Preparar repositorio para despliegue: código fuente limpio

utils/helpers.py

@@ -13,8 +13,14 @@ def safe_html(texto: Optional[str]) -> str:
         return ""
     
     # Sanitize content to prevent layout breakage (e.g. unclosed divs)
-    allowed_tags = ['b', 'i', 'strong', 'em', 'p', 'br', 'span', 'a']
-    allowed_attrs = {'a': ['href', 'target', 'rel']}
+    allowed_tags = [
+        'b', 'i', 'strong', 'em', 'p', 'br', 'span', 'a', 'img',
+        'h1', 'h2', 'h3', 'h4', 'ul', 'ol', 'li', 'blockquote'
+    ]
+    allowed_attrs = {
+        'a': ['href', 'target', 'rel', 'title'],
+        'img': ['src', 'alt', 'title', 'width', 'height', 'style']
+    }
     
     cleaned = bleach.clean(texto, tags=allowed_tags, attributes=allowed_attrs, strip=True)
     return Markup(cleaned)

utils/qdrant_search.py

@@ -6,7 +6,7 @@ import os
 import time
 from typing import List, Dict, Any, Optional
 from qdrant_client import QdrantClient
-from sentence_transformers import SentenceTransformer
+# from sentence_transformers import SentenceTransformer (Moved to function)
 
 # Configuración
 QDRANT_HOST = os.environ.get("QDRANT_HOST", "localhost")
@@ -16,7 +16,7 @@ EMB_MODEL = os.environ.get("EMB_MODEL", "sentence-transformers/paraphrase-multil
 
 # Singleton para clientes globales
 _qdrant_client: Optional[QdrantClient] = None
-_embedding_model: Optional[SentenceTransformer] = None
+_embedding_model: Optional[Any] = None
 
 
 def get_qdrant_client() -> QdrantClient:
@@ -40,12 +40,13 @@ def get_qdrant_client() -> QdrantClient:
     return _qdrant_client
 
 
-def get_embedding_model() -> SentenceTransformer:
+def get_embedding_model() -> Any:
     """
     Obtiene el modelo de embeddings (singleton).
     """
     global _embedding_model
     if _embedding_model is None:
+        from sentence_transformers import SentenceTransformer
         _embedding_model = SentenceTransformer(EMB_MODEL, device='cpu')
     return _embedding_model