_col}' column found for comparison.")
                else:
                     print(f"\\n  📈 {var_type.upper()}:")
                     print(f"    No data found for this variable type.")


        self.best_models_by_experiment = results_by_exp
        return results_by_exp

    def _determine_overall_best_models(self):
        \"\"\"Determine the best models across all experiments\"\"\"
        if not hasattr(self, 'best_models_by_experiment') or not self.best_models_by_experiment:
             print("⚠️ No experimental analysis available to determine overall models.")
             return {}

        print("\\n" + "="*80)
        print("🏆 OVERALL BEST MODELS ACROSS ALL EXPERIMENTS")
        print("="*80)

        model_performance = {}

        for exp, exp_results in self.best_models_by_experiment.items():
            for var_type, var_results in exp_results.items():
                if var_type not in model_performance:
                    model_performance[var_type] = {}

                # Use the list of all models analyzed for this type in this experiment
                models_in_exp_type = var_results.get('all_models', [])

                for model_data in models_in_exp_type:
                    model_name = model_data.get('Model', 'Unknown Model') # Use .get for safety
                    r2_val = model_data.get('R2')
                    rmse_val = model_data.get('RMSE')

                    if model_name not in model_performance[var_type]:
                        model_performance[var_type][model_name] = {
                            'r2_values': [],
                            'rmse_values': [],
                            'experiments': []
                        }

                    if r2_val is not None:
                         model_performance[var_type][model_name]['r2_values'].append(r2_val)
                    if rmse_val is not None:
                        model_performance[var_type][model_name]['rmse_values'].append(rmse_val)

                    if exp not in model_performance[var_type][model_name]['experiments']:
                         model_performance[var_type][model_name]['experiments'].append(exp)

        overall_best_models = {}

        # Calculate average performance and select best
        for var_type, models in model_performance.items():
            best_avg_r2 = -np.inf # Use -infinity to ensure any valid R2 is better
            best_model_info = None

            print(f"\\n📊 {var_type.upper()} MODELS:")
            if not models:
                 print("  No models found for this type.")
                 continue

            for model_name, perf_data in models.items():
                # Calculate average R2, ignoring NaNs
                r2_values = [v for v in perf_data['r2_values'] if v is not None and not np.isnan(v)]
                avg_r2 = np.mean(r2_values) if r2_values else -np.inf # Handle case with no valid R2

                # Calculate average RMSE, ignoring NaNs
                rmse_values = [v for v in perf_data['rmse_values'] if v is not None and not np.isnan(v)]
                avg_rmse = np.mean(rmse_values) if rmse_values else np.inf # Handle case with no valid RMSE

                n_exp = len(perf_data['experiments'])

                print(f"  {model_name}:")
                print(f"    Average R² = {avg_r2:.4f}" if avg_r2 > -np.inf else "    Average R² = N/A")
                print(f"    Average RMSE = {avg_rmse:.4f}" if avg_rmse < np.inf else "    Average RMSE = N/A")
                print(f"    Tested in {n_exp} experiments")

                # Selection logic: prioritize higher average R2. Could add secondary criteria (e.g., lower RMSE, consistency)
                if avg_r2 > best_avg_r2:
                    best_avg_r2 = avg_r2
                    best_model_info = {
                        'name': model_name,
                        'avg_r2': avg_r2,
                        'avg_rmse': avg_rmse,
                        'n_experiments': n_exp
                    }
                elif avg_r2 == best_avg_r2 and avg_rmse < (best_model_info['avg_rmse'] if best_model_info and best_model_info['avg_rmse'] < np.inf else np.inf):
                     # Tie-breaking: prefer lower average RMSE if R2 is the same
                     best_model_info = {
                        'name': model_name,
                        'avg_r2': avg_r2,
                        'avg_rmse': avg_rmse,
                        'n_experiments': n_exp
                    }


