Update app.py

app.py

@@ -93,9 +93,14 @@ WRITABLE_BASE = _pick_writable_base()
 LOCAL_BASE = WRITABLE_BASE / "my_app_cache" / "index"
 LOCAL_BASE.mkdir(parents=True, exist_ok=True)
 # Recreate the SAME embedding model used to build the index
+try:
+    import torch
+    _emb_device = "cuda" if torch.cuda.is_available() else "cpu"
+except Exception:
+    _emb_device = "cpu"
 emb = HuggingFaceEmbeddings(
     model_name="google/embeddinggemma-300m",
-    model_kwargs={"device": "cuda"},
+    model_kwargs={"device": _emb_device},
     encode_kwargs={"normalize_embeddings": True},
 )