This is a template for secure electron applications. Written following the latest safety requirements, recommendations and best practices.
Under the hood is Vite — A next-generation blazing fast bundler, and electron-builder for packaging.
This repo was originally forked from an excellent starter by Alex Kozack that can be found here: vite-electron-builder
The focus of this repo is to add features that support development under WSL for builds targeting the Windows operating system. Issues pertaining to Linux and macOS are likely to be ignored and unknown. For the target audience, the workflow is roughly thus:
- Make changes.
- Test initially using a browser with
npm run dev
ornpm run dev:vite
. - Run
npm run dev:electron
and if you have an X server (i.e. VcXsrv) installed and running in the Windows OS then a window will pop open showing the Linux executable running out of WSL. Hot module replacement works so changes can be made debugged live in many cases. - Run
npm run compile:windows
to regenerate the files indist/win-unpacked
- Copy the
win-unpacked
folder outside of the WSL filesystem; desktop is fine. A script is even better. - Run the executable within the
win-unpacked\
folder you copied the files into.
One of the main changes in this version is that when compiling for Windows 10 or higher it utilizes MicaBrowserWindow
from GregVido/mica-electron. And, by default, MicaBrowserWindow is configured to apply the Acrylic frosted-glass effect to the electron window!
- Changed
npm run dev
for convenience intonpm run dev:vite
- Added
npm run dev:vite
to open a local Vite dev server - Added
npm run dev:electron
for what was previouslynpm run dev
- Add
npm run dev
to build and watch in one step - Add
npm run compile:windows
to build only for windows - Add support for MicaBrowserWindow and acrylic
- This template uses the latest electron version with all the latest security patches.
- The architecture of the application is built according to the security guides and best practices.
- The latest version of the electron-builder is used to package the application.
- Vite is used to bundle all source codes. It's an extremely fast bundler, that has a vast array of amazing features. You can learn more about how it is arranged in this video.
- Vite supports reading
.env
files. You can also specify the types of your environment variables intypes/env.d.ts
. - Automatic hot-reloads for the
Main
andRenderer
processes.
Vite provides many useful features, such as: TypeScript
, TSX/JSX
, CSS/JSON Importing
, CSS Modules
, Web Assembly
and much more.
- The latest version of TypeScript is used for all the source code.
- Vite supports TypeScript out of the box. However, it does not support type checking.
- Code formatting rules follow the latest TypeScript recommendations and best practices thanks to @typescript-eslint/eslint-plugin.
- By default, web pages are built using Vue. However, you can easily change that. Or not use additional frameworks at all.
- Code formatting rules follow the latest Vue recommendations and best practices thanks to eslint-plugin-vue.
The structure of this template is very similar to a monorepo. The entire source code of the project is divided into three modules (packages) that are each bundled independently:
packages/renderer
. Responsible for the contents of the application window. In fact, it is a regular web application. In developer mode, you can even open it in a browser. The development and build process is the same as for classic web applications. Access to low-level API electrons or Node.js is done through the preload layer.packages/preload
. Contain Electron preload scripts. Acts as an intermediate bridge between the renderer process and the API exposed by electron and Node.js. Runs in an isolated browser context, but has direct access to the full Node.js functionality.packages/main
Contain Electron main script. This is the main process that powers the application. It manages creating and handling the spawned BrowserWindow, setting and enforcing secure permissions and request handlers. You can also configure it to do much more as per your need, such as: logging, reporting statistics and health status among others.
Schematically, the structure of the application and the method of communication between packages can be depicted as follows:
flowchart TB;
packages/preload <-. IPC Messages .-> packages/main
subgraph packages/main["packages/main (Shared beatween all windows)"]
M[index.ts] --> EM[Electron Main Process Modules]
M --> N2[Node.js API]
end
subgraph Window["Browser Window"]
subgraph packages/preload["packages/preload (Works in isolated context)"]
P[index.ts] --> N[Node.js API]
P --> ED[External dependencies]
P --> ER[Electron Renderer Process Modules]
end
subgraph packages/renderer
R[index.html] --> W[Web API]
R --> BD[Bundled dependencies]
R --> F[Web Frameforks]
end
end
packages/renderer -- Call Exposed API --> P
The main
and preload
packages are built in library mode as it is
simple javascript.