            if best_model_info and var_type.lower() in ['biomass', 'substrate', 'product', 'all_types']:
                 # Assign to standard keys if they exist
                 target_key = var_type.lower() if var_type.lower() in ['biomass', 'substrate', 'product'] else 'overall'
                 overall_best_models[target_key] = best_model_info
                 print(f"\\n  🏆 BEST {var_type.upper()} MODEL: {best_model_info['name']} (Avg R²={best_model_info['avg_r2']:.4f})" if best_model_info['avg_r2'] > -np.inf else f"\\n  🏆 BEST {var_type.upper()} MODEL: {best_model_info['name']} (Avg R²=N/A)")
            elif best_model_info:
                 # Add other types found
                 overall_best_models[var_type] = best_model_info


        self.overall_best_models = overall_best_models
        return overall_best_models


    def create_comparison_visualizations(self):
        \"\"\"Create visualizations comparing models across experiments\"\"\"
        if not hasattr(self, 'best_models_by_experiment') or not self.best_models_by_experiment:
            print("⚠️ No analysis results to visualize.")
            return # Exit if no data

        # Prepare data for visualization - focusing on R2 for best models per experiment/type
        plot_data = []
        for exp, results in self.best_models_by_experiment.items():
            for var_type, var_results in results.items():
                plot_data.append({
                    'Experiment': exp,
                    'Variable_Type': var_type,
                    'Best_Model': var_results.get('best_model'),
                    'R2': var_results.get('r2')
                })

        plot_df = pd.DataFrame(plot_data)
        plot_df = plot_df.dropna(subset=['R2']) # Only plot entries with R2

        if plot_df.empty:
            print("⚠️ No valid R² data available for visualization.")
            return

        # Use Seaborn for better aesthetics
        plt.figure(figsize=(14, 8))
        sns.barplot(data=plot_df, x='Experiment', y='R2', hue='Variable_Type', palette='viridis')

        plt.title('Best Model R² Comparison by Experiment and Variable Type', fontsize=16)
        plt.xlabel('Experiment', fontsize=12)
        plt.ylabel('R²', fontsize=12)
        plt.xticks(rotation=45, ha='right')
        plt.ylim(0, 1.05) # R2 is typically between 0 and 1
        plt.legend(title='Variable Type')
        plt.tight_layout() # Adjust layout to prevent labels overlapping
        plt.grid(axis='y', linestyle='--', alpha=0.7)
        plt.show()

        # Optional: Add more plots if needed, e.g., parameter trends


    def generate_summary_table(self) -> pd.DataFrame:
        \"\"\"Generate a summary table of best models by experiment and type\"\"\"
        if not hasattr(self, 'best_models_by_experiment') or not self.best_models_by_experiment:
            print("⚠️ No analysis results to generate summary table.")
            return pd.DataFrame()

        summary_data = []

        for exp, results in self.best_models_by_experiment.items():
            for var_type, var_results in results.items():
                summary_data.append({
                    'Experiment': exp,
                    'Variable_Type': var_type,
                    'Best_Model': var_results.get('best_model', 'N/A'),
                    'R2': var_results.get('r2', np.nan),
                    'RMSE': var_results.get('rmse', np.nan)
                })

        summary_df = pd.DataFrame(summary_data)

        print("\\n📋 SUMMARY TABLE: BEST MODELS BY EXPERIMENT AND VARIABLE TYPE")
        print("="*80)
        if not summary_df.empty:
             # Format R2 and RMSE for display
             summary_df_display = summary_df.copy()
             if 'R2' in summary_df_display.columns:
                  summary_df_display['R2'] = summary_df_display['R2'].apply(lambda x: f'{x:.4f}' if pd.notna(x) else 'N/A')
             if 'RMSE' in summary_df_display.columns:
                  summary_df_display['RMSE'] = summary_df_display['RMSE'].apply(lambda x: f'{x:.4f}' if pd.notna(x) else 'N/A')

             print(summary_df_display.to_string(index=False))
        else:
             print("No data to display in the summary table.")

        return summary_df

# Example usage for the fallback code structure
# Note: The AI-generated code should ideally replace this example usage
# but this part demonstrates how the generated code might be used.
if __name__ == "__main__":
    print("🧬 Experimental Model Comparison System (Fallback Code Example)")
    print("="*60)

