@agentprodready/vector-store

Provider-independent vector store contracts (+ in-memory reference) for AgentProdReady semantic / hybrid memory search.

| | | |---|---| | Status | Production contracts published (1.0.x) | | Install | npm install @agentprodready/vector-store | | Module | ESM · Node.js >=22 <25 | | License | MIT |

When to use

Enable semantic search behind @agentprodready/memory. Prefer Simple memory: true first if you only need ephemeral wiring.

Install

npm install @agentprodready/vector-store
# Durable pgvector provider (optional)
npm install @agentprodready/vector-store-pgvector

Sample

import { InMemoryVectorStore } from '@agentprodready/vector-store';

const store = new InMemoryVectorStore({
  dimensions: 32,
  embeddingModelId: 'reference-embedding-32',
});

await store.upsert({
  id: 'mem-1',
  vector: new Array(32).fill(0.1),
  metadata: { tenantId: 't1' },
});

Composition selects the concrete adapter. Raw vendor distance types never escape the adapter boundary.

Ownership

| Owns | Does not own | |---|---| | VectorStorePort, normalized scores, in-memory reference | Memory lifecycle / ranking / auth | | Dimension / model / metric mismatch errors | Embedding generation (AI Provider) | | | pgvector SQL (@agentprodready/vector-store-pgvector) |

Docs

• Vector search guide
• Memory package
• Repository README

License

MIT © 2026 ameenmari