@orka-js/memory

In-memory vector database adapter for OrkaJS — ideal for development, testing, and small datasets.

Installation

npm install @orka-js/memory

Quick Start

import { MemoryVectorAdapter } from '@orka-js/memory'
import { Orka } from '@orka-js/core'

const vectorDB = new MemoryVectorAdapter()

const orka = new Orka({ llm, vectorDB })

await orka.knowledge.ingest([
  { id: 'doc-1', content: 'Hello world' },
  { id: 'doc-2', content: 'Goodbye world' },
])

const results = await orka.knowledge.query('hello', 2)

Direct Usage

const adapter = new MemoryVectorAdapter({ similarity: 'cosine' })

await adapter.createCollection('docs', 3)
await adapter.upsert('docs', [
  { id: '1', vector: [0.1, 0.2, 0.3], metadata: { text: 'foo' } },
  { id: '2', vector: [0.4, 0.5, 0.6], metadata: { text: 'bar' } },
])

const results = await adapter.search('docs', [0.1, 0.2, 0.3], 1)
const stats = await adapter.getCollectionStats('docs')
// { count: 2, dimension: 3 }

Note: Data is stored in process memory only and is not persisted across restarts.

Configuration

| Option | Type | Default | Description | |--------|------|---------|-------------| | similarity | 'cosine' \| 'euclidean' \| 'dotProduct' | 'cosine' | Similarity metric for vector search |

API

MemoryVectorAdapter

Also exported as MemoryVectorDBAdapter (alias).

adapter.createCollection(name, dimension)     // Promise<void>
adapter.upsert(collection, docs)              // Promise<void>
adapter.search(collection, vector, topK)      // Promise<SearchResult[]>
adapter.delete(collection, ids)               // Promise<void>
adapter.getCollectionStats(name)              // Promise<{ count, dimension }>

Implements: VectorDBAdapter

Related Packages

• @orka-js/core — VectorDBAdapter interface
• @orka-js/qdrant — Qdrant for production use
• @orka-js/pgvector — PostgreSQL adapter
• orkajs — Full bundle