add cache for embeddings

retrieval.py

@@ -83,7 +83,7 @@ class Retrieval:
         
         # Инициализируем CrossEncoder
         # self.cross_encoder = CrossEncoder('DiTy/cross-encoder-russian-msmarco')
-        self.embedder = SentenceTransformer('cointegrated/rubert-tiny2')
+        self.embedder = SentenceTransformer('cointegrated/rubert-tiny2', cache_folder="/tmp")
         self.embeddings_of_summary = self.embedder.encode(self.paragraphs_df['summary'], convert_to_tensor=True)
 
         print("✅ RAG система готова к использованию")