@astermind/astermind-pro/dist/retrieval/index-builder.d.ts

Astermind Pro - Premium ML Toolkit with Advanced RAG, Reranking, Summarization, and Information Flow Analysis

import { type SparseVec } from './vectorization.js';
import type { Kernel } from '../types.js';
export interface IndexState {
    vocabMap: Map<string, number>;
    idf: number[];
    tfidfDocs: SparseVec[];
    landmarksIdx: number[];
    landmarkMat: Float64Array[];
    denