buoy: A TypeScript repository from derhub

Electron

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

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 in types/env.d.ts.
Automatic hot-reloads for the Main and Renderer processes.

Vite provides many useful features, such as: TypeScript, TSX/JSX, CSS/JSON Importing, CSS Modules , Web Assembly and much more.

See all Vite features.

TypeScript (optional)

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.

Guide to disable typescript and remove dependencies

Vue (optional)

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.

Find more forks 🔱 for others frameworks or setups

Continuous Integration

The configured workflow will check the types for each push and PR.
The configured workflow will check the code style for each push and PR.
Automatic tests used Vitest -- A blazing fast test framework powered by Vite.
- Unit tests are placed within each package and are ran separately.
- End-to-end tests are placed in the root tests directory and use playwright.

Publishing

Each time you push changes to the main branch, the release workflow starts, which creates a new draft release. For each next commit will be created and replaced artifacts. That way you will always have draft with latest artifacts, and the release can be published once it is ready.
- Code signing supported. See release workflow.
- Auto-update is supported. After the release is published, all client applications will download the new version and install updates silently.

Note: This template configured only for GitHub public repository, but electron-builder also supports other update distribution servers. Find more in electron-builder docs.

How it works

The template requires a minimum amount dependencies. Only Vite is used for building, nothing more.

Project Structure

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

Build web resources

The main and preload packages are built in library mode as it is simple javascript. The renderer package builds as a regular web app.

Compile 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.

Working with dependencies

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.

Expose in main world

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.

Working with Electron API

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.

Modes and Environment Variables

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 in types/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 default
development is used by npm 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

Contribution

See Contributing Guide.

derhub/buoy