dotfiles

My personal dotfiles for zsh.

• 🧙‍ Leverages antidote for Zsh plugin management:
  • zsh-completions: Additional completion definitions.
  • powerlevel10k: Fast and flexible theme.
  • zsh-autosuggestions: As-you-type command suggestions.
  • zsh-syntax-highlighting: Command syntax highlighting.
• 👻 Optimized for the ghostty terminal emulator.
• 🐍 Uses uv for lightning-fast Python management.
• 💎 Manages environments for nodenv and rbenv.
• 🤵 Tailored for macOS with surgical defaults write configuration in script/macos.
• ✅ Validated with shellcheck.

Installation

Run the bootstrap script:

./script/bootstrap

This will:

1. Install Homebrew (if missing).
2. Install dependencies via brew bundle (using local Brewfile).
3. Set your shell to zsh.
4. Symlink configuration files.
5. Apply macOS system preferences surgically.
6. Initialize the local LLM environment.

Local LLM Workflow

This project is optimized for local agentic coding on Apple Silicon (M5 Max recommended) using llama.cpp and aider.

🚀 Getting Started

1. Initialize: The LLM environment is set up during ./script/bootstrap. You can manually trigger it with:

   ./script/llm setup

2. Download Model: Retrieve the optimized Qwen3-Coder-Next-Q4_K_M model (~48GB split across 4 GGUF shards) from the official GGUF repo:

   ./script/llm download

3. Start Server: Launch the llama.cpp server with full GPU acceleration:

   llm-start

4. Launch Agent: Start a coding session in any repository:

   aider

🛠️ Tools

• llama.cpp: Efficient inference engine for Apple Silicon.
• aider: CLI coding agent that works with your local model.
• Qwen3-Coder-Next-80B: High-performance local Mixture of Experts model for coding tasks.
• script/llm: Management utility for your local LLM environment.