- Update header to show base URL as a separate element - Include memory budget details in context length info in Settings - Add Hugging Face link to model context menu - Remove --no-mmap to enable memory mapping - Update llama.cpp to b7652 - Fix Nemotron KV cache footprint calculation