UPSTREAM PR #20428: server: fix multi-turn cache reuse for hybrid/recurrent models#1248

loci-dev wants to merge 2 commits intomainauroralabs-loci/llama.cpp:mainfrom

loci/pr-20428-hybrid-cache-reuseauroralabs-loci/llama.cpp:loci/pr-20428-hybrid-cache-reuseCopy head branch name to clipboard