bug fixed; code polished; win amd64 (python 3.8.x-3.13.x supported), mac x86 (intel) and arm64 (silicon) binary packs provided in this release
0.3.9
reconstruct the prompt engine and wheels (for 3.10-3.13 python versions); win amd64, mac x86 (intel) and arm64 (silicon) binary pack provided in this release
0.3.8
trim up; code polished; win amd64, mac x86 (intel) & arm64 (silicon/metal) wheels plus llama_core-0.3.8.tar.gz (source code) included in this pack
0.3.7
descriptor updated; with brief instruction on how to build from source file for device(s) with Metal and/or CUDA; overall code scheme revised; win amd64, mac x86 & arm64 (metal) wheels plus llama_core-0.3.7.tar.gz (source code) included in this release
0.3.6
new engine implemented; win amd64, mac x86 & arm64 (metal) wheels plus llama_core-0.3.6.tar.gz (source code) included in this pack
0.3.5
second attempt; new engine on test; metal & x86 wheels for mac included; if you cannot build the wheel for win, can take version 0.3.3 (pip install llama-core==0.3.3) instead right away