Get started

Use Node.js 22 or later. @anvia/pinecone uses Pinecone SDK 9 and targets Pinecone API 2026-07.

Provision a Pinecone index with the embedding model's dimension before connecting.

pnpm add @anvia/core @anvia/pinecone @pinecone-database/pinecone

import { retrieveDocuments } from '@anvia/core/vector-store';
import { PineconeVectorClient } from '@anvia/pinecone';
const storeClient = new PineconeVectorClient({
    client
});
const store = storeClient.vectorStore({
    indexName: 'support-docs',
    namespace: 'production',
    metric: 'cosine',
    dimensions: embeddings.dimensions!
});
await store.validate();
await store.upsert({
    documents: documents
});
const results = await retrieveDocuments({ store, model: embeddings, query, topK: 5 });

Documents must already contain embeddings. The adapter uses the standard Pinecone client when client is omitted. Keep credentials and endpoint configuration in that SDK or inject a centrally managed client.

Next

• Collections and indexing
• Search and filters
• Production