@dreamlake/dreamdb

TypeScript driver for DreamDB — a storage and retrieval protocol for multimodal signals anchored to a shared timeline.

Install

npm install @dreamlake/dreamdb

Quick start

import { Dataset, Schema, MemoryBackend } from '@dreamlake/dreamdb'

// Define a schema
const schema = new Schema()
schema.addEmbedding('embedding', 16)
schema.addImage('image', 'jpeg')
schema.addScalarCategorical('label')

// Create a dataset (in-memory for testing, or use S3Backend for production)
const backend = new MemoryBackend()
const ds = await Dataset.create('my-dataset', schema, backend)

// Append samples
await ds.appendMany([
  { embedding: new Float32Array([...]), image: new Uint8Array([...]), label: 'cat' },
  { embedding: new Float32Array([...]), image: new Uint8Array([...]), label: 'dog' },
])

// Vector query — find the 5 nearest neighbors
const batches = await ds.iterVector({
  field: 'embedding',
  query: new Float32Array([...]),
  topK: 5,
})

Browser usage

import { S3Backend, Space } from '@dreamlake/dreamdb'

const backend = new S3Backend('https://your-bucket.s3.amazonaws.com')
const space = new Space(backend)

// Resolve a ref to its manifest
const manifest = await space.resolveRef('my-dataset')
console.log(manifest.tracks)

Documentation

• Installation
• TypeScript SDK Reference
• Protocol Specification
• Browser Demo

License

MIT