Comfy-Launcher-MultiOS

A single Bash script (comfy.sh) to install, launch, update, and manage ComfyUI on Linux and macOS — with multi-OS support, Python version selection, HTTPS, saved launch profiles, and a clean terminal interface.

---

✨ Features

• 🎛️ All-in-one menu — one script, one interactive menu for every action
• 🖥️ Multi-OS support — Fedora/RHEL, Arch Linux, Ubuntu/Debian, macOS (auto-detected)
• 🐍 Python version selection — Choose between 3.12, 3.13, or 3.14 at install time
• ⚡ Automatic CUDA detection — PyTorch installed with the right backend for your hardware
• 🔀 Python switcher — Switch the venv Python version without reinstalling ComfyUI
• 🔐 HTTPS launch — Optional self-signed certificate, generated automatically
• 🗂️ Saved launch profiles — Save addresses (e.g. a Tailscale IP) once, relaunch in one keystroke
• 📊 System info dashboard — Full diagnostic display in a clean terminal UI

---

🚀 Quick Start

1. Clone this repository

git clone https://github.com/Black0S/Comfy-Launcher-MultiOS
cd Comfy-Launcher-MultiOS

2. Make the script executable

chmod +x comfy.sh

3. Run it

./comfy.sh

That's it — everything happens through the interactive menu:

  ╔══════════════════════════════════════════════════╗
  ║          Comfy-Launcher  —  comfy.sh             ║
  ╚══════════════════════════════════════════════════╝

    1) 📦  Install / Reinstall ComfyUI
    2) 🚀  Launch ComfyUI  (HTTP / HTTPS / custom address)
    3) 🔄  Update ComfyUI
    4) 🔥  Update PyTorch
    5) 🐍  Switch Python version
    6) 📊  System info
    0) 🚪  Quit

---

📦 Installing (menu → 1)

Your operating system is auto-detected (Fedora/RHEL, Arch, Ubuntu/Debian, macOS). You will then be prompted to select your Python version (3.12 / 3.13 / 3.14).

If auto-detection fails on an unusual distro, the installer falls back to a manual OS menu.

The installer will:

• Check and install git if missing
• Check and install the selected Python version if missing
• Clone ComfyUI and ComfyUI-Manager into ./comfyui
• Create a virtual environment (comfyui/.venv)
• Install PyTorch with the right backend (CUDA on Linux, MPS nightly on macOS)
• Install all ComfyUI dependencies

Older CUDA driver? PyTorch ships from the cu130 (CUDA 13.0) wheel channel by default. If your driver is too old, override it before launching:

TORCH_CUDA_CHANNEL=cu128 ./comfy.sh

This applies to Install, Switch Python, and Update PyTorch.

---

🚀 Launching (menu → 2)

Choose how to serve ComfyUI:

Mode	Result
Local	http://127.0.0.1:8188 (default)
Custom address	Choose IP + port over HTTP (e.g. 0.0.0.0:8080 to expose on your LAN)
Local HTTPS	https://127.0.0.1:8188 with an auto-generated self-signed certificate
Custom HTTPS	Choose IP + port over HTTPS (self-signed)

You can also add extra launch flags (e.g. --lowvram, --use-split-cross-attention) when configuring a launch.

The self-signed certificate is created once with openssl into ./certs/ (git-ignored). Your browser will show a one-time warning you can accept.

Saved launch profiles

No more retyping addresses. When you create a launch configuration, you can save it as a named profile (e.g. tailscale, lan, local). On the next launch the menu lists your profiles — just pick a number to start instantly:

  Saved profiles (pick a number to launch)
    1) tailscale      →  https://100.64.1.5:8188
    2) lan            →  http://<this-machine-IP>:8080

  n) New launch configuration
  e) Edit / delete a saved profile
  0) Back to main menu

• Select — type the profile number to launch with no further questions.
• Create — choose n, configure address/HTTPS/extra flags, and optionally give it a name to save.
• Manage — choose e to edit (address, port, HTTPS, flags) or delete a profile.

