Vertex AI plugin for Genkit

This package provides Vertex AI integrations for Genkit, including Model Garden, Rerankers, Evaluation, and Vector Search.

⚠️ Deprecation notice: The main vertexAI plugin export (Gemini, Imagen, and embedder models) is deprecated. Please migrate to @genkit-ai/google-genai:

// Before (deprecated)
import { vertexAI } from '@genkit-ai/vertexai';

// After
import { vertexAI } from '@genkit-ai/google-genai';
Copy

Installation

npm i --save @genkit-ai/vertexai
Copy

Sub-packages

This package provides the following sub-package imports:

Model Garden

Access third-party models (Anthropic Claude, Mistral, Llama) hosted on Vertex AI Model Garden:

import { genkit } from 'genkit';
import { vertexModelGarden } from '@genkit-ai/vertexai/modelgarden';

const ai = genkit({
  plugins: [
    vertexModelGarden({ projectId: 'my-project', location: 'us-central1' }),
  ],
});

const { text } = await ai.generate({
  model: vertexModelGarden.model('claude-sonnet-4'),
  prompt: 'Write a haiku about cloud computing',
});

console.log(text);
Copy

Rerankers

Use Vertex AI's reranking API to reorder documents by relevance:

import { genkit } from 'genkit';
import { vertexRerankers } from '@genkit-ai/vertexai/rerankers';

const ai = genkit({
  plugins: [
    vertexRerankers({ projectId: 'my-project', location: 'us-central1' }),
  ],
});
Copy

Evaluation

Evaluate AI output quality using Vertex AI's built-in metrics:

import { vertexAIEvaluation } from '@genkit-ai/vertexai/evaluation';
import { VertexAIEvaluationMetricType } from '@genkit-ai/vertexai/evaluation';

const ai = genkit({
  plugins: [
    vertexAIEvaluation({
      projectId: 'my-project',
      location: 'us-central1',
      metrics: [
        VertexAIEvaluationMetricType.BLEU,
        VertexAIEvaluationMetricType.ROUGE,
        VertexAIEvaluationMetricType.SAFETY,
        VertexAIEvaluationMetricType.GROUNDEDNESS,
      ],
    }),
  ],
});
Copy

Vector Search

Use Vertex AI Vector Search for retrieval-augmented generation (RAG) with BigQuery or Firestore document stores:

import { vertexAIVectorSearch } from '@genkit-ai/vertexai/vectorsearch';

const ai = genkit({
  plugins: [
    vertexAIVectorSearch({
      projectId: 'my-project',
      location: 'us-central1',
      vectorSearchOptions: [
        {
          publicDomainName: 'my-public-endpoint.vdb.vertexai.goog',
          indexEndpointId: 'my-index-endpoint-id',
          indexId: 'my-index-id',
          deployedIndexId: 'my-deployed-index-id',
          documentRetriever: myDocRetriever,
          documentIndexer: myDocIndexer,
          embedder: myEmbedder,
        },
      ],
    }),
  ],
});
Copy

More information

The sources for this package are in the main Genkit repo. Please file issues and pull requests against that repo.

Usage information and reference details can be found in official Genkit documentation.

License: Apache 2.0