    # --- Placeholder Example Data ---
    # This data structure should match the format the AI expects and uses
    # in the generated code. It includes 'Experiment' and 'Type'.
    fallback_example_data = {
        'Experiment': ['pH_7.0', 'pH_7.0', 'pH_7.0', 'pH_7.5', 'pH_7.5', 'pH_7.5',
                      'pH_7.0', 'pH_7.0', 'pH_7.5', 'pH_7.5',
                      'pH_7.0', 'pH_7.0', 'pH_7.5', 'pH_7.5'],
        'Model': ['Monod', 'Logistic', 'Gompertz', 'Monod', 'Logistic', 'Gompertz',
                 'First_Order', 'Monod_Substrate', 'First_Order', 'Monod_Substrate',
                 'Luedeking_Piret', 'Linear', 'Luedeking_Piret', 'Linear'],
        'Type': ['Biomass', 'Biomass', 'Biomass', 'Biomass', 'Biomass', 'Biomass',
                'Substrate', 'Substrate', 'Substrate', 'Substrate',
                'Product', 'Product', 'Product', 'Product'],
        'R2': [0.9845, 0.9912, 0.9956, 0.9789, 0.9834, 0.9901,
              0.9723, 0.9856, 0.9698, 0.9812,
              0.9634, 0.9512, 0.9687, 0.9423],
        'RMSE': [0.0234, 0.0189, 0.0145, 0.0267, 0.0223, 0.0178,
                0.0312, 0.0245, 0.0334, 0.0289,
                0.0412, 0.0523, 0.0389, 0.0567],
        'mu_max': [0.45, 0.48, 0.52, 0.42, 0.44, 0.49,
                  np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan],
        'Ks': [np.nan, np.nan, np.nan, np.nan, np.nan, np.nan,
              2.1, 1.8, 2.3, 1.9, np.nan, np.nan, np.nan, np.nan]
    }

    # Create analyzer instance using the fallback data
    analyzer = ExperimentalModelAnalyzer(results_df=pd.DataFrame(fallback_example_data))

    # Analyze by experiment
    analysis_results = analyzer.analyze_by_experiment()

    # Determine overall best models
    overall_best = analyzer._determine_overall_best_models()
    print(f"Overall Best Models (Determined by Fallback): {overall_best}")


    # Create visualizations (will use the best_models_by_experiment attribute)
    print("\\nAttempting to create visualizations...")
    try:
        analyzer.create_comparison_visualizations()
    except Exception as e:
        print(f"Error creating visualization: {e}")
        print("This might happen if data structure or plotting logic is not fully compatible.")


    # Generate summary table
    summary_table = analyzer.generate_summary_table()

    print("\\n✨ Fallback Analysis complete!")

# --- End of Fallback Code Example ---


# Estado global para almacenar resultados
class AppState:
    def __init__(self):
        self.current_analysis = ""
        self.current_code = ""
        self.current_language = "en"

app_state = AppState()

def export_report(export_format: str, language: str) -> Tuple[str, str]:
    """Exporta el reporte al formato seleccionado"""
    if not app_state.current_analysis:
        error_msg = {
            'en': "No analysis available to export",
            'es': "No hay análisis disponible para exportar",
            'fr': "Aucune analyse disponible pour exporter",
            'de': "Keine Analyse zum Exportieren verfügbar",
            'pt': "Nenhuma análise disponível para exportar"
        }
        return error_msg.get(language, error_msg['en']), ""

    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")

    try:
        # Ensure ReportExporter is used correctly with its static methods
        if export_format == "DOCX":
            filename = f"biotech_analysis_report_{timestamp}.docx"
            ReportExporter.export_to_docx(app_state.current_analysis, filename, language)
        else:  # PDF
            filename = f"biotech_analysis_report_{timestamp}.pdf"
            ReportExporter.export_to_pdf(app_state.current_analysis, filename, language)

        success_msg = TRANSLATIONS[language]['report_exported']
        return f"{success_msg} {filename}", filename
    except Exception as e:
        # Provide more specific error details for export
        return f"Error exporting report: {str(e)}", ""


