Skip to content
dbx tools

Interface: CreateIndexOptions

Options for SearchClient.createIndex / SearchClient.ensureIndex. The goal is the same as the rest of the client: name a source table and a text column, and everything else - the endpoint, the embedding model, the index type, the sync mode - has a sensible default that infers from the workspace, overridable when a deployment needs to go deeper.

Extended by

Section titled “Extended by”

Properties

Section titled “Properties”

columnsToSync?

Section titled “columnsToSync?”

optional columnsToSync?: string[]

Extra columns to sync alongside the embedding source (Delta Sync).

embeddingDimension?

Section titled “embeddingDimension?”

optional embeddingDimension?: number

For a direct-access index: the embedding vector dimension (self-managed vectors, no source table).

embeddingModel?

Section titled “embeddingModel?”

optional embeddingModel?: string

Embedding model endpoint. A loose name is fuzzy-matched; when omitted the best embedding endpoint in the workspace is chosen (resolveEmbeddingModel).

embeddingSourceColumn?

Section titled “embeddingSourceColumn?”

optional embeddingSourceColumn?: string

The text column embeddings are computed from (Delta Sync). Defaults to the first of text / content / body present, else the caller must set it.

embeddingVectorColumn?

Section titled “embeddingVectorColumn?”

optional embeddingVectorColumn?: string

For a direct-access index: the column the vector is stored in. Defaults to embedding.

endpoint?

Section titled “endpoint?”

optional endpoint?: string

Vector Search endpoint to host the index on. Defaults to the plugin’s endpoint.

pipelineType?

Section titled “pipelineType?”

optional pipelineType?: "TRIGGERED" | "CONTINUOUS"

Sync mode for a Delta Sync index. TRIGGERED (default) syncs on demand; CONTINUOUS keeps fresh.

primaryKey?

Section titled “primaryKey?”

optional primaryKey?: string

Primary-key column. Defaults to id.

schema?

Section titled “schema?”

optional schema?: Record<string, string>

The column-name -> type map used to build a direct-access index’s schema_json (types: string, int, long, float, double, boolean, date, timestamp). Must include the primary key and any columns you upsert. Defaults to { id: "string", text: "string" }. The embedding vector column is added automatically.

signal?

Section titled “signal?”

optional signal?: AbortSignal

External cancellation.

sourceTable?

Section titled “sourceTable?”

optional sourceTable?: string

Source Delta table (catalog.schema.table) for a Delta Sync index. Provide this for the common case: Databricks computes and syncs embeddings from it. Omit it (and pass embeddingDimension) to create a direct-access index you write vectors to yourself.