Each profile stores the listen address, port, HTTPS on/off, and any extra flags. Profiles live in profiles.conf (git-ignored, since it may contain private addresses like a Tailscale IP).

Tailscale tip: a profile pointing at your Tailscale IP lets you reach ComfyUI from any device on your tailnet. The self-signed cert will still trigger a browser warning (the cert is issued for localhost); for a trusted certificate, consider tailscale cert / tailscale serve instead.

⚠️ Security: ComfyUI has no authentication. Listening on anything other than 127.0.0.1 exposes the full UI — and code execution via custom nodes — to your network. Only do this on a trusted network, or behind an authenticated reverse proxy. comfy.sh warns you when the listen address is not local.

---

🐍 Python Version Guide

Version	Status	Notes
3.12	⚠️ Older	Maximum compatibility with custom nodes
3.13	✅ Recommended	Compatible with the vast majority of custom nodes
3.14	🧪 Experimental	Some custom nodes may not work

Switching Python version (menu → 5)

Switch the Python version used by ComfyUI without reinstalling everything from scratch. The script will:

• Display the currently active Python version (highlighted)
• Let you choose a new version, installing it if missing
• Delete the old virtual environment and recreate a clean one
• Reinstall PyTorch and all ComfyUI dependencies automatically

---

🏗️ Arch Linux — Note on Python installation

Arch Linux only packages a single version of Python in its official repos. To support multiple specific versions (3.12, 3.13, 3.14), the installer uses pyenv, installed via your AUR helper.

Requirements:

• yay or paru must be installed before running comfy.sh on Arch

The installer will:

1. Install all required build dependencies via pacman
2. Install pyenv via your AUR helper
3. Automatically configure pyenv in your .bashrc or .zshrc
4. Resolve and install the latest patch version of your chosen Python
5. Create the venv using that exact Python binary

---

🔄 Updating

• Update ComfyUI (menu → 3) — Pulls the latest commits from the ComfyUI repository and updates all venv dependencies. (ComfyUI-Manager updates itself from within the ComfyUI interface.)
• Update PyTorch (menu → 4) — Reinstalls the latest version of PyTorch in the venv, for your OS and hardware.

---

📊 System Info (menu → 6)

Displays a full diagnostic dashboard in the terminal:

• 🖥️ System — OS, kernel, CPU, RAM
• ⚡ GPU / CUDA — GPU name, VRAM, driver version, CUDA toolkit (or MPS on macOS)
• 🐍 Python — All versions installed on the system (including pyenv), with the ComfyUI venv version highlighted
• 🔥 PyTorch — Version, CUDA/MPS availability, GPU seen by PyTorch
• 🎨 ComfyUI — Branch, commit, date, ComfyUI-Manager status

---

📋 Requirements

Linux

• bash 4.0+
• A package manager: dnf, pacman, or apt
• Arch only: yay or paru (AUR helper) must be installed
• NVIDIA GPU with drivers installed (recommended)
• CUDA runtime (optional — PyTorch works without nvcc)

macOS

• bash 3.2+ (the system default works; the script is tested on it)
• Homebrew
• openssl (preinstalled) for HTTPS
• Apple Silicon recommended for MPS acceleration

---

📂 Directory Structure

After installation, the following structure exists next to the script:

.
├── comfy.sh                  ← the only script (interactive menu)
├── certs/                    ← self-signed TLS certificate (generated, git-ignored)
├── profiles.conf             ← saved launch profiles (generated, git-ignored)
└── comfyui/                  ← ComfyUI repository (cloned by the installer)
    ├── .venv/                ← Python virtual environment
    ├── main.py
    ├── requirements.txt
    └── custom_nodes/
        └── comfyui-manager/  ← ComfyUI-Manager

---

🤝 Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

---

📄 License

This project is released under the MIT License.

---

ComfyUI is developed by @comfyanonymous — this project is an independent launcher and is not officially affiliated with ComfyUI.