# Interfaz Gradio con soporte multiidioma y temas
def create_interface():
    # Estado inicial
    current_theme = "light"
    current_language = "en"

    def update_interface_language(language):
        """Actualiza el idioma de la interfaz"""
        app_state.current_language = language
        t = TRANSLATIONS[language]

        # Build model choices string with descriptions for info text
        model_info_str = ""
        # Default model might change based on QWEN_MODELS keys
        default_model_key = list(QWEN_MODELS.keys())[0] if QWEN_MODELS else "Qwen/Qwen3-14B"
        if default_model_key in QWEN_MODELS:
             model_info_str = f"{t['best_for']}: {QWEN_MODELS[default_model_key]['best_for']}"


        return [
            gr.update(value=f"# {t['title']}"),  # title_text
            gr.update(value=t['subtitle']),       # subtitle_text
            gr.update(label=t['upload_files']),   # files_input
            gr.update(label=t['select_model'], info=model_info_str),   # model_selector
            gr.update(label=t['select_language']), # language_selector
            gr.update(label=t['select_theme']),   # theme_selector
            gr.update(label=t['detail_level'], choices=[(t['detailed'], "detailed"), (t['summarized'], "summarized")]),   # detail_level
            gr.update(label=t['additional_specs'], placeholder=t['additional_specs_placeholder']), # additional_specs
            gr.update(value=t['analyze_button']), # analyze_btn
            gr.update(label=t['export_format']),  # export_format
            gr.update(value=t['export_button']),  # export_btn
            gr.update(label=t['comparative_analysis']), # analysis_output
            gr.update(label=t['implementation_code']),  # code_output
            gr.update(label=t['data_format'])    # data_format_accordion
        ]

    def process_and_store(files, model, detail, language, additional_specs):
        """Procesa files y almacena resultados"""
        if not files:
            error_msg = TRANSLATIONS[language]['error_no_files']
            app_state.current_analysis = error_msg
            app_state.current_code = generate_implementation_code(error_msg) # Provide fallback code even on file error
            return error_msg, app_state.current_code

        # Assuming files is a list of strings (filepaths) from Gradio
        analysis, code = process_files(files, model, detail, language, additional_specs)

        # Store results in app state
        app_state.current_analysis = analysis
        app_state.current_code = code

        return analysis, code

    # Get default model key safely
    default_qwen_model_key = list(QWEN_MODELS.keys())[0] if QWEN_MODELS else ""
    default_qwen_model_info = QWEN_MODELS.get(default_qwen_model_key, {})
    default_model_info_str = f"{TRANSLATIONS[current_language]['best_for']}: {default_qwen_model_info.get('best_for', 'N/A')}"


    with gr.Blocks(theme=THEMES[current_theme]) as demo:
        # Componentes de UI
        with gr.Row():
            with gr.Column(scale=3):
                title_text = gr.Markdown(f"# {TRANSLATIONS[current_language]['title']}")
                subtitle_text = gr.Markdown(TRANSLATIONS[current_language]['subtitle'])
            with gr.Column(scale=1):
                with gr.Row():
                    language_selector = gr.Dropdown(
                        choices=[("English", "en"), ("Español", "es"), ("Français", "fr"),
                                ("Deutsch", "de"), ("Português", "pt")],
                        value=current_language,
                        label=TRANSLATIONS[current_language]['select_language'],
                        interactive=True
                    )
                    theme_selector = gr.Dropdown(
                        choices=[("Light", "light"), ("Dark", "dark")],
                        value=current_theme,
                        label=TRANSLATIONS[current_language]['select_theme'],
                        interactive=True
                    )

        with gr.Row():
            with gr.Column(scale=1):
                files_input = gr.File(
                    label=TRANSLATIONS[current_language]['upload_files'],
                    file_count="multiple",
                    file_types=[".csv", ".xlsx", ".xls"], # Focusing on data files
                    type="filepath" # Get file path as string
                )

