Update app.py

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-#import anthropic
+# import anthropic # Removed Anthropic import
 import PyPDF2
 import pandas as pd
 import numpy as np
@@ -11,31 +11,43 @@ import tempfile
 from typing import Dict, List, Tuple, Union, Optional
 import re
 from pathlib import Path
-import openpyxl
+import openpyxl # Needed for reading .xlsx
 from dataclasses import dataclass
 from enum import Enum
-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
-from reportlab.pdfbase import pdfmetrics
-from reportlab.pdfbase.ttfonts import TTFont
-import matplotlib.pyplot as plt
+# 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'
 
-# Inicializar cliente Anthropic
-#client = anthropic.Anthropic()
+# 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.")
 
-# Inicializar cliente Nebius
-from openai import OpenAI
-client = OpenAI(api_key=os.getenv("NEBIUS_API_KEY"))
 
 # Sistema de traducción - Actualizado con nuevas entradas
 TRANSLATIONS = {
@@ -43,7 +55,7 @@ TRANSLATIONS = {
         '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': '🤖 Claude Model',
+        'select_model': '🤖 Qwen Model', # Changed label
         'select_language': '🌐 Language',
         'select_theme': '🎨 Theme',
         'detail_level': '📋 Analysis detail level',
@@ -60,7 +72,7 @@ TRANSLATIONS = {
         'dark': 'Dark',
         'best_for': 'Best for',
         'loading': 'Loading...',
-        'error_no_api': 'Please configure ANTHROPIC_API_KEY in HuggingFace Space secrets',
+        '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:',
@@ -74,7 +86,7 @@ TRANSLATIONS = {
         '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 Claude',
+        'select_model': '🤖 Modelo Qwen', # Changed label
         'select_language': '🌐 Idioma',
         'select_theme': '🎨 Tema',
         'detail_level': '📋 Nivel de detalle del análisis',
@@ -91,7 +103,7 @@ TRANSLATIONS = {
         'dark': 'Oscuro',
         'best_for': 'Mejor para',
         'loading': 'Cargando...',
-        'error_no_api': 'Por favor configura ANTHROPIC_API_KEY en los secretos del Space',
+        '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:',
@@ -105,7 +117,7 @@ TRANSLATIONS = {
         '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 Claude',
+        'select_model': '🤖 Modèle Qwen', # Changed label
         'select_language': '🌐 Langue',
         'select_theme': '🎨 Thème',
         'detail_level': '📋 Niveau de détail',
@@ -122,7 +134,7 @@ TRANSLATIONS = {
         'dark': 'Sombre',
         'best_for': 'Meilleur pour',
         'loading': 'Chargement...',
-        'error_no_api': 'Veuillez configurer ANTHROPIC_API_KEY',
+        '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:',
@@ -136,7 +148,7 @@ TRANSLATIONS = {
         'title': '🧬 Vergleichender Analysator für Biotechnologische Modelle',
         'subtitle': 'Spezialisiert auf vergleichende Analyse von Modellanpassungsergebnissen',
         'upload_files': '📁 Ergebnisse hochladen (CSV/Excel)',
-        'select_model': '🤖 Claude Modell',
+        'select_model': '🤖 Qwen Modell', # Changed label
         'select_language': '🌐 Sprache',
         'select_theme': '🎨 Thema',
         'detail_level': '📋 Detailgrad der Analyse',
@@ -153,7 +165,7 @@ TRANSLATIONS = {
         'dark': 'Dunkel',
         'best_for': 'Am besten für',
         'loading': 'Laden...',
-        'error_no_api': 'Bitte konfigurieren Sie ANTHROPIC_API_KEY',
+        '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:',
@@ -167,7 +179,7 @@ TRANSLATIONS = {
         'title': '🧬 Analisador Comparativo de Modelos Biotecnológicos',
         'subtitle': 'Especializado em análise comparativa de resultados de ajuste',
         'upload_files': '📁 Carregar resultados (CSV/Excel)',
-        'select_model': '🤖 Modelo Claude',
+        'select_model': '🤖 Modelo Qwen', # Changed label
         'select_language': '🌐 Idioma',
         'select_theme': '🎨 Tema',
         'detail_level': '📋 Nível de detalhe',
@@ -184,7 +196,7 @@ TRANSLATIONS = {
         'dark': 'Escuro',
         'best_for': 'Melhor para',
         'loading': 'Carregando...',
-        'error_no_api': 'Por favor configure ANTHROPIC_API_KEY',
+        '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:',
@@ -245,21 +257,21 @@ 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
@@ -272,7 +284,7 @@ class ModelRegistry:
             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)",
