Indexing site with authentication

[erzz]

In one deployment scenario we need to wrap the application in our SSO provide (azure) which means that the site is built with authentication enabled.

Lunr seems to do its indexing after the site is built (understandably) which means that in this scenario it cannot access any of the URLs (as it has no SSO credentials). This results in zero docs getting indexed of course.

The only workaround I can think of for this would be:

• Run the build once with SSO disabled to generate the index and search-docs.
• Run build again with SSO enabled for deployment
• Copy the index and search docs over to the build folder of the deployment run

I dont expect that there is a lot you can do about solving SSO requirements here - but do wonder if you have any brainwaves on a better approach or method of generating the index in a different way?

[lxix]

This is how I do it:
I have a bare (meaning that I have not modified/wrapped docusaurus or docusaurus-lunr-search in any way, especially with auth) docusaurus project in my private repository, which I build a docker image from (with nixpacks, but not important). Then I push the image to a (private) registry and deploy it to a server. On this server there is a traefik container configured to do the authentication before requests can reach the docusaurus container.

[erzz]

Thanks @lxix :)

Whilst you wrote your answer I also found an answer thanks to the tip in #105 (which this is essentially a dupe of!). Here is my solution when using azure's msal-browser auth package.

export default function Root({children}) {
  /* this allows the search indexing to work in a CI context */
  const isBrowser = useIsBrowser();
  if (!isBrowser) {
    return <>{children}</>;
  } else {
    /* otherwise require SSO */
    return (
      <MsalProvider instance={pca}>
        <UnauthenticatedTemplate>
          <Login />
        </UnauthenticatedTemplate>
        <AuthenticatedTemplate>{children};</AuthenticatedTemplate>
      </MsalProvider>
    );
  }
}

So essentially the build process is not a browser context, so we can detect that and allow it to bypass auth. But when its being served as a CSR the auth kicks in.