Route macOS at unslothai/llama.cpp instead of ggml-org/llama.cpp and drop --simple-policy for Darwin, so macOS installs read the fork's published manifest (the macos-arm64 / macos-x64 bundles built with deployment target 13.3) instead of upstream filenames. Upstream's arm64 release binaries are currently stamped minos=26 and fail to load on macOS < 26; the fork bundles load on macOS 13.3+. Non-macOS routing and setup.ps1 are unchanged, and the installer still falls back to upstream and then a source build if the fork lacks a compatible asset. Add a consumer-contract test that the fork's manifest + checksum schema parses and selects both macOS slices. |
||
|---|---|---|
| .. | ||
| backend | ||
| frontend | ||
| src-tauri | ||
| __init__.py | ||
| install_llama_prebuilt.py | ||
| install_python_stack.py | ||
| LICENSE.AGPL-3.0 | ||
| package-lock.json | ||
| package.json | ||
| setup.bat | ||
| setup.ps1 | ||
| setup.sh | ||
| Unsloth_Studio_Colab.ipynb | ||