Get started

pnpm add @anvia/core @anvia/weaviate weaviate-client

import { retrieveDocuments } from '@anvia/core/vector-store';
import { WeaviateVectorClient } from '@anvia/weaviate';
const storeClient = new WeaviateVectorClient({});
const store = storeClient.vectorStore({
    collectionName: 'SupportDocs',
    dimensions: 1536,
    metric: "cosine"
});
await store.validate();
await store.upsert({
    documents: documents
});
const results = await retrieveDocuments({ store, model: embeddings, query, topK: 5 });

Anvia supplies precomputed vectors, so provision the collection without a vectorizer. Without a client, the adapter uses local HTTP and gRPC environment defaults with insecure connections; inject a configured client for remote deployments.

Next

• Collections and indexing
• Search and filters
• Production