Skip to content

Interface: LakebaseAiSearchIndexConfig

Per-alias Lakebase search configuration.

  • Omit<IndexConfig, "embeddingFn" | "endpointName" | "pagination" | "reranker">

optional auth?: "service-principal" | "on-behalf-of-user"

Auth mode for the built-in HTTP routes — “service-principal” (default) uses the app’s SP, “on-behalf-of-user” proxies the logged-in user’s token. Programmatic callers select per call via appkit.aiSearch.asUser(req).

Omit.auth


optional columns?: string[]

Columns to return in results. Optional: in development the plugin auto-discovers them from the index’s source table when omitted (and warns that they should be set explicitly for production).

Omit.columns


optional documents?: Record<string, unknown>[]

Documents used to seed an empty full-text table during setup.


optional indexName?: string

Three-level UC name: catalog.schema.index_name. Defaults to the DATABRICKS_VS_INDEX_NAME env var when omitted — so multiple aliases that omit it all resolve to that same physical index. Set it explicitly per alias when they should point at distinct indexes.

Omit.indexName


optional numResults?: number

Max results per query

Omit.numResults


optional queryType?: SearchQueryType

Default search mode

Omit.queryType


optional textColumn?: string

Document field indexed first. Defaults to text.