Skip to main content

Install and Backends

SLMCortex requires Python 3.11+.

On macOS and Linux, install the base CLI with Homebrew first:

brew tap alvarolorentedev/SLMCortex
brew install slmcortex

This installs both public launchers:

slmcortex --help
slmcortex-composer --help

Source Checkout Alternative​

Use a source checkout only when you need repo-local demo scripts, editable development, or direct access to the checked-in fixtures:

python3 -m venv .venv
. .venv/bin/activate
pip install --upgrade pip
pip install -e '.[test]'

This gives you the public CLI plus the repo-local no-model demo flow without forcing a real inference backend.

Optional Real-Model Backends​

Install one backend only when you need actual local training or inference:

pip install -e '.[mlx]'   # macOS Apple Silicon
pip install -e '.[gguf]'  # Linux, Windows, macOS Intel, or explicit GGUF use

Backend selection comes from base config:

  • backend: auto chooses MLX on macOS arm64/aarch64
  • backend: auto chooses GGUF everywhere else
  • explicit backend: mlx is rejected outside macOS Apple Silicon
  • GGUF configs must use a .gguf runtime model path
  • GGUF training or import conversion needs gguf_converter

Practical Guidance​

Start with the base install and the no-model demo. Add a backend only when:

  • infer must stop being dry-run
  • serve must answer with real model output
  • train-slm or adaptive training must run for real

That keeps setup problems separate from product-flow problems.