Bound llama.cpp HTTP and compute threads; enable four slots d861a68 verified ShoaibSSM commited on 11 days ago