@@ -282,7 +294,7 @@ class ModelRegistry:
             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))",
@@ -296,18 +308,26 @@ class ModelRegistry:
 # Instancia global del registro
 model_registry = ModelRegistry()
 
-CLAUDE_MODELS = {
-    "Qwen/Qwen3-14B": {
+# Available Qwen Models (Updated from Claude)
+QWEN_MODELS = {
+    "Qwen/Qwen3-14B": { # Using the model specified by the user
         "name": "Qwen 3-14B",
-        "description": "Modelo Qwen 3-14B para análisis detallado",
-        "max_tokens": 4096,
-        "best_for": "Análisis técnico y científico"
-    }
+        "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"""
@@ -319,23 +339,25 @@ class FileProcessor:
             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"""
@@ -352,21 +374,36 @@ class FileProcessor:
 
 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 = Document()
-        
+        doc = ReportExporter.Document()
+
         # Configurar estilos
         title_style = doc.styles['Title']
-        title_style.font.size = Pt(24)
+        title_style.font.size = ReportExporter.Pt(24)
         title_style.font.bold = True
-        
+
         heading_style = doc.styles['Heading 1']
-        heading_style.font.size = Pt(18)
+        heading_style.font.size = ReportExporter.Pt(18)
         heading_style.font.bold = True
-        
+
         # Título
         title_text = {
             'en': 'Comparative Analysis Report - Biotechnological Models',
@@ -375,9 +412,9 @@ class ReportExporter:
             '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',
@@ -388,14 +425,14 @@ class ReportExporter:
         }
         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('##'):
@@ -419,36 +456,36 @@ class ReportExporter:
             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 = SimpleDocTemplate(filename, pagesize=letter)
+        doc = ReportExporter.SimpleDocTemplate(filename, pagesize=ReportExporter.letter)
         story = []
-        styles = getSampleStyleSheet()
-        
+        styles = ReportExporter.getSampleStyleSheet()
+
         # Estilos personalizados
-        title_style = ParagraphStyle(
+        title_style = ReportExporter.ParagraphStyle(
             'CustomTitle',
             parent=styles['Title'],
             fontSize=24,
-            textColor=colors.HexColor('#1f4788'),
+            textColor=ReportExporter.colors.HexColor('#1f4788'),
             spaceAfter=30
         )
-        
-        heading_style = ParagraphStyle(
+
+        heading_style = ReportExporter.ParagraphStyle(
             'CustomHeading',
             parent=styles['Heading1'],
             fontSize=16,
-            textColor=colors.HexColor('#2e5090'),
+            textColor=ReportExporter.colors.HexColor('#2e5090'),
             spaceAfter=12
         )
-        
+
         # Título
         title_text = {
             'en': 'Comparative Analysis Report - Biotechnological Models',
@@ -457,9 +494,9 @@ class ReportExporter:
             'de': 'Vergleichsanalysebericht - Biotechnologische Modelle',
             'pt': 'Relatório de Análise Comparativa - Modelos Biotecnológicos'
         }
-        
-        story.append(Paragraph(title_text.get(language, title_text['en']), title_style))
-        
+
+        story.append(ReportExporter.Paragraph(title_text.get(language, title_text['en']), title_style))
+
         # Fecha
         date_text = {
             'en': 'Generated on',
@@ -468,483 +505,781 @@ class ReportExporter:
             'de': 'Erstellt am',
             'pt': 'Gerado em'
         }
-        story.append(Paragraph(f"{date_text.get(language, date_text['en'])}: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}", styles['Normal']))
-        story.append(Spacer(1, 0.5*inch))
-        
+        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(Spacer(1, 0.2*inch))
+                story.append(ReportExporter.Spacer(1, 0.2*ReportExporter.inch))
             elif line.startswith('###'):
-                story.append(Paragraph(line.replace('###', '').strip(), styles['Heading3']))
+                story.append(ReportExporter.Paragraph(line.replace('###', '').strip(), styles['Heading3']))
             elif line.startswith('##'):
-                story.append(Paragraph(line.replace('##', '').strip(), styles['Heading2']))
+                story.append(ReportExporter.Paragraph(line.replace('##', '').strip(), styles['Heading2']))
