Skip to content

macOS Setup and Local AI Guide

This guide answers four things: - What is missing from your current setup - How to install everything quickly - What each tool is for and when to use it - What to install for local AI on a Mac mini M4 (16GB/256GB)

Current State (from scripts/dev-check.sh)

Core environment is healthy. Missing items are mostly optional productivity tools: - gh, wget - bat, eza, fd, fzf, zoxide, tlrc, httpie, git-delta - nvm, pnpm, code - starship, optional runtimes (go, rust)

Install Everything You Listed

brew install gh wget bat eza fd fzf zoxide tlrc httpie git-delta nvm pnpm starship

GUI app (if you want VS Code)

brew install --cask visual-studio-code

Optional language toolchains

brew install go rust

After installing nvm

mkdir -p ~/.nvm
source ~/.zshrc
nvm install --lts
nvm alias default 'lts/*'

Do Your Dotfiles Handle This?

Mostly yes: - Installers already include essentials and optional tools in os/mac/ - scripts/dev-check.sh checks health and now shows accurate required vs optional status - scripts/sync.sh handles symlinks and config refresh

What is not fully automatic yet: - Installing every optional GUI app by default (intentionally interactive/optional) - Local AI stack (recommended to keep optional because of disk/RAM tradeoffs)

Tool Cheat Sheet (What + When)

  • gh: GitHub from terminal. Use for PRs, issues, auth.
  • wget: quick file downloads, recursive/mirror-style fetch.
  • bat: syntax-highlighted file viewing.
  • eza: better ls with git status/icons.
  • rg (ripgrep): fast code/text search.
  • fd: fast file discovery.
  • fzf: fuzzy picker for history/files/branches.
  • zoxide: smart directory jumping (z project).
  • tlrc: concise examples for commands (maintained replacement for old tldr formula on Homebrew).
  • httpie: readable API calls (http GET ...).
  • jq: JSON parsing/transform.
  • git-delta: better colored git diffs.
  • nvm: manage multiple Node versions safely.
  • pnpm: fast, space-efficient Node package manager.
  • starship: cross-shell prompt.

Local AI on Mac mini M4 (16GB RAM, 256GB SSD)

For this hardware, prioritize efficient 7B/8B models and keep disk usage controlled.

  1. Ollama for model management and serving
  2. Open WebUI (or another lightweight UI) for chat UX
  3. Optional llama.cpp for low-level tuning/benchmarking

Install

brew install ollama
brew install llama.cpp

Start Ollama:

ollama serve

Pull practical models (start small):

ollama pull llama3.1:8b
ollama pull qwen2.5-coder:7b

About the tools you mentioned

  • Ollama: best default choice for local model lifecycle.
  • OpenClaw: UI option; use if you prefer this UX, but keep one UI to avoid bloat.
  • GPT4All: all-in-one desktop app; good for quick start, less flexible than CLI-first stack.
  • llama.cpp: best for advanced control, quantization experiments, benchmarking.
  • Core ML variants: useful for Apple Silicon optimization; good once baseline setup is stable.

Storage and Performance Advice for 16GB/256GB

  • Keep only 2-3 active models locally.
  • Prefer 7B/8B quantized models for responsiveness.
  • Avoid many 14B+ models on internal disk.
  • Consider external SSD for model cache if you expand model library.

Suggested Next Steps

  1. Install baseline CLI tools now (brew install ... above).
  2. Re-run scripts/dev-check.sh.
  3. Add Ollama + one coding model first (qwen2.5-coder:7b).
  4. Add a UI only if you actually need browser/chat workflows.