# Firebase Retires the Vertex AI Name, Changes Your Default Region

> Source: <https://firerun.io/firebase-agent-platform-backend-rename-2026/>
> Published: 2026-08-15 00:00:00+00:00

# Firebase Retires the Vertex AI Name, Changes Your Default Region

Firebase's JS SDK deprecates VertexAIBackend for AgentPlatformBackend, and the new class defaults to global instead of us-central1.

Firebase AI Logic’s JavaScript SDK deprecated `VertexAIBackend`

in version 12.17.0, released July 30, 2026, replacing it with a new `AgentPlatformBackend`

class ([Firebase JS SDK releases, July 2026](https://github.com/firebase/firebase-js-sdk/releases)). The rename isn’t cosmetic: the new class defaults to a different region than the old one, a behavior change developers can miss if they treat this as a find-and-replace.

## What actually changed

The swap follows Google’s April 22, 2026 rebrand of Vertex AI into the Gemini Enterprise Agent Platform, announced at Cloud Next 2026 ([Google Cloud, April 2026](https://cloud.google.com/blog/products/ai-machine-learning/introducing-gemini-enterprise-agent-platform)). Google said at the time that future Vertex AI capability would ship exclusively through the new Agent Platform rather than as a standalone product. Firebase’s client SDK is now catching up to that shift, three months later: what Firebase’s docs called the “Vertex AI Gemini API” is now the “Agent Platform Gemini API,” one of the two backend options in Firebase AI Logic alongside the Gemini Developer API ([Firebase AI Logic docs](https://firebase.google.com/docs/ai-logic/faq-and-troubleshooting)).

`VertexAIBackend`

still works. Google hasn’t pulled it, but it’s deprecated, and new features will only land on `AgentPlatformBackend`

going forward.

## The part that actually breaks things

The two classes aren’t drop-in equivalents. `VertexAIBackend`

defaults to the `us-central1`

region if you don’t specify one. `AgentPlatformBackend`

defaults to `global`

([Firebase AI Logic docs, location reference](https://firebase.google.com/docs/ai-logic/locations?api=vertex)). Swap the class name without setting a location explicitly, and a call that ran in `us-central1`

now runs wherever Google routes “global” traffic, which matters for latency-sensitive apps or anyone with a data-residency requirement tied to a specific region.

Firebase’s own docs point out why the default flipped rather than just carrying `us-central1`

forward: most of Google’s newer Gemini models aren’t available in `us-central1`

at all. Pinning the old regional default into the new class would have quietly cut off access to current models. Developers who need `us-central1`

specifically have to ask for it: `new AgentPlatformBackend('us-central1')`

.

## Where this fits

This is Firebase’s second rename in this exact spot in about a year. “Vertex AI in Firebase” became “Firebase AI Logic” in May 2025, and the underlying package went from `firebase-vertexai`

to `firebase-ai`

at the same time. That first rename was mostly cosmetic: it reflected Firebase AI Logic adding support for the Gemini Developer API alongside Vertex AI, not a change in how existing code behaved. This one is different: it’s Firebase’s SDK inheriting a platform-wide rebrand from Google Cloud, and it ships with a default that changes where your requests actually go.

Firerun’s take: the version bump itself is easy to skip past. SDK release notes are exactly the kind of change that gets batch-approved in a dependency update PR without a close read. Anyone on `VertexAIBackend`

today should treat this one differently. Read the region section of the migration notes before swapping classes, and if the target region matters to your app, set it explicitly rather than trust the new default to match the old one.