             elif line.startswith('#'):
-                story.append(Paragraph(line.replace('#', '').strip(), heading_style))
+                story.append(ReportExporter.Paragraph(line.replace('#', '').strip(), heading_style))
             elif line.startswith('**') and line.endswith('**'):
                 text = line.replace('**', '')
-                story.append(Paragraph(f"<b>{text}</b>", styles['Normal']))
+                story.append(ReportExporter.Paragraph(f"<b>{text}</b>", styles['Normal']))
             elif line.startswith('- ') or line.startswith('* '):
-                story.append(Paragraph(f"• {line[2:]}", styles['Normal']))
+                story.append(ReportExporter.Paragraph(f"• {line[2:]}", styles['Normal']))
             elif line == '---' or line.startswith('==='):
-                story.append(Spacer(1, 0.3*inch))
-                story.append(Paragraph("_" * 70, styles['Normal']))
-                story.append(Spacer(1, 0.3*inch))
+                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(Paragraph(clean_line, styles['Normal']))
-        
+                story.append(ReportExporter.Paragraph(clean_line, styles['Normal']))
+
         # Construir PDF
         doc.build(story)
         return filename
 
+
 class AIAnalyzer:
-    """Clase para análisis con IA"""
-    
+    """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', 
+                '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
-        
-        # Ejemplo de prompt mejorado:
-        prompt = f"""
-        {lang_prefix}
-        Based on the following data:
-        {data.to_string()}
-        
-        Perform a comparative analysis of mathematical models. Include:
-        1. Best model for each variable type (biomass, substrate, product)
-        2. Comparison metrics (R², RMSE, AIC, BIC)
-        3. Justification for model selection
-        4. Implementation code with embedded data values
+
+        # 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.
         """
-        
-        response = self.client.chat.completions.create(
-            model="Qwen/Qwen3-14B",
-            messages=[{"role": "user", "content": prompt}],
-            max_tokens=4096
-        )
-    
+
+        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 in English. ",
-            'es': "Por favor responde en español. ",
-            'fr': "Veuillez répondre en français. ",
-            'de': "Bitte antworten Sie auf Deutsch. ",
-            'pt': "Por favor responda em português. "
+            '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, claude_model: str, detail_level: str = "detailed", 
+
+    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"""
-        
-        # Preparar resumen completo de los datos
+
+        # Prepare comprehensive data summary for the model
         data_summary = f"""
-        FITTING RESULTS DATA:
-        
-        Data structure:
+        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)}
-        
-        Complete data:
-        {data.to_string()}
-        
-        Descriptive statistics:
-        {data.describe().to_string()}
         """
-        
-        # Extraer valores para usar en el código
-        data_dict = data.to_dict('records')
-        
-        # Obtener prefijo de idioma
+
+        # Get language prefix
         lang_prefix = self.get_language_prompt_prefix(language)
-        
-        # Agregar especificaciones adicionales del usuario si existen
+
+        # Add user additional specifications if they exist
         user_specs_section = f"""
-        
-        USER ADDITIONAL SPECIFICATIONS:
+
+        USER ADDITIONAL SPECIFICATIONS / FOCUS AREAS:
         {additional_specs}
-        
-        Please ensure to address these specific requirements in your analysis.
+
+        Please ensure your analysis incorporates these specific requirements and focus areas.
         """ if additional_specs else ""
-        
-        # Prompt mejorado con instrucciones específicas para cada nivel
+
+        # Prompt enhanced with specific instructions for each level
+        # Added system message for better role adherence
         if detail_level == "detailed":
-            prompt = f"""
-            {lang_prefix}
-            
-            You are an expert in biotechnology and mathematical modeling. Analyze these kinetic/biotechnological model fitting results.
