Indexes

A vector index is declared once in application code. The index configuration is immutable after creation — changing dimensions or metric requires a new index and re-ingestion.

Required Options

name

Stable name for the index inside the Layeron app. Layeron uses the name for binding derivation and Cloudflare resource naming.

vector({ name: "docs", dimensions: 768 })

dimensions

Number of dimensions in each vector. Must match the output of the embedding model.

• Type: integer
• Range: 1 — 1536

vector({ name: "docs", dimensions: 768 })

Optional Configuration

metric

Similarity metric for distance calculations.

Value	Description
"cosine" (default)	Cosine similarity
"euclidean"	Euclidean distance
"dot_product"	Dot product similarity

vector({ name: "docs", dimensions: 768, metric: "euclidean" })

namespace

Platform namespace for the index. Defaults to "default".

Namespaces keep multiple indexes with the same name isolated from each other. They are part of the stable platform identity and affect binding and Cloudflare resource naming.

vector({ name: "docs", namespace: "search", dimensions: 768 })

binding

Worker binding name for the Cloudflare Vectorize binding. Omit to let Layeron derive a stable binding name automatically.

vector({ name: "docs", binding: "MY_BINDING", dimensions: 768 })

metadata.indexes

Metadata fields indexed for query filtering. Indexed fields let you filter query results by metadata value.

Each entry requires a propertyName and a type:

Type	Description
"string"	String metadata field
`“number”	Number metadata field
"boolean"	Boolean metadata field

vector({
  name: "docs",
  dimensions: 768,
  metadata: {
    indexes: [
      { propertyName: "tenantId", type: "string" },
      { propertyName: "sourceId", type: "string" },
    ],
  },
})

Full Example

const docs = vector({
  name: "docs",
  namespace: "search",
  dimensions: 768,
  metric: "cosine",
  metadata: {
    indexes: [
      { propertyName: "tenantId", type: "string" },
    ],
  },
})