Function vertexAIVectorSearch

VertexAI vector search plugin

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

const ai = genkit({
  plugins: [
    vertexAI({ ... }),
    vertexAIVectorSearch({
       projectId: PROJECT_ID,
       location: LOCATION,
       vectorSearchOptions: [
         {
           publicDomainName: VECTOR_SEARCH_PUBLIC_DOMAIN_NAME,
           indexEndpointId: VECTOR_SEARCH_INDEX_ENDPOINT_ID,
           indexId: VECTOR_SEARCH_INDEX_ID,
           deployedIndexId: VECTOR_SEARCH_DEPLOYED_INDEX_ID,
           documentRetriever: VECTOR_SEARCH_DOCUMENT_RETRIEVER,
           documentIndexer: VECTOR_SEARCH_DOCUMENT_INDEXER,
           embedder: VECTOR_SEARCH_EMBEDDER,
         },
       ],
     }),
  ],
});

const metadata1 = {
  restricts: [{
    namespace: "colour",
    allowList: ["green", "blue, "purple"],
    denyList:  ["red", "grey"],
  }],
  numericRestricts: [
  {
    namespace: "price",
    valueFloat: 4199.99,
  },
  {
    namespace: "weight",
    valueDouble: 987.6543,
  },
  {
    namespace: "ports",
    valueInt: 3,
  },
],
}
const productDescription1 = "The 'Synapse Slate' seamlessly integrates neural pathways, allowing users to control applications with thought alone. Its holographic display adapts to any environment, projecting interactive interfaces onto any surface."
const doc1 = Document.fromText(productDescription1, metadata1);

// Index the document along with its restricts and numericRestricts
const indexResponse = await ai.index({
  indexer: vertexAiIndexerRef({ ... }),
  [doc1],
});


// Later, construct a query using restricts and numeric restricts
const queryMetadata = {
  restricts: [{
    namespace: "colour",
    allowList: ["purple"],
    denyList: ["red"],
  }],
  numericRestricts: [{
    namespace: "price",
    valueFloat: 5000.00,
    op: LESS,
  }]
};
const query = "I'm looking for something with a projected display";
const queryDoc = new Document(query, queryMetadata);

const response = await ai.retrieve({
  retriever: vertexAIRetrieverRef({ ... }),
  query: queryDocument,
  options: { k },
});

console.log(`response: ${response}`);
Copy