What is the difference between firebase_vertexai and firebase_ai?

[htsuruo]

I found the PR for creating firebase_ai.
#17332

It looks like firebase_ai is the package copied by firebase_vertexai and both are quite similar. Is that accurate?
I'm a bit confused about which Dart package I should use for implementing generative AI.

I understand there have been a lot of changes since the Google I/O '25 announcements and the introduction of Firebase AI Logic. Here's my current understanding:

Before:

• google_generative_ai: This package directly used the Gemini Developer API. This meant developers had to store the Gemini API Key directly in the app, making it suitable only for prototypes, not production applications.
• firebase_vertexai: This was an SDK for Vertex AI for Firebase. It supported Firebase App Check, which made it suitable for production apps.

After:

• firebase_ai: This is the SDK for Firebase AI Logic. google_generative_ai has been deprecated and replaced by firebase_ai since Firebase AI Logic launched.
• firebase_vertexai: This package seems to have remained unchanged.

In reality, firebase_ai supports using both the Gemini Developer API and Vertex AI.
firebase_ai = google_generative_ai + firebase_vertexai

Given this, it seems firebase_ai is superior to firebase_vertexai.
Should developers now use firebase_ai exclusively for all cases, or is there still a reason to use both packages depending on the situation?

Firebase AI Logic was formerly called "Vertex AI in Firebase" with the plugin firebase_vertexai.
https://firebase.google.com/docs/flutter/setup?platform=ios#available-plugins

[cynthiajoan]

@htsuruo, thanks for this question. In general your understanding is correct. More details and explanation can be found in our migration doc

The reason that we keep firebase_vertexai is try not to break existing developer experience, and we recommend developers to pick up with firebase_ai or migrate to firebase_ai.

We will start to put in appropriate labels in corresponding packages for better clarification and help developers making choices.

[htsuruo]

@cynthiajoan

Thanks you quick reply,

we recommend developers to pick up with firebase_ai or migrate to firebase_ai.

OK, understand well.

We will start to put in appropriate labels in corresponding packages for better clarification and help developers making choices.

Cool. That could help all developers. I've opened an issue ( #17397 ) so new developers don't get confused. If you agree with my suggestion, please go ahead with it.