Skip to content

Latest commit

 

History

History
292 lines (187 loc) · 9.15 KB

README.md

File metadata and controls

292 lines (187 loc) · 9.15 KB
Raw

ComfyUI Desktop (beta)

codecov Beta

Currently Supported Platforms

  • Windows (NVIDIA)
  • macOS (Apple silicon)

Overview

This desktop app is a packaged way to use ComfyUI and comes bundled with a few things:

On startup, it will install all the necessary python dependencies with uv and start the ComfyUI server. The app will automatically update with stable releases of ComfyUI, ComfyUI-Manager, and the uv executable as well as some desktop specific features.

Users please read the guide. Developers, read on.

Installed Files

Electron

The desktop application comes bundled with:

  • ComfyUI source code
  • ComfyUI-Manager
  • Electron, Chromium binaries, and node modules

Windows

We use the NSIS installer for Windows and it will install files in these locations:

Bundled Resources: %APPDATA%\Local\Programs\comfyui-electron

screenshot of resources directory

User files are stored here: %APPDATA%\ComfyUI

Automatic Updates: %APPDATA%\Local\comfyui-electron-updater

macOS

The macOS application is distributed as a DMG and will install files in:

~/Library/Application Support/ComfyUI

The application will be dragged into /Applications

Linux

~/.config/ComfyUI

ComfyUI

ComfyUI will also write files to disk as you use it. You will also be asked to select a location to store ComfyUI files like models, inputs, outputs, custom_nodes and saved workflows.

An extra_model_config.yaml is created to tell ComfyUI where to look for these files. You can edit this file to do things like add additional model files to the search path.

On Windows: %APPDATA%\ComfyUI\extra_model_config.yaml

On macOS: ~/Library/Application Support/ComfyUI/extra_model_config.yaml

On Linux: ~/.config/ComfyUI/extra_model_config.yaml

Logs

We use electron-log to log everything. Electron main process logs are in main.log, and ComfyUI server logs are in comfyui_<date>.log.

on Linux: ~/.config/{app name}/logs
on macOS: ~/Library/Logs/{app name}
on Windows: %AppData%\{app name}\logs

Development

Setup Python

Make sure you have python 3.12+ installed. It is recommended to setup a virtual environment.

Linux/MacOS:

python -m venv venv
source venv/bin/activate

Windows:

py -3.12 -m venv venv
.\venv\Scripts\Activate.ps1

Windows

Visual Studio

Visual studio 2019 or later with the Desktop C++ workload is required for node-gyp. See the node-gyp windows installation notes. Also requires the spectre-mitigated libraries, found in the individual components section of the VS installer.

Confirmed working:

  • Visual Studio Community 2022 - 17.12.1
  • Desktop development with C++ workload
  • MSVC v143 x64 spectre-mitigated libraries (Latest / v14.42-17.12)
    • Open the Visual Studio Installer
    • Click "Modify" on your Visual Studio 2022 Community installation
    • Go to the "Individual Components" tab
    • Search for "Spectre"
    • Check the boxes for the Spectre-mitigated libraries that match your project's architecture (x86 and/or x64)
    • image

Look for "MSVC v143 - VS 2022 C++ x64/x86 Spectre-mitigated libs" If you're using other toolsets, you may need their corresponding Spectre-mitigated libraries as well

NPM Dependencies

Node

We recommend using nvm to manage node versions. This project uses node v20.x.

Windows

Microsoft recommends nvm-windows on their Node.js on Windows page.

nvm install 20
nvm use 20

Yarn

This project uses yarn as its package manager. If you do not already have a yarn binary available on your PATH, run:

# corepack is a set of utilities included with all recent distributions of node
corepack enable
yarn set version v4.5.0 # Look at the packageManager key in package.json for the exact version.

This will install a usable yarn binary. Then, in the root directory of this repo (ie adjacent to the top-level package.json file), run:

yarn install

ComfyUI Assets

Before you can start the electron application, you need to download the ComfyUI source code and other things that are usually bundled with the application. We use comfy-cli to install everything.

Install comfy-cli

With the python environment activated, install comfy-cli:

pip install -U comfy-cli

ComfyUI and other dependencies

First, initialize the application resources by running make:assets:<gpu>:

# populate the assets/ dir (Installs a fresh ComfyUI instance under assets/)
yarn make:assets:[amd|cpu|nvidia|macos]

This command will install ComfyUI under assets, as well ComfyUI-Manager, and the frontend extension responsible for electron settings menu. The exact versions of each package is defined in package.json.

Second, you need to install uv. This will be bundled with the distributable, but we also need it locally.

yarn download:uv

You can then run start to build/launch the code and a live buildserver that will automatically rebuild the code on any changes:

deactivate # Deactivate your existing python env to avoid influencing the
yarn start

You can also build the package and/or distributables using the make command:

# build the platform-dependent package and any distributables
yarn make
# build cross-platform, e.g. windows from linux
yarn make --windows

Troubleshooting

If you get an error similar to:

The module '/electron/node_modules/node-pty/build/Release/pty.node' was compiled against a different Node.js version using NODE_MODULE_VERSION 115. This version of Node.js requires NODE_MODULE_VERSION 125. Please try re-compiling or re-installing the module (for instance, using `npm rebuild` or `npm install`).

You will need to rebuild the node-pty using electron-rebuild, for example:

npx electron-rebuild

or if that fails

yarn install -D @electron/rebuild
rm -rf node_modules
rm yarn.lock
yarn install
electron-rebuild

Missing libraries

You may get errors reporting that the build is unable to find e.g. libnss3.so if electron prerequisites are not included in your distro. Find the correct package for your distro and install.

apt example:

apt-get install libnss3

Debugger

There are helpful debug launch scripts for VSCode / Cursor under .vscode/launch.json. It will start the dev server as defined in .vscode/tasks.json. Then attach the debugger.

This can be used simply by pressing F5 in VSCode or VSCode derivative.

The launch environment can be customised, e.g. add a "linux" section to source your ~/.profile (and other interactive config) when debugging in linux:

{
  "version": "2.0.0",
  "tasks": [
    {
      "linux": { "options": { "shell": { "args": ["-ci"] } } }
    }
  ]
}

Release

We use Todesktop to build and codesign our distributables. To make a new release:

  1. Make a PR with the semantic version. eg. v1.0.1
  2. Add the label Release to the PR.
  3. Merge the PR
  4. A build will automatically start and you can view it at https://app.todesktop.com

Publish Locally

Follow the above steps for local development setup first.

# Authentication will be required.
yarn publish

Utility scripts

A number of utility scripts are defined under the "scripts" field of package.json. For example, to clean up the build artifacts you can run:

yarn clean

# Remove files created by yarn make:assets:<gpu>
yarn:clean:assets

# clean:slate also removes node_modules
yarn clean:slate

Crash Reports

At the onboarding step, you can opt-in to send us crash reports, which only include stack trace for unhandled exceptions and native crashes. These are really helpful to help us debug issues in lieu of an actual Github bug report. However, you can opt-out and nothing will ever be sent. We use Sentry to send crash reports. Code here.

In either case, no personal data, workflows or logs will be sent.

Download

Windows (NVIDIA) NSIS x64: Download

macOS ARM: Download