Skip to content

Interface: SearchPluginConfig

The AI Search plugin’s slice of AppKit config. Every field is optional - the plugin works with nothing but an index name in the environment.

  • BasePluginConfig

[key: string]: unknown

optional allowWrite?: boolean

Enable the write surface (add_documents tool + POST /documents). Off by default.


optional columns?: string[]

Default columns to return per hit. Falls back to each index’s own columns.


optional embeddingModel?: string

Embedding model endpoint used when the plugin CREATES a Delta Sync index. Fuzzy-resolved.


optional endpoint?: string

The Vector Search serving endpoint indexes live on (only needed to create an index).


optional ensureOnSetup?: EnsureOnSetupConfig

Provision a REAL index at boot: ensure the endpoint + index exist and seed documents when the index is empty. Runs in the background during setup() (a slow first-time endpoint/index build never blocks the server) using the boot-time SDK auth (env / config profile). Idempotent - safe every boot.

The default is a MANAGED direct-access index (Databricks embeds the text column), so documents are plain rows and search-by-text works with no Delta table, no warehouse, and no vectors to compute. Point it at a sourceTable to provision a Delta Sync index instead.


optional host?: string

BasePluginConfig.host


optional index?: string

The default index searched when a request omits one (name or alias).


optional indexes?: (string | SearchIndexConfig)[]

Indexes the plugin should know about (for aliases, universal search, and the UI catalogue).


optional mode?: "hybrid" | "vector" | "keyword"

Default match mode.


optional name?: string

BasePluginConfig.name


optional pageSize?: number

Default number of hits a search returns.


optional telemetry?: TelemetryOptions

BasePluginConfig.telemetry


optional timeoutMs?: number

Per-call timeout in milliseconds.