ABR often doesn't match Prowlarr when titles are simple

[pekechew]

Example:

Title is "Damned" by Genevieve Cogman

ABR finds the correct book

ABR sends Prowlarr only the title to search for in the "Audiobook" category

2025-06-01 00:00:42,069 [INFO] app.internal.prowlarr.prowlarr: Querying prowlarr: http://127.0.0.1:9696/api/v1/search?query=Damned&type=search&limit=100&offset=0&categories=3030

Prowlarr returns "Wrath of the Damned" by Ella Fields

I think it would be better to include the author in the query (or have an option to do so) as that works in a manual search.

Or maybe it searches "Title Author" and if that fails, then it falls back to just "Title", or there could be some sort of validation between the two.

[bdunn44]

+1 for this! I frequently have the same issue. I think adding the author would help in most cases. Adding a configuration page to tweak the Prowlarr query parameters would be a nice feature as well

[trsmith78]

I'm having the same or similar issue with Dust by Huge Howey. If I search prowlarr directly with this query, it returns several of the correct audiobooks. However, through audiobookrequest, it finds the book so I can select and add it to my requests, but when I actually initiate the search, the books doesn't show at all. Instead I get a page full of similar matches with the word "Dust", but none with the author requested Hugh Howey. I mean a page full of results none even close. Again, if I issue the query in prowlarr, the books is literally the first several results.

[markbeep]

That seems to be a problem on the heuristics I use for scoring sources. Currently it already tries to determine which result from prowlarr is the most likely by comparing the title, subtitle, author, file size, etc. When you view the "Sources" of a book on the wishlist, the sources are shown ordered according to these heuristics.

I've been thinking about slightly updating it anyhow. One problem is that I don't want to search too specifically, because that then can result in queries not returning anything from prowlarr.

So as you suggested, it might make sense to look into creating multiple searches over Prowlarr, i.e.:

• title only
• subtitle only
• title + author

Then combine all the sources and run some updated heuristics over them to better determine what the best results are.

[michmerr]

Just set this up, and the first search turned up a list of titles (all part of a series). I selected a title, and it's download icon disappeared...but I hadn't enabled automatic download, so nothing happened. I enabled auto-download, and went back to the list. The title was still missing the download link (showing a very faint checkmark, instead). I selected another title, and after a few gears turned, I could see the download being processed.

I checked Prowlarr to find the search for the download operation was, as already discussed, the simple title of the book. The first search result was used for the download attempt. This is fair in that the title string matched a string in the result, so there wasn't a whole lot to evaluate without something more than the title. The download failed, complaining of missing items, but the title remains unselectable in the ABR search results, just like the one selected before auto-download was selected.

Even if something were in place to try the next matching source in the event of a download failure, things would have run into problems because there were several different matches near the top that were from different authors. Perhaps a little out of scope for the search discussion, but another reason for improving it.

It's interesting looking at the subtitles of the books in the search results. They vary in how they associate the title with the series. Some just say that they are in the series, some have the number, etc. All manageable when evaluating the strings, but it makes me wonder if the book object returned by the Audible API has a property that identifies the book as a part of a series (or even multiple series). That might be an additional string used to refine a search.

[michmerr]

I just looked at the original issue description, and for common/simple titles, that page length is going to be the limiting factor. Prowlarr doesn't return any pagination metadata that I can see, which makes it hard to use something like a total result count to gauge how expensive it would be to page through all the results if no satisfactory match is found.

Defaulting to title and author will probably get the best results. Not sure how interesting that gets with anthologies, or other books with a bunch of authors. That should net few enough results that you can trying matching them all, with the option of falling back to a title-only search if there were no results. I'd hesitate to match any result that doesn't have at least the author's last name, though.