Full text search is finally here for Firestore!

Added: 2026-05-21

[00:00:02][music] >> Eight years ago, we introduced Firestore to developers. And eight years ago, we started to frequently get this one feature request.

[00:00:13]>> [music] >> More Morgan Chen in videos.

[00:00:18]No.

[00:00:19]Full-text search.

[00:00:21]We are happy to announce that full-text search is making its way to Firestore Enterprise Edition.

[00:00:27]Let's take a look at integrating this within our application.

[00:00:31]I have an application [music] that I made as an issue tracker.

[00:00:35]I'm going to quickly be able to filter through issues that contain keywords.

[00:00:39]Prior to full-text search being available [music] in Firestore Enterprise, I would have to work with a third-party search provider to index my data. This was fine. However, integrating directly with Firestore reduces my compute costs because I don't need to export this data to an external search service. And also eliminates billing [music] costs from the external service as well.

[00:00:59]Everything is integrated.

[00:01:01]I also could have used something like vector search to help narrow requests.

[00:01:05]As well, by using vector search, I can easily determine semantic meaning from requests.

[00:01:10]This has the caveat that it also adds additional overhead for reaching out to the Gemini API to generate embeddings for the content.

[00:01:17]>> [music] >> And also generate embeddings on the request for the content.

[00:01:22]And if my content was too large to generate embeddings from, I would need to come up with a chunking strategy to divide my content into smaller chunks.

[00:01:30]Having said that, let's examine my example a little more closely to get started.

[00:01:36]To get full-text search working, I need to first create an index of the search terms I am interested in.

[00:01:42]I head to the Firestore page in the Google Cloud Console and go to the index screen. On the index screen, I select the options create an index and to create a search index.

[00:01:52]On the create search index screen, I give my index a name, a search type, either text or geo, the collection ID, in my case it's issues since that is where my issues are stored, [music] and the field I want to index on for full text search.

[00:02:07]I then want to select my query scope. In my example, I will use the collection scope instead of the collection group.

[00:02:14]Morgan has a great video describing the differences between these two query scopes, and you should definitely check out the video linked below. And we click create.

[00:02:23]Now we wait for the index to be built.

[00:02:25]This takes a couple minutes.

[00:02:28]Once the index is built, we can start querying our collection using the search modifier for pipeline queries.

[00:02:34]And we can also negate search terms or combine multiple search terms within a query. The code looks something like this, >> [music] >> where you execute a pipeline query on a collection with a search request.

[00:02:46]In my application, we can see I have about 500 total issues, and when I query for storage, I get 67 issues.

[00:02:54]When I add a minus token, since I cannot help with token issues, I get a total of 62 issues.

[00:03:00]And there you have it. With a few short lines, we are able to query our Firestore enterprise database using full text search.

[00:03:07]I have extra documentation linked in the description below.

[00:03:11]I am Noe, and I cannot wait to see what you build next.

[00:03:19]>> [music] [music] [music]