                model_selector = gr.Dropdown(
                    choices=list(QWEN_MODELS.keys()),
                    value=default_qwen_model_key,
                    label=TRANSLATIONS[current_language]['select_model'],
                    info=default_model_info_str
                )

                detail_level = gr.Radio(
                    choices=[
                        (TRANSLATIONS[current_language]['detailed'], "detailed"),
                        (TRANSLATIONS[current_language]['summarized'], "summarized")
                    ],
                    value="detailed",
                    label=TRANSLATIONS[current_language]['detail_level']
                )

                # Nueva entrada para especificaciones adicionales
                additional_specs = gr.Textbox(
                    label=TRANSLATIONS[current_language]['additional_specs'],
                    placeholder=TRANSLATIONS[current_language]['additional_specs_placeholder'],
                    lines=3,
                    max_lines=5,
                    interactive=True
                )

                analyze_btn = gr.Button(
                    TRANSLATIONS[current_language]['analyze_button'],
                    variant="primary",
                    size="lg"
                )

                gr.Markdown("---")

                export_format = gr.Radio(
                    choices=["DOCX", "PDF"],
                    value="PDF",
                    label=TRANSLATIONS[current_language]['export_format']
                )

                export_btn = gr.Button(
                    TRANSLATIONS[current_language]['export_button'],
                    variant="secondary"
                )

                export_status = gr.Textbox(
                    label="Export Status",
                    interactive=False,
                    visible=False
                )

                export_file = gr.File(
                    label="Download Report",
                    visible=False
                )

            with gr.Column(scale=2):
                analysis_output = gr.Markdown(
                    label=TRANSLATIONS[current_language]['comparative_analysis']
                )

                code_output = gr.Code(
                    label=TRANSLATIONS[current_language]['implementation_code'],
                    language="python",
                    interactive=True,
                    lines=20
                )

        data_format_accordion = gr.Accordion(
            label=TRANSLATIONS[current_language]['data_format'],
            open=False
        )

        with data_format_accordion:
            gr.Markdown("""
            ### Expected CSV/Excel structure:

            | Experiment | Model | Type | R2 | RMSE | AIC | BIC | mu_max | Ks | Parameters |
            |------------|-------|------|-----|------|-----|-----|--------|-------|------------|
            | pH_7.0 | Monod | Biomass | 0.985 | 0.023 | -45.2 | -42.1 | 0.45 | 2.1 | {...} |
            | pH_7.0 | Logistic | Biomass | 0.976 | 0.031 | -42.1 | -39.5 | 0.42 | - | {...} |
            | pH_7.0 | First_Order | Substrate | 0.992 | 0.018 | -48.5 | -45.2 | - | 1.8 | {...} |
            | pH_7.5 | Monod | Biomass | 0.978 | 0.027 | -44.1 | -41.2 | 0.43 | 2.2 | {...} |

            **Important columns:**
            - **Experiment**: Experimental condition identifier (Optional, but recommended for comparative analysis)
            - **Model**: Model name (e.g., Monod, Logistic)
            - **Type**: Variable type (Biomass, Substrate, Product) (Optional, but recommended for analysis by type)
            - **R2, RMSE**: Fit quality metrics (At least one needed for comparison)
            - **Parameters**: Columns for model-specific parameters (e.g., mu_max, Ks, Xmax)
            """)

        # Definir ejemplos (Update example paths if necessary)
        examples = gr.Examples(
            examples=[
                [["examples/biomass_models_comparison.csv"], list(QWEN_MODELS.keys())[0] if QWEN_MODELS else "", "detailed", ""],
                [["examples/substrate_kinetics_results.xlsx"], list(QWEN_MODELS.keys())[0] if QWEN_MODELS else "", "summarized", "Focus on temperature effects"]
            ],
            inputs=[files_input, model_selector, detail_level, additional_specs],
            label=TRANSLATIONS[current_language]['examples']
        )


