feat(backend): add buun-llama-cpp fork (DFlash + TCQ KV-cache) - #9532

Open

mudler wants to merge 10 commits into

feat/buun-llama-cpp-backend

Commits on Jul 31, 2026

feat(backend): add buun-llama-cpp fork (DFlash + TCQ KV-cache)

mudler
authored and
localai-org-maint-bot
committed
test(gallery): extend importer specs to cover buun-llama-cpp

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): backport logit_bias_eog field to grpc-server copy

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): drop logit_bias_eog arg from params_from_json_cmpl

mudler
authored and
localai-org-maint-bot
committed
ci(buun-llama-cpp): wire backend into test-extra + build matrix

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): shim atomicAdd(double*,double) for pre-sm_60 CUDA

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): pass WARP_SIZE to argmax __shfl_xor_sync calls

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): shim cudaMemcpy{To,From}Symbol + WARP_SIZE on fwht128 shuffles

mudler
authored and
localai-org-maint-bot
committed
fix(buun-llama-cpp): match speculative draft split anchor
localai-org-maint-bot
committed
fix(buun-llama-cpp): isolate fork patch series
localai-org-maint-bot
committed