use pre-installed MiniLM (no cache, no internet)

rag.py

@@ -103,23 +103,15 @@ def get_texts() -> List[str]:
         print(f"⚠ Dataset fetch failed: {e} – using empty corpus")
         return []
 
+# ------------------------------------------------------------------
 # ------------------------------------------------------------------
 @lru_cache(maxsize=1)
 def get_vectorstore() -> FAISS:
     texts = get_texts()
 
-    # --- FINAL: load local MiniLM (no internet, no cache) -----------------
-    import os
-    local_model_path = os.path.abspath(
-        os.path.join(os.path.dirname(__file__), "st_model")
-    )
-
-    # force transformers to read local files only
-    os.environ["TRANSFORMERS_OFFLINE"] = "1"
-    os.environ["HF_DATASETS_OFFLINE"] = "1"
-
+    # --- FINAL: use pre-installed MiniLM (no cache, no internet) ----------
     from sentence_transformers import SentenceTransformer
-    model = SentenceTransformer(local_model_path, device="cpu", cache_folder=None)
+    model = SentenceTransformer("all-MiniLM-L6-v2", device="cpu", cache_folder=None)
 
     from langchain.embeddings import SentenceTransformerEmbeddings
     embeddings = SentenceTransformerEmbeddings(model=model)