-            
+            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 BY EXPERIMENT
-            
-            PERFORM A COMPREHENSIVE COMPARATIVE ANALYSIS PER EXPERIMENT:
-            
-            1. **EXPERIMENT IDENTIFICATION AND OVERVIEW**
-               - List ALL experiments/conditions tested (e.g., pH levels, temperatures, time points)
-               - For EACH experiment, identify:
-                 * Experimental conditions
-                 * Number of models tested
-                 * Variables measured (biomass, substrate, product)
-            
-            2. **MODEL IDENTIFICATION AND CLASSIFICATION BY EXPERIMENT**
-               For EACH EXPERIMENT separately:
-               - Identify ALL fitted mathematical models BY NAME
-               - Classify them: biomass growth, substrate consumption, product formation
-               - Show the mathematical equation of each model
-               - List parameter values obtained for that specific experiment
-            
-            3. **COMPARATIVE ANALYSIS PER EXPERIMENT**
-               Create a section for EACH EXPERIMENT showing:
-               
-               **EXPERIMENT [Name/Condition]:**
-               
-               a) **BIOMASS MODELS** (if applicable):
-                  - Best model: [Name] with R²=[value], RMSE=[value]
-                  - Parameters: μmax=[value], Xmax=[value], etc.
-                  - Ranking of all biomass models tested
-                  
-               b) **SUBSTRATE MODELS** (if applicable):
-                  - Best model: [Name] with R²=[value], RMSE=[value]
-                  - Parameters: Ks=[value], Yxs=[value], etc.
-                  - Ranking of all substrate models tested
-                  
-               c) **PRODUCT MODELS** (if applicable):
-                  - Best model: [Name] with R²=[value], RMSE=[value]
-                  - Parameters: α=[value], β=[value], etc.
-                  - Ranking of all product models tested
-            
-            4. **DETAILED COMPARATIVE TABLES**
-               
-               **Table 1: Summary by Experiment and Variable Type**
-               | Experiment | Variable | Best Model | R² | RMSE | Key Parameters | Ranking |
-               |------------|----------|------------|-------|------|----------------|---------|
-               | Exp1       | Biomass  | [Name]     | [val] | [val]| μmax=X         | 1       |
-               | Exp1       | Substrate| [Name]     | [val] | [val]| Ks=Y           | 1       |
-               | Exp1       | Product  | [Name]     | [val] | [val]| α=Z            | 1       |
-               | Exp2       | Biomass  | [Name]     | [val] | [val]| μmax=X2        | 1       |
-               
-               **Table 2: Complete Model Comparison Across All Experiments**
-               | Model Name | Type | Exp1_R² | Exp1_RMSE | Exp2_R² | Exp2_RMSE | Avg_R² | Best_For |
-            
-            5. **PARAMETER ANALYSIS ACROSS EXPERIMENTS**
-               - Compare how parameters change between experiments
-               - Identify trends (e.g., μmax increases with temperature)
-               - Calculate average parameters and variability
-               - Suggest optimal conditions based on parameters
-            
-            6. **BIOLOGICAL INTERPRETATION BY EXPERIMENT**
-               For each experiment, explain:
-               - What the parameter values mean biologically
-               - Whether values are realistic for the conditions
-               - Key differences between experiments
-               - Critical control parameters identified
-            
-            7. **OVERALL BEST MODELS DETERMINATION**
-               - **BEST BIOMASS MODEL OVERALL**: [Name] - performs best in [X] out of [Y] experiments
-               - **BEST SUBSTRATE MODEL OVERALL**: [Name] - average R²=[value]
-               - **BEST PRODUCT MODEL OVERALL**: [Name] - most consistent across conditions
-               
-               Justify with numerical evidence from multiple experiments.
-            
-            8. **CONCLUSIONS AND RECOMMENDATIONS**
-               - Which models are most robust across different conditions
-               - Specific models to use for each experimental condition
-               - Confidence intervals and prediction reliability
-               - Scale-up recommendations with specific values
-            
-            Use Markdown format with clear structure. Include ALL numerical values from the data.
-            Create clear sections for EACH EXPERIMENT.
+
+            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
-            prompt = f"""
-            {lang_prefix}
-            
-            You are an expert in biotechnology. Provide a CONCISE but COMPLETE analysis BY EXPERIMENT.
-            
+             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
-            
+
+            DETAIL LEVEL: SUMMARIZED - Be concise but include all experiments and essential information.
+
             PROVIDE A FOCUSED COMPARATIVE ANALYSIS:
-            
-            1. **EXPERIMENTS OVERVIEW**
-               - Total experiments analyzed: [number]
-               - Conditions tested: [list]
-               - Variables measured: biomass/substrate/product
-            
-            2. **BEST MODELS BY EXPERIMENT - QUICK SUMMARY**
-               
-               📊 **EXPERIMENT 1 [Name/Condition]:**
-               - Biomass: [Model] (R²=[value])
-               - Substrate: [Model] (R²=[value])  
-               - Product: [Model] (R²=[value])
-               
-               📊 **EXPERIMENT 2 [Name/Condition]:**
-               - Biomass: [Model] (R²=[value])
-               - Substrate: [Model] (R²=[value])
-               - Product: [Model] (R²=[value])
-               
-               [Continue for all experiments...]
