Firestore-level queries, enterprise sign-in, point-in-time recovery
Documents gain OR filters, aggregations, collection groups and vector search. Auth gains SCIM and blocking hooks. Dedicated databases can be restored to any second.
- DocumentsOR and AND filter groups, nested like Firestore's
or()/and(), andcount,sumandaverageaggregations computed on the server. Read more - DocumentsCollection-group queries across every subcollection with the same name, and live listeners on collection groups. Read more
- DocumentsVector search on documents (
findNearest) on pgvector, under the same rules as every other query. Read more - AuthSCIM 2.0 provisioning, so identity providers such as Okta and Microsoft Entra can create, update and remove app users. Read more
- AuthBlocking hooks that run before a user is created or signs in, and can refuse or enrich the request. Read more
- AuthAnonymous users are faster to create, and policies and SQL can tell them apart (
auth.jwt()->>'is_anonymous'). Read more - AuthToken signing does less work per request, so sign-in and refresh are quicker under load.
- DatabasePoint-in-time recovery for dedicated project databases: restore to a chosen moment, not only to a daily backup. Read more
- DatabaseFaster SQL: repeated statements skip re-analysis, each project gets more pooled connections, and busy projects wait briefly for a slot instead of failing. Read more
- RealtimeA faster connect path and message fan-out (each message is encoded once for all subscribers), with the measured numbers published. Read more
- FunctionsNode.js 24, Go 1.27 and Bun 1.4 runtimes. A function that stops without a result now says why: a signal, a sandbox limit or its own error output. Read more
- SecurityEvery engine has its own database and cache logins, platform images are signed in CI, and a security scan runs weekly. Read more
- DocsReference pages for the Swift, Kotlin and Dart SDKs, eight framework quickstarts, and tutorials for chat, payments, RAG and migrating from Supabase or Firebase. Read more