Recipe

Pinecone Vector DB

Store and query high-dimensional embeddings at scale with Pinecone's managed vector database. Ideal for semantic search, RAG pipelines, and recommendation engines.

Index Setup

Create a pod-based or serverless index with cosine similarity. Choose dimensions matching your embedding model — 1536 for OpenAI, 768 for all-MiniLM.

Upsert & Query

Batch upsert vectors with metadata payloads. Query via top-K nearest neighbor search with optional metadata filters for hybrid retrieval.

Namespaces

Partition vectors into namespaces for multi-tenant isolation or staging environments without spinning up separate indexes.

Sparse-Dense

Combine dense embeddings with sparse BM25 vectors using Pinecone's hybrid search for keyword-aware semantic retrieval.

Quick Start

import { Pinecone } from '@pinecone-database/pinecone'

const pc = new Pinecone({ apiKey: process.env.PINECONE_API_KEY })
const index = pc.index('meridian-embeddings')

await index.upsert([
  { id: 'doc-1', values: embedding, metadata: { source: 'docs' } }
])

const results = await index.query({
  vector: queryEmbedding,
  topK: 10,
  includeMetadata: true
})

← Back to Docs Pinecone Docs ↗