-            
-            3. **OVERALL WINNERS ACROSS ALL EXPERIMENTS**
-               🏆 **Best Models Overall:**
-               - **Biomass**: [Model] - Best in [X]/[Y] experiments
-               - **Substrate**: [Model] - Average R²=[value]
-               - **Product**: [Model] - Most consistent performance
-            
-            4. **QUICK COMPARISON TABLE**
-               | Experiment | Best Biomass | Best Substrate | Best Product | Overall R² |
-               |------------|--------------|----------------|--------------|------------|
-               | Exp1       | [Model]      | [Model]        | [Model]      | [avg]      |
-               | Exp2       | [Model]      | [Model]        | [Model]      | [avg]      |
-            
-            5. **KEY FINDINGS**
-               - Parameter ranges across experiments: μmax=[min-max], Ks=[min-max]
-               - Best conditions identified: [specific values]
-               - Most robust models: [list with reasons]
-            
-            6. **PRACTICAL RECOMMENDATIONS**
-               - For biomass prediction: Use [Model]
-               - For substrate monitoring: Use [Model]
-               - For product estimation: Use [Model]
-               - Critical parameters: [list with values]
-            
-            Keep it concise but include ALL experiments and model names with their key metrics.
+
+            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/Qwen3-14B",  # o el modelo específico
-                messages=[{"role": "user", "content": prompt}],
-                max_tokens=3000,
-                temperature=0.7
+                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
             )
-            
-            # Análisis adicional para generar código con valores numéricos reales
-            code_prompt = f"""
-            {lang_prefix}
-            
-            Based on the analysis and this actual data:
-            {data.to_string()}
-            
-            Generate Python code that:
-            
-            1. Creates a complete analysis system with the ACTUAL NUMERICAL VALUES from the data
-            2. Implements analysis BY EXPERIMENT showing:
-               - Best models for each experiment
-               - Comparison across experiments
-               - Parameter evolution between conditions
-            3. Includes visualization functions that:
-               - Show results PER EXPERIMENT
-               - Compare models across experiments
-               - Display parameter trends
-            4. Shows the best model for biomass, substrate, and product separately
-            
-            The code must include:
-            - Data loading with experiment identification
-            - Model comparison by experiment and variable type
-            - Visualization showing results per experiment
-            - Overall best model selection with justification
-            - Functions to predict using the best models for each category
-            
-            Make sure to include comments indicating which model won for each variable type and why.
-            
-            Format: Complete, executable Python code with actual data values embedded.
+
+            # 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_response = self.client.messages.create(
-                model=claude_model,
-                max_tokens=3000,
-                messages=[{
-                    "role": "user",
-                    "content": code_prompt
-                }]
+            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": response.choices[0].message.content,
-                "codigo_implementacion": code_response.content[0].text,
+                "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() 
+                    "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_dict  # Incluir todos los datos para el código
+                    "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, claude_model: str, detail_level: str = "detailed", 
+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(client, model_registry)
+    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()
-        
-        with open(file.name, 'rb') as f:
-            file_content = f.read()
-        
-        if file_ext in ['.csv', '.xlsx', '.xls']:
-            if language == 'es':
-                results.append(f"## 📊 Análisis de Resultados: {file_name}")
+
+        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:
-                results.append(f"## 📊 Results Analysis: {file_name}")
-            
-            if file_ext == '.csv':
-                df = processor.read_csv(file_content)
+                 # 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:
-                df = processor.read_excel(file_content)
-            
-            if df is not None:
-                analysis_type = analyzer.detect_analysis_type(df)
-                
-                if analysis_type == AnalysisType.FITTING_RESULTS:
-                    result = analyzer.analyze_fitting_results(
-                        df, claude_model, detail_level, language, additional_specs
