How it works

Unlighthouse has multiple steps, aimed at only running logic when it's needed.

The core logic is as follows

1. Instantiation

Configuration is loaded from unlighthouse.config.ts or a custom configFile.

const unlighthouse = await createUnlighthouse(config)

Worker

A new puppeteer-cluster is created.

2. Server Context Provided

Once the context is provided for which site is being scanned and server details.

// create a server to serve the client from
const { server, app } = await createServer()
await unlighthouse.setServerContext({ url: server.url, server: server.server, app })

API

An unrouted API instance is created.

Client

Vite client is copied from the node_modules and is injected with static configuration for the scan

3. Start

For integrations, Unlighthouse will only start when accessed. Otherwise, Unlighthouse is started straight away.

hooks.hookOnce('visited-client', () => {
  unlighthouse.start()
})

Context

Discovery of the route definitions is attempted. A virtual router for the route definitions is created.

Attempt to read the robots.txt to disocver the sitemap and excluded routes.

With the sitemap.xml, we collect the list of URLs to work with, if no sitemap is discovered, the home page will be scanned.

Worker

For each URL:

1. perform a GET and extract HTML payload for basic SEO data and discover new internal links
2. In a new thread, perform the lighthouse scan, the HTML and JSON payload of the report will be saved to the filesystem

Client

Broadcasting setup on worker events.

3. Using the client

API

The unrouted API instance routes requests to the worker to perform actions. Static files such as the full-page screenshot and lighthouse HTML report are served.