@eidentic/pinecone

Pinecone vector store adapter for Eidentic — fully-managed vector search for agent memory using the Pinecone API. Implements VectorPort from @eidentic/types. The Pinecone index must be pre-created with the matching dimension and cosine metric; the adapter does not create indexes.

Install

pnpm add @eidentic/pinecone @pinecone-database/pinecone

Usage

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

const pinecone = new Pinecone({ apiKey: process.env.PINECONE_API_KEY! });
const index = pinecone.index("eidentic-memory");

const vector = PineconeVectorStore.create({
  index,
  dim: 1536,
});

// Use with Memory
import { Memory } from "@eidentic/memory";
const memory = new Memory({ store, vector, embedder });

Live conformance test

Set EIDENTIC_TEST_PINECONE_API_KEY to run the real-service vector conformance suite. Without EIDENTIC_TEST_PINECONE_INDEX, the suite creates a disposable dimension-4 cosine serverless index, waits until it is ready, and deletes it in teardown. An explicitly supplied index is preserved and only random test namespaces are deleted. Never target a production index.

The disposable index defaults to AWS us-east-1; override with EIDENTIC_TEST_PINECONE_CLOUD and EIDENTIC_TEST_PINECONE_REGION when required by the project. Because Pinecone is eventually consistent, the live suite waits 5 seconds between mutations and dependent reads. Override this only for provider-specific testing with EIDENTIC_TEST_PINECONE_CONSISTENCY_DELAY_MS (a non-negative millisecond value).

Links

• GitHub
• Issue tracker
• Root README

Apache-2.0