Replacement search for self-hosted Confluence?

Question

Anyone who has enough data in their Confluence instance will know why I'm asking this question. I saw something that could be used for Confluence Cloud, and am (desperately!) hoping to find something that would replace the search engine in self-hosted editions.

Atlassian Team: Confluence search is bad enough that I'm starting to consider recommending we go back to MediaWiki. Finding information is, at the end of the day, far more important by far than having tie-ins to Jira and whatnot - and those tie-ins can be scripted into MediaWiki anyhow.

Mike Ely · Answer

Thanks to everyone who took the time to respond, and I will definitely look into the specific suggestions offered for enhancements or replacements for the built-in search.

While I know it wasn't the intention, responses focusing on structure, tagging, etc. actually aren't helpful. I've experienced large internal wikis on both Confluence and MediaWiki (with SphynxSearch) and the difference between the two platforms in terms of search quality was profound. There was no great need to carefully tend to wiki structure, tag each page as though it were some sort of artisanal garden of fragile delights, the search on MW Just Worked

In terms of specifics, I'd narrow them down to two significant categories: presentation and relevance.

On the presentation end, having search results appear as a partial drop-down just doesn't work visually. When done that way the search results are necessarily truncated, making it harder to determine whether a given result is what one is looking for, and leading to annoying multiple-scrollbar problems in the browser. Just make the results a regular page for Pete's sake.

On the question of relevance, Confluence search really falls on its face. When I'm looking for, say, "Juniper LACP" my top results are just as likely to include a fact sheet about a particular model of Juniper switch somebody uploaded when speccing out a project five years ago as they are to show the nice tutorial one of the network engineers wrote a couple months back containing our shop standard for how to do this work. MediaWiki? You're going to see the tutorial, top and center on the normal easy-to-read results page.

Suggestion: Confluence search should prioritize results based on privilege. So assuming I'm an Editor on spaces named "Elephants," "Camels" and "Buffalo" and I'm a Reader on spaces named "Turkeys," "Walrus" and "Chinchillas," it stands to reason most of my work is going to revolve around elephants, camels and buffalo so whatever I'm searching for is more likely to be in those spaces than the others. Put those at the top.

I was griping internally about the general fail that is Confluence search and my coworkers have mostly landed on two workarounds:

Memorize the page titles of important-to-them Confluence pages so they can search for them directly.
Keep a text file of same, occasionally with hotlinks to the pages.

Both of those workarounds shine a bright and very clear light on how Confluence search Does. Not. Work.

Huh. Is there a SphinxSearch plugin for Confluence? Because I know that stuff works just fine. Also something that presents the search results in a regular page rather than a dropdown menu.

Mike Ely · Answer

Preamble: can we get search to return in a single page rather than a drop-down, and can we integrate SphynxSearch into Confluence as it's a much better engine? Specific suggestions appreciated and will be looked into.

Kristian Klima_gtcuat · Answer

Hi @Mike Ely

Confluence Server search was pretty bad but I find search ond Cloud really good out of the box.

Of course, a lot depends on organizing content - having guidelines and discipline to follow them to avoid turning Confluence into a cesspool of 'user-generated/outdated/duplicated' content. But that can happen in any tool.