Skip to content

config

Configuration for the AI Search plugin: the typed SearchPluginConfig (the plugin’s slice of AppKit config), the JSON Schema the manifest publishes for it, and resolveSearchConfig which layers that config over environment defaults into the concrete ResolvedSearchConfig the runtime, client, tools, and routes read.

The design goal is “one index, zero config”: name a default index (or set DATABRICKS_VECTOR_SEARCH_INDEX) and everything else - the columns to return, the page size, the match mode, the mounted route path, the embedding model used when creating an index - has a sensible default that can be overridden when a deployment needs to go deeper.

Which index / endpoint is used follows the standard precedence: explicit plugin config, then environment, then a default.

Env fallbacks: DATABRICKS_VECTOR_SEARCH_INDEX (the default index), SEARCH_INDEX, SEARCH_ENDPOINT, SEARCH_COLUMNS, SEARCH_PAGE_SIZE, SEARCH_MODE, SEARCH_EMBEDDING_MODEL, SEARCH_TIMEOUT_MS, SEARCH_WRITE.