okay

rag.py

@@ -104,20 +104,14 @@ def get_texts() -> List[str]:
         return []
 
 ## ------------------------------------------------------------------------------rtutu
-# ------# --------# ------------------------------------------------------------------
+# ------------------------------------------------------------------
 @lru_cache(maxsize=1)
 def get_vectorstore() -> FAISS:
     texts = get_texts()
 
-    # --- FINAL: load model from repo folder (no internet, no cache) --------
-    import os
-    local_model_path = os.path.join(os.path.dirname(__file__), "st_model")
-
-    from sentence_transformers import SentenceTransformer
-    model = SentenceTransformer(local_model_path, device="cpu", cache_folder=None)
-
-    from langchain.embeddings import SentenceTransformerEmbeddings
-    embeddings = SentenceTransformerEmbeddings(model=model)
+    # --- FINAL: use optimum ONNX MiniLM (already on disk in image) --------
+    from langchain.embeddings import OptimumEmbeddings
+    embeddings = OptimumEmbeddings.from_pretrained("optimum/all-MiniLM-L6-v2")
     # ------------------------------------------------------------------------
 
     if not texts:                       # no data → empty FAISS