The renderer
package builds as a regular web app.
The next step is to package a ready to distribute Electron app for macOS, Windows and Linux with "auto update" support out of the box.
To do this, use electron-builder:
- Using the npm script
compile
: This script is configured to compile the application as quickly as possible. It is not ready for distribution, it is compiled only for the current platform and is used for debugging. - Using GitHub Actions: The application is compiled for any platform and ready-to-distribute files are automatically added as a draft to the GitHub releases page.
Because the renderer
works and builds like a regular web application, you can only use dependencies that support the
browser or compile to a browser-friendly format.
This means that in the renderer
you are free to use any frontend dependencies such as Vue, React, lodash, axios and so
on. However, you CANNOT use any native Node.js APIs, such as, systeminformation
. These APIs are only available in
a Node.js runtime environment and will cause your application to crash if used in the renderer
layer. Instead, if you
need access to Node.js runtime APIs in your frontend, export a function form the preload
package.
All dependencies that require Node.js api can be used in
the preload
script.
Here is an example. Let's say you need to read some data from the file system or database in the renderer.
In the preload context, create a function that reads and returns data. To make the function announced in the preload
available in the render, you usually need to call
the electron.contextBridge.exposeInMainWorld
. However,
this template uses the unplugin-auto-expose plugin, so you just need
to export the method from the preload. The exposeInMainWorld
will be called automatically.
// preload/index.ts
import { readFile } from 'node:fs/promises';
// Encapsulate types if you use typescript
interface UserData {
prop: string
}
// Encapsulate all node.js api
// Everything you exported from preload/index.ts may be called in renderer
export function getUserData(): Promise<UserData> {
return readFile('/path/to/file/in/user/filesystem.json', {encoding:'utf8'}).then(JSON.parse);
}
Now you can import and call the method in renderer
// renderer/anywere/component.ts
import { getUserData } from '#preload'
const userData = await getUserData()
Find more in Context Isolation tutorial.
Although the preload has access to all of Node.js's API, it still runs in the BrowserWindow context, so a limited electron modules are available in it. Check the electron docs for full list of available methods.
All other electron methods can be invoked in the main
.
As a result, the architecture of interaction between all modules is as follows:
sequenceDiagram
renderer->>+preload: Read data from file system
preload->>-renderer: Data
renderer->>preload: Maximize window
activate preload
preload-->>main: Invoke IPC command
activate main
main-->>preload: IPC response
deactivate main
preload->>renderer: Window maximized
deactivate preload
Find more in Inter-Process Communication tutorial.
All environment variables are set as part of the import.meta
, so you can access them vie the following
way: import.meta.env
.
Note: If you are using TypeScript and want to get code completion you must add all the environment variables to the
ImportMetaEnv
intypes/env.d.ts
.
The mode option is used to specify the value of import.meta.env.MODE
and the corresponding environment variables files
that need to be loaded.
By default, there are two modes:
production
is used by defaultdevelopment
is used bynpm run watch
script
When running the build script, the environment variables are loaded from the following files in your project root:
.env # loaded in all cases
.env.local # loaded in all cases, ignored by git
.env.[mode] # only loaded in specified env mode
.env.[mode].local # only loaded in specified env mode, ignored by git
Warning: To prevent accidentally leaking env variables to the client, only variables prefixed with
VITE_
are exposed to your Vite-processed code.
For example let's take the following .env
file:
DB_PASSWORD=foobar
VITE_SOME_KEY=123
Only VITE_SOME_KEY
will be exposed as import.meta.env.VITE_SOME_KEY
to your client source code, but DB_PASSWORD
will not.
You can change that prefix or add another. See envPrefix
See Contributing Guide.