feat(vertexai): Migrate to AI and add GoogleAI support

[dlarocque]

API: http://goto.google.com/googleai-in-vinf-api (internal)

This PR integrates support for the Google AI API alongside the existing Vertex AI backend. To accommodate this, the SDK has been refactored to use the more generic AI naming convention instead of VertexAI.

Changes

Google AI Support

• Adds a googleAIBackend() function to configure the SDK to use the Google AI API. This enables usage with models available in a free tier.
• Introduces mappers (googleAIMappers.ts) to translate requests and responses between the SDK's unified format (prioritizing Vertex AI) and the specific format required by the Google AI API. This handles differences in features like safety settings (method unsupported), citation metadata (citationSources vs citations), and response properties (defaulting missing safety scores).
• Adds checks and throws AIError with code UNSUPPORTED or logs warnings for features/properties not available or different in the Google AI backend (e.g., SafetySetting.method, rounding topK, Part.videoMetadata).

Firebase AI Migration

• Renames the core service interface and related classes/types from VertexAI to AI (e.g., AI, AIError, AIModel, AIErrorCode, AIOptions).
• Maintains backwards compatibility by exporting the old names as aliases to the new ones (e.g., export type VertexAI = AI;, export const VertexAIError = AIError;).
• Introduces Backend (GoogleAIBackend | VertexAIBackend) and BackendType types to abstract the backend configuration.

Initialization

• Introduces the primary getAI(app?, options?) function for initializing the SDK. It accepts AIOptions to specify the desired backend.
• Keeps the existing getVertexAI(app?, options?) function for backward compatibility, initializing a Vertex AI backend instance.

Internal

• Adds an internal encoding/decoding scheme (helpers.ts) for instance identifiers based on backend type and location, allowing the component system to manage different service instances.
• Updates internal request logic (request.ts, stream-reader.ts) to handle backend differences and use the appropriate mappers.

Packaging

• Added a ./ai export path in firebase/package.json to allow the package to be imported with firebase/ai.

Testing

• Added new unit tests covering backend configuration (googleAIBackend, vertexAIBackend), Google AI request/response mappers, and instance identifier helpers.
• Added specific tests for core methods (generateContent, countTokens) using the Google AI backend path to verify request mapping and error handling.
• Included tests explicitly for backwards compatibility for types and functions.

Sample usage

import { getAI, getVertexAI getGenerativeModel, vertexAIBackend } from 'firebase/ai';

// Google AI
let ai = getAI(app); // Google AI is the default backend.
let model = getGenerativeModel(ai, { model: 'gemini-2.0-flash' });
const result = await model.generateContent('Why is the sky blue?');
console.log(result.response.text());

// Vertex AI
let ai = getAI(app, { backend: vertexAIBackend() });
let model = getGenerativeModel(ai, { model: 'gemini-2.0-flash' });
const result = await model.generateContent('Why is the sky blue?');
console.log(result.response.text());

// Backwards compatibility
let vertexAI = getVertexAI(app); // Vertex AI backend
let model = getGenerativeModel(vertexAI, { model: 'gemini-2.0-flash' });
const result = await model.generateContent('Why is the sky blue');
console.log(result.response.text());

[changeset-bot]

🦋 Changeset detected

Latest commit: 41b0385

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
firebase	Minor
@firebase/vertexai	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[google-oss-bot]

Size Report ¹

Affected Products

• @firebase/vertexai

  Type	Base (1363ecc)	Merge (79f6ae1)	Diff
  browser	34.7 kB	41.3 kB	+6.54 kB (+18.8%)
  main	35.7 kB	42.4 kB	+6.71 kB (+18.8%)
  module	34.7 kB	41.3 kB	+6.54 kB (+18.8%)

• firebase

  Type	Base (1363ecc)	Merge (79f6ae1)	Diff
  firebase-vertexai.js	28.3 kB	32.7 kB	+4.48 kB (+15.9%)

Test Logs

• Base (1363ecc): https://github.com/firebase/firebase-js-sdk/actions/runs/14484800684
• Merge (79f6ae1): https://github.com/firebase/firebase-js-sdk/actions/runs/14621866391

1. https://storage.googleapis.com/firebase-sdk-metric-reports/scoMiZxJiw.html

[google-oss-bot]

Size Analysis Report ¹

This report is too large (116,370 characters) to be displayed here in a GitHub comment. Please use the below link to see the full report on Google Cloud Storage.

Test Logs

• Base (1363ecc): https://github.com/firebase/firebase-js-sdk/actions/runs/14484800684
• Merge (79f6ae1): https://github.com/firebase/firebase-js-sdk/actions/runs/14621866391

1. https://storage.googleapis.com/firebase-sdk-metric-reports/Y06hh4YpoW.html

[hsubox76]

Why do we want that?

[dlarocque]

We currently don't do any validation on location passed, so users can pass things like ' ' , 'x', etc without getting errors in the SDK or falling back to the default. I believe allowing '' fits within this approach.

Do you think '' should be treated the same as null or undefined and fallback to the default? I have no preference.

[hsubox76]

I guess it doesn't really matter either way but I do think it's weird to treat null and empty string differently as both of them have to be put there deliberately and with the empty string we're assuming someone putting it there deliberately does not want the default, while with null, we're assuming they do? If you want null to cause it to be default I think anything falsy should also cause default (so maybe if (!location))