import { OpenAssistantTool } from '@openassistant/utils';
import { z } from 'zod';
import { LocalMoranResult } from '@geoda/lisa';
import { GetGeometries, GetValues } from '../types';
export type LisaFunctionArgs = z.ZodObject<{
    method: z.ZodEnum<['localMoran', 'localGeary', 'localG', 'localGStar', 'quantileLisa']>;
    weightsID: z.ZodOptional<z.ZodString>;
    variableName: z.ZodString;
    multiVariableNames: z.ZodOptional<z.ZodArray<z.ZodString>>;
    biVariableNames: z.ZodOptional<z.ZodArray<z.ZodString>>;
    permutation: z.ZodOptional<z.ZodNumber>;
    significanceThreshold: z.ZodOptional<z.ZodNumber>;
    datasetName: z.ZodString;
    k?: z.ZodOptional<z.ZodNumber>;
    quantile?: z.ZodOptional<z.ZodNumber>;
    mapBounds?: z.ZodOptional<z.ZodArray<z.ZodNumber>>;
}>;
export type LisaLlmResult = {
    success: boolean;
    globalMoranI?: number;
    clusterColorAndLabels?: Array<{
        value: number;
        label: string;
        color: string;
        numberOfObservations: number;
    }>;
    error?: string;
    instructions?: string;
    datasetName?: string;
};
export type LisaAdditionalData = Partial<LocalMoranResult> & {
    datasetName: string;
    originalDatasetName: string;
    significanceThreshold: number;
};
export type LisaFunctionContext = {
    getValues: GetValues;
    getGeometries?: GetGeometries;
};
/**
 * ## lisa Tool
 *
 * This tool applies Local Indicators of Spatial Association (LISA) statistics to identify local clusters and spatial outliers.
 * It helps detect areas where similar values cluster together or where outliers exist in the spatial distribution.
 *
 * ### LISA Methods
 *
 * The LISA method can be one of the following types:
 * - **localMoran**: Local Moran's I statistic for detecting clusters and outliers
 * - **localGeary**: Local Geary's C statistic for detecting spatial heterogeneity
 * - **localG**: Local Getis-Ord G statistic for detecting hot and cold spots
 * - **localGStar**: Local Getis-Ord G* statistic (includes the observation itself)
 * - **quantileLisa**: Quantile-based LISA analysis
 *
 * ### Parameters
 * - `datasetName`: Name of the dataset containing the variable
 * - `variableName`: Name of the numerical variable to analyze
 * - `method`: LISA method to use (see above)
 * - `weightsID`: ID of spatial weights matrix (optional, will use cached weights if available)
 * - `permutation`: Number of permutations for significance testing (default: 999)
 * - `significanceThreshold`: Significance threshold for filtering results (default: 0.05)
 * - `k`: Number of quantiles for quantile LISA (required for quantileLisa method)
 * - `quantile`: Quantile value for quantile LISA (required for quantileLisa method)
 *
 * **Example user prompts:**
 * - "Are young population clustering over the zipcode areas?"
 * - "Can you perform a local Moran's I analysis on the population data?"
 * - "What are the local clusters in the population data?"
 * - "How many significant clusters are there in the population data?"
 *
 * ### Example
 * ```typescript
 * import { lisa } from "@openassistant/geoda";
 * import { convertToVercelAiTool } from "@openassistant/utils";
 *
 * const lisaTool = {
 *   ...lisa,
 *   context: {
 *     getValues: async (datasetName: string, variableName: string) => {
 *       // Implementation to retrieve values from your data source
 *       return [100, 200, 150, 300, 250, 180, 220, 190, 280, 210];
 *     },
 *   },
 * };
 *
 * const result = await generateText({
 *   model: openai('gpt-4.1', { apiKey: key }),
 *   prompt: 'Can you perform a local Moran analysis on the population data?',
 *   tools: { lisa: convertToVercelAiTool(lisaTool) },
 * });
 * ```
 *
 * :::note
 * The LISA tool should always be used with the spatialWeights tool. The LLM models know how to use the spatialWeights tool for the LISA analysis.
 * :::
 */
export declare const lisa: OpenAssistantTool<LisaFunctionArgs, LisaLlmResult, LisaAdditionalData, LisaFunctionContext>;
export type LisaTool = typeof lisa;