import { OpenAssistantTool } from '@openassistant/utils';
import { z } from 'zod';
import { GetValues } from '../types';
export type DataClassifyFunctionArgs = z.ZodObject<{
    datasetName: z.ZodString;
    variableName: z.ZodString;
    method: z.ZodEnum<['quantile', 'natural breaks', 'equal interval', 'percentile', 'box', 'standard deviation', 'unique values']>;
    k: z.ZodOptional<z.ZodNumber>;
    hinge: z.ZodOptional<z.ZodNumber>;
}>;
export type DataClassifyLlmResult = {
    success: boolean;
    result?: {
        datasetName: string;
        variableName: string;
        method: string;
        k?: number;
        hinge?: number;
        breaks: number[];
    };
    error?: string;
    instruction?: string;
};
export type DataClassifyAdditionalData = {
    originalDatasetName: string;
    variableName: string;
    method: string;
    k: number;
    hinge?: number;
    breaks: number[];
};
export type DataClassifyFunctionContext = {
    getValues: GetValues;
};
/**
 * ## dataClassify Tool
 *
 * This tool is used to classify numerical data into k bins or classes using various statistical methods.
 * It returns break points that can be used to categorize continuous data into discrete intervals.
 *
 * ### Classification Methods
 *
 * The classification method can be one of the following types:
 * - **quantile**: Divides data into equal-sized groups based on quantiles
 * - **natural breaks**: Uses Jenks' algorithm to minimize within-group variance
 * - **equal interval**: Creates intervals of equal width across the data range
 * - **percentile**: Uses percentile-based breaks (25th, 50th, 75th percentiles)
 * - **box**: Uses box plot statistics (hinge = 1.5 or 3.0)
 * - **standard deviation**: Creates breaks based on standard deviation intervals
 * - **unique values**: Returns all unique values in the dataset
 *
 * ### Parameters
 * - `datasetName`: Name of the dataset containing the variable
 * - `variableName`: Name of the numerical variable to classify
 * - `method`: Classification method (see above)
 * - `k`: Number of bins/classes (required for quantile, natural breaks, equal interval)
 * - `hinge`: Hinge value for box method (default: 1.5)
 *
 * **Example user prompts:**
 * - "Can you classify the population data into 5 classes using natural breaks?"
 * - "Classify the income variable using quantile method with 4 bins"
 * - "Use box plot method to classify the housing prices"
 *
 * ### Example
 * ```typescript
 * import { dataClassify } from "@openassistant/geoda";
 * import { convertToVercelAiTool } from "@openassistant/utils";
 *
 * const classifyTool = {
 *   ...dataClassify,
 *   context: {
 *     getValues: async (datasetName: string, variableName: string) => {
 *       // Implementation to retrieve values from your data source
 *       return [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20];
 *     },
 *   },
 * };
 *
 * // Usage with AI model
 * const result = await generateText({
 *   model: yourModel,
 *   prompt: 'Can you classify the population data into 5 classes using natural breaks?',
 *   tools: { dataClassify: convertToVercelAiTool(classifyTool) },
 * });
 * ```
 */
export declare const dataClassify: OpenAssistantTool<DataClassifyFunctionArgs, DataClassifyLlmResult, DataClassifyAdditionalData, DataClassifyFunctionContext>;
export type DataClassifyTool = typeof dataClassify;
export declare function runDataClassify({ datasetName, variableName, method, k, hinge, getValues, }: {
    datasetName: string;
    variableName: string;
    method: string;
    k: number;
    hinge?: number;
    getValues: GetValues;
}): Promise<{
    llmResult: {
        success: boolean;
        result: any;
        error?: undefined;
    };
    additionalData: any;
} | {
    llmResult: {
        success: boolean;
        error: string;
        result?: undefined;
    };
    additionalData?: undefined;
}>;