        # Eventos - Actualizado para incluir additional_specs
        language_selector.change(
            update_interface_language,
            inputs=[language_selector],
            outputs=[
                title_text, subtitle_text, files_input, model_selector,
                language_selector, theme_selector, detail_level, additional_specs,
                analyze_btn, export_format, export_btn, analysis_output,
                code_output, data_format_accordion
            ]
        )

        def change_theme(theme_name):
            """Cambia el tema de la interfaz"""
            # Note: Dynamic theme switching in Gradio might require a page reload for full effect.
            # This function primarily triggers the UI update but the theme itself is set at gr.Blocks creation.
            # Returning gr.Info is a common way to indicate the change.
            # To truly change theme dynamically, you might need Javascript or specific Gradio features.
            return gr.Info("Theme applied. May require page refresh for full effect on all components.")

        theme_selector.change(
            change_theme,
            inputs=[theme_selector],
            outputs=[] # No direct UI output change from this function in the current structure
        )

        analyze_btn.click(
            fn=process_and_store,
            inputs=[files_input, model_selector, detail_level, language_selector, additional_specs],
            outputs=[analysis_output, code_output]
        )

        def handle_export(format, language):
            status, file = export_report(format, language)
            # Check if the file was successfully created before making the download button visible
            if file and os.path.exists(file):
                return gr.update(value=status, visible=True), gr.update(value=file, visible=True, label=f"Download {format}")
            else:
                # Hide the download button if no file was created
                return gr.update(value=status, visible=True), gr.update(value=None, visible=False)


        export_btn.click(
            fn=handle_export,
            inputs=[export_format, language_selector],
            outputs=[export_status, export_file]
        )

    return demo

# Función principal
def main():
    # Check for the specific API key required for Qwen
    if openai_client is None:
        print("⚠️ NEBIUS_API_KEY environment variable not found. Please configure it.")
        return gr.Interface(
            fn=lambda: TRANSLATIONS['en']['error_no_api'], # Display error message in UI
            inputs=None, # No inputs needed for just showing error
            outputs=gr.Textbox(label="Configuration Error"),
            title=TRANSLATIONS['en']['title'],
            description="Failed to initialize AI client.",
            theme=THEMES['light'] # Use a default theme
        )

    # Proceed with creating the interface if client is initialized
    return create_interface()

# Para ejecución local
if __name__ == "__main__":
    # Ensure Gradio example paths exist for the examples section
    if not os.path.exists("examples"):
         os.makedirs("examples")
    # Create dummy example files if they don't exist
    if not os.path.exists("examples/biomass_models_comparison.csv"):
        dummy_csv_data = {'Experiment': ['ExpA', 'ExpA', 'ExpB', 'ExpB'],
                          'Model': ['Monod', 'Logistic', 'Monod', 'Logistic'],
                          'Type': ['Biomass', 'Biomass', 'Biomass', 'Biomass'],
                          'R2': [0.98, 0.97, 0.95, 0.96],
                          'RMSE': [0.02, 0.03, 0.04, 0.035],
                          'mu_max': [0.5, 0.48, 0.4, 0.38]}
        pd.DataFrame(dummy_csv_data).to_csv("examples/biomass_models_comparison.csv", index=False)

    if not os.path.exists("examples/substrate_kinetics_results.xlsx"):
         dummy_excel_data = {'Experiment': ['Temp25', 'Temp25', 'Temp30', 'Temp30'],
                             'Model': ['First_Order', 'Monod_Substrate', 'First_Order', 'Monod_Substrate'],
                             'Type': ['Substrate', 'Substrate', 'Substrate', 'Substrate'],
                             'R2': [0.99, 0.98, 0.97, 0.985],
                             'RMSE': [0.015, 0.02, 0.025, 0.018],
                             'Ks': [1.5, 1.2, 1.8, 1.4]}
         pd.DataFrame(dummy_excel_data).to_excel("examples/substrate_kinetics_results.xlsx", index=False)


    demo = main()
    if demo:
        demo.launch(
            server_name="0.0.0.0",
            server_port=7860,
            share=False
        )