Update app.py

app.py

@@ -294,51 +294,70 @@ def extract_and_infer_with_gemini(query, condiciones):
         return None
 def find_best_matches_hybrid(entities, data):
     if not entities or not data: return []
+    
+    # 1. Extracción de entidades del usuario (sin cambios)
     user_symptoms = set(sanitize_text(s) for s in entities.get("sintomas", []))
     user_foods = set(sanitize_text(f) for f in entities.get("alimentos", []))
     inferred_condition_raw = sanitize_text(entities.get("condicion_probable", ""))
+    
+    # 2. Obtención de términos de búsqueda (sin cambios)
     candidate_terms = set(user_foods)
     for food in user_foods:
         food_sanitized = sanitize_text(food)
         if food_sanitized in FOOD_TO_COMPOUND_MAP:
             candidate_terms.update(c.lower() for c in FOOD_TO_COMPOUND_MAP[food_sanitized])
-    target_data = data
-    if inferred_condition_raw:
-        filtered_data = []
-        for entry in data:
-            entry_condition = sanitize_text(entry.get("condicion_asociada", ""))
-            if entry_condition == inferred_condition_raw:
-                filtered_data.append(entry)
-                continue
-            if entry_condition in CONDITION_SYNONYMS:
-                if inferred_condition_raw in [sanitize_text(s) for s in CONDITION_SYNONYMS[entry_condition]]:
-                    filtered_data.append(entry)
-        if filtered_data:
-            target_data = filtered_data
+
+    # 3. Lógica de puntuación MEJORADA (el cambio clave está aquí)
     results = []
-    for entry in target_data:
+    # ¡Importante! Siempre iteramos sobre 'data' (la base de datos completa)
+    for entry in data:
         score_details = {'condition': 0, 'food': 0, 'symptoms': 0, 'total': 0}
-        matched_symptoms = []
-        if inferred_condition_raw and sanitize_text(entry.get("condicion_asociada", "")) == inferred_condition_raw:
-            score_details['condition'] = 100
+        
+        # Bono de Condición: Si la entrada coincide con la inferencia de la IA, recibe un gran bono.
+        if inferred_condition_raw:
+            entry_condition_sanitized = sanitize_text(entry.get("condicion_asociada", ""))
+            is_match = (entry_condition_sanitized == inferred_condition_raw)
+            # Comprobar también sinónimos
+            if not is_match and entry_condition_sanitized in CONDITION_SYNONYMS:
+                if inferred_condition_raw in [sanitize_text(s) for s in CONDITION_SYNONYMS[entry_condition_sanitized]]:
+                    is_match = True
+            if is_match:
+                score_details['condition'] = 100
+
+        # Puntuación por Alimento/Compuesto (sin cambios)
         entry_compounds_text = sanitize_text(entry.get("compuesto_alimento", ""))
         if any(term in entry_compounds_text for term in candidate_terms):
             score_details['food'] = 15
+            
+        # Puntuación por Síntomas (sin cambios)
         entry_symptoms_keys = set(sanitize_text(s) for s in entry.get("sintomas_clave", []))
         symptom_score = 0
+        matched_symptoms = []
         for user_symptom in user_symptoms:
             for key in entry_symptoms_keys:
                 if key in user_symptom or user_symptom in key:
                     symptom_score += 10
                     matched_symptoms.append(key)
-                    break
+                    break 
         score_details['symptoms'] = symptom_score
+        
+        # Cálculo final
         total_score = sum(score_details.values())
+        
+        # Solo añadimos resultados que tengan alguna puntuación
         if total_score > 0:
-            results.append({'entry': entry, 'score': score_details, 'matched_symptoms': list(set(matched_symptoms))})
+            results.append({
+                'entry': entry, 
+                'score': score_details, 
+                'matched_symptoms': list(set(matched_symptoms))
+            })
+
     if not results: return []
+    
+    # 4. Ordenar y devolver (sin cambios)
     sorted_results = sorted(results, key=lambda x: x['score']['total'], reverse=True)
     return sorted_results
+
 def generate_detailed_analysis(query, match):
     if not model: return "Error: El modelo de IA no está disponible."
     prompt_parts = [