-                    )
-                    
-                    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"])
-        
-        results.append("\n---\n")
-    
+                 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)
-    code_text = "\n\n# " + "="*50 + "\n\n".join(all_code) if all_code else generate_implementation_code(analysis_text)
-    
+    # 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, data: pd.DataFrame) -> str:
-    """
-    Genera código de implementación con análisis por experimento y valores numéricos embebidos
-    """
-    # Extraer datos únicos por experimento
-    experiments = data['Experiment'].unique()
-    
-    # Iniciar construcción del código
-    code_parts = [
-        "\"\"\"\nSistema de Análisis de Modelos Biotecnológicos\nGenerado automáticamente con valores reales de datos\n\"\"\"",
-        "import numpy as np",
-        "import pandas as pd",
-        "import matplotlib.pyplot as plt",
-        "from scipy.optimize import curve_fit",
-        "from sklearn.metrics import r2_score, mean_squared_error",
-        ""
-    ]
-    
-    # Agregar datos embebidos
-    code_parts.append("# Datos embebidos con resultados de ajuste de modelos")
-    code_parts.append("model_data = {")
-    for col in data.columns:
-        code_parts.append(f"    '{col}': {data[col].tolist()},")
-    code_parts.append("}")
-    code_parts.append("df = pd.DataFrame(model_data)")
-    code_parts.append("")
-    
-    # Función para cargar datos
-    code_parts.append("def load_data():")
-    code_parts.append("    \"\"\"Carga los datos de ajuste de modelos\"\"\"")
-    code_parts.append("    print('Datos cargados con éxito. Estructura:')")
-    code_parts.append("    print(df.head())")
-    code_parts.append("    return df")
-    code_parts.append("")
-    
-    # Función para comparar modelos por experimento
-    code_parts.append("def compare_models_by_experiment(df):")
-    code_parts.append("    \"\"\"Compara modelos por experimento y tipo de variable\"\"\"")
-    code_parts.append("    results = {}")
-    code_parts.append("    for exp in df['Experiment'].unique():")
-    code_parts.append("        exp_data = df[df['Experiment'] == exp]")
-    code_parts.append("        best_models = {}")
-    code_parts.append("        for var_type in ['biomass', 'substrate', 'product']:")
-    code_parts.append("            # Filtrar por tipo de variable si existe")
-    code_parts.append("            var_data = exp_data[exp_data['Type'].str.lower() == var_type] if 'Type' in exp_data.columns else exp_data")
-    code_parts.append("            if not var_data.empty:")
-    code_parts.append("                best_idx = var_data['R2'].idxmax()")
-    code_parts.append("                best_model = var_data.iloc[best_idx]")
-    code_parts.append("                best_models[var_type] = {")
-    code_parts.append("                    'model': best_model['Model'],")
-    code_parts.append("                    'r2': best_model['R2'],")
-    code_parts.append("                    'rmse': best_model['RMSE'],")
-    code_parts.append("                    'params': {k: v for k, v in best_model.items() if k not in ['Model', 'R2', 'RMSE', 'Experiment', 'Type']}")
-    code_parts.append("                }")
-    code_parts.append("        results[exp] = best_models")
-    code_parts.append("    print('\\nModelos seleccionados por experimento:')")
-    code_parts.append("    for exp, models in results.items():")
-    code_parts.append("        print(f'\\n{exp}:')")
-    code_parts.append("        for var, info in models.items():")
-    code_parts.append("            print(f'  {var}: {info[\"model\"]}' + (f' (R²={info[\"r2\"]:.3f})' if info[\"r2\"] else ''))")
-    code_parts.append("    return results")
-    code_parts.append("")
-    
-    # Función para visualizar resultados
-    code_parts.append("def plot_results_by_experiment(df, results):")
-    code_parts.append("    \"\"\"Visualiza los resultados por experimento\"\"\"")
-    code_parts.append("    n_experiments = len(results)")
-    code_parts.append("    fig, axes = plt.subplots(nrows=1, ncols=n_experiments, figsize=(6*n_experiments, 5))")
-    code_parts.append("    if n_experiments == 1: axes = [axes]")
-    code_parts.append("    for ax, (exp, models) in zip(axes, results.items()):")
-    code_parts.append("        exp_data = df[df['Experiment'] == exp]")
-    code_parts.append("        exp_data.plot(kind='bar', x='Model', y='R2', ax=ax, title=f'{exp} - R² Comparison', legend=False)")
-    code_parts.append("        ax.set_ylabel('R² Score')")
-    code_parts.append("        ax.axhline(y=0.95, color='r', linestyle='--', label='Threshold (0.95)')")
-    code_parts.append("        ax.legend()")
-    code_parts.append("    plt.tight_layout()")
-    code_parts.append("    plt.show()")
-    code_parts.append("")
-    
-    # Ejemplo de uso
-    code_parts.append("if __name__ == '__main__':")
-    code_parts.append("    df = load_data()")
-    code_parts.append("    results = compare_models_by_experiment(df)")
-    code_parts.append("    plot_results_by_experiment(df, results)")
