use optimum ONNX MiniLM (no disk, no internet)

rag.py

@@ -105,21 +105,16 @@ def get_texts() -> List[str]:
 
 ## ------------------------------------------------------------------------------rtutu
 # ------------------------------------------------------------------
+# ------------------------------------------------------------------
 @lru_cache(maxsize=1)
 def get_vectorstore() -> FAISS:
     texts = get_texts()
 
-    # --- FINAL: load local ONNX MiniLM (no internet, no disk write) -------
-    import os
-    local_model_path = os.path.abspath(
-        os.path.join(os.path.dirname(__file__), "st_model")
+    # --- FINAL: use optimum ONNX MiniLM (already on disk) -----------------
+    from langchain_community.embeddings import OptimumSentenceEmbeddings
+    embeddings = OptimumSentenceEmbeddings(
+        model_name="optimum/all-MiniLM-L6-v2"
     )
-
-    from sentence_transformers import SentenceTransformer
-    model = SentenceTransformer(local_model_path, device="cpu", cache_folder=None)
-
-    from langchain.embeddings import SentenceTransformerEmbeddings
-    embeddings = SentenceTransformerEmbeddings(model=model)
     # ------------------------------------------------------------------------
 
     if not texts:                       # no data → empty FAISS