Pinecone JS SDK Cheat Sheet

This quick reference guide summarizes the most common patterns and configurations in Pinecone using the official Node.js SDK.

1. Client Initialization and Index Operations

import { Pinecone } from "@pinecone-database/pinecone";

// 1. Initialize client instance
const pc = new Pinecone({ apiKey: "YOUR_API_KEY" });

// 2. Create a serverless index
await pc.createIndex({
  name: "new-index",
  dimension: 1536,
  metric: "cosine",
  spec: {
    serverless: {
      cloud: "aws",
      region: "us-east-1",
    },
  },
});

// 3. Drop an index
await pc.deleteIndex("new-index");

2. Ingesting Vectors (Upsert)

const index = pc.index("new-index");

// Upsert vectors with unique IDs and custom JSON metadata
await index.upsert([
  {
    id: "item_a",
    values: [0.012, -0.045, 0.089], // Dimension values array
    metadata: { category: "web", price: 29.99 },
  },
  {
    id: "item_b",
    values: [-0.005, 0.098, -0.012],
    metadata: { category: "database", price: 14.50 },
  },
]);

3. Querying Similar Vectors with Metadata Filters

const index = pc.index("new-index");

// Execute a similarity search
const results = await index.query({
  vector: [0.010, -0.040, 0.080],
  topK: 5,
  includeMetadata: true,
  filter: {
    category: { $eq: "web" },
    price: { $lte: 30.00 },
  },
});

4. Querying inside a Namespace Partition

// Target a specific client namespace partition
const userSpace = pc.index("new-index").namespace("user-123");

// Query will search only vectors within user-123 partition
const userResults = await userSpace.query({
  vector: [0.010, -0.040, 0.080],
  topK: 3,
  includeMetadata: true,
});