-    
-    return "\n".join(code_parts)
+
+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:
@@ -966,41 +1301,52 @@ def export_report(export_format: str, language: str) -> Tuple[str, str]:
             '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:
-        return f"Error: {str(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']),   # model_selector
+            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']),   # detail_level
+            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
@@ -1009,18 +1355,30 @@ def create_interface():
             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 archivos y almacena resultados"""
+        """Procesa files y almacena resultados"""
         if not files:
             error_msg = TRANSLATIONS[language]['error_no_files']
-            return error_msg, ""
-        
+            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():
@@ -1030,35 +1388,35 @@ def create_interface():
             with gr.Column(scale=1):
                 with gr.Row():
                     language_selector = gr.Dropdown(
-                        choices=[("English", "en"), ("Español", "es"), ("Français", "fr"), 
+                        choices=[("English", "en"), ("Español", "es"), ("Français", "fr"),
                                 ("Deutsch", "de"), ("Português", "pt")],
-                        value="en",
+                        value=current_language,
                         label=TRANSLATIONS[current_language]['select_language'],
                         interactive=True
                     )
                     theme_selector = gr.Dropdown(
                         choices=[("Light", "light"), ("Dark", "dark")],
-                        value="light",
+                        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", ".pdf", ".zip"],
-                    type="filepath"
+                    file_types=[".csv", ".xlsx", ".xls"], # Focusing on data files
+                    type="filepath" # Get file path as string
                 )
 
                 model_selector = gr.Dropdown(
-                    choices=list(CLAUDE_MODELS.keys()),
-                    value="Qwen/Qwen3-14B",  # ✅ Modelo válido
+                    choices=list(QWEN_MODELS.keys()),
+                    value=default_qwen_model_key,
                     label=TRANSLATIONS[current_language]['select_model'],
-                    info=f"{TRANSLATIONS[current_language]['best_for']}: {CLAUDE_MODELS['Qwen/Qwen3-14B']['best_for']}"
+                    info=default_model_info_str
                 )
-                
+
                 detail_level = gr.Radio(
                     choices=[
                         (TRANSLATIONS[current_language]['detailed'], "detailed"),
@@ -1067,7 +1425,7 @@ def create_interface():
                     value="detailed",
                     label=TRANSLATIONS[current_language]['detail_level']
                 )
-                
+
                 # Nueva entrada para especificaciones adicionales
                 additional_specs = gr.Textbox(
                     label=TRANSLATIONS[current_language]['additional_specs'],
@@ -1076,83 +1434,84 @@ def create_interface():
                     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
-            - **Model**: Model name
-            - **Type**: Variable type (Biomass/Substrate/Product)
-            - **R2, RMSE**: Fit quality metrics
-            - **Parameters**: Model-specific parameters
+            - **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
+
+        # Definir ejemplos (Update example paths if necessary)
         examples = gr.Examples(
             examples=[
-                [["examples/biomass_models_comparison.csv"], "Qwen/Qwen3-14B", "detailed", ""],
-                [["examples/substrate_kinetics_results.xlsx"], "Qwen/Qwen3-14B", "summarized", "Focus on temperature effects"]
+                [["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,
@@ -1160,55 +1519,95 @@ def create_interface():
             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, 
+                analyze_btn, export_format, export_btn, analysis_output,
                 code_output, data_format_accordion
             ]
         )
-        
+
         def change_theme(theme_name):
             """Cambia el tema de la interfaz"""
-            # Nota: En Gradio actual, cambiar el tema dinámicamente requiere recargar
-            # Esta es una limitación conocida
-            return gr.Info("Theme will be applied on next page load")
-        
+            # 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=[]
+            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)
-            if file:
-                return gr.update(value=status, visible=True), gr.update(value=file, visible=True)
+            # 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:
-                return gr.update(value=status, visible=True), gr.update(visible=False)
-        
+                # 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,
-            debug=True,
             share=False
         )