Databricks SDK for JavaScript
    Preparing search index...

    Interface AiExtractFunction

    Configuration for a task that runs the ai_extract function.

    The fields to extract come from exactly one of two places: an existing AI Extract agent, named by agent_id, or the inline schema and instructions. Set agent_id or the inline fields, not both.

    interface AiExtractFunction {
        enableCitations?: boolean;
        enableConfidenceScores?: boolean;
        instructions?: string;
        mode?: AiExtractFunction_Mode;
        schema?: string;
        version?: string;
    }
    Index

    Properties

    enableCitations? enableConfidenceScores? instructions? mode? schema? version?

    Properties

    enableCitations?: boolean

    When true, includes citation metadata locating each extracted value in the source.

    enableConfidenceScores?: boolean

    When true, includes a per-field confidence score in the response.

    instructions?: string

    Global description of the task and domain, passed to the model to improve extraction quality.

    mode?: AiExtractFunction_Mode

    The extraction mode. Defaults to standard extraction when unspecified.

    schema?: string

    The extraction schema defining the fields to extract, as a JSON string. Either an array of field names (e.g. ["company", "valuation"]) or an object mapping each field name to a type and an optional description. Supported types are string, integer, number, boolean, enum, object, and array; all fields are optional.

    version?: string

    The ai_extract function version to invoke (for example, 2.1).

    Settings

    Member Visibility

    On This Page

    Properties