@absolutejs/rag-pinecone

Pinecone vector-store adapter for @absolutejs/rag. Treats Pinecone as the backend boundary: server-side filtering, native vector search, and explicit opt-in index provisioning helpers.

Install

bun add @absolutejs/rag @absolutejs/rag-pinecone @pinecone-database/pinecone

Usage

import { createPineconeRAG } from "@absolutejs/rag-pinecone";

const rag = createPineconeRAG({
  apiKey: process.env.PINECONE_API_KEY,
  indexName: "absolute-rag-demo",
  namespace: "production",
  vector: {
    provider: "pinecone",
    dimensions: 1536,
    distanceMetric: "cosine",
  },
});

await rag.store.upsert({
  chunks: [
    {
      chunkId: "doc-1#0",
      text: "Pinecone stores vectors with attached metadata.",
      title: "Pinecone overview",
      source: "https://docs.pinecone.io",
      metadata: { tags: ["vector", "managed"] },
    },
  ],
});

const hits = await rag.collection.search({ query: "vector database", topK: 4 });

Index provisioning

The Pinecone index is not auto-provisioned at runtime. For explicit, opt-in provisioning use the exported helpers:

import {
  describePineconeIndex,
  ensurePineconeIndex,
} from "@absolutejs/rag-pinecone";

await ensurePineconeIndex({
  apiKey: process.env.PINECONE_API_KEY,
  indexName: "absolute-rag-demo",
  dimensions: 1536,
  metric: "cosine",
  waitUntilReady: true,
});

License

Apache License 2.0. See LICENSE.