use local model files in  cache dir

Try to deploy on an internal network, so I manually downloaded the embedding model files from the external network and placed them under cache/model.
However, the system still reports a model download failure on startup.
Hope to use the local model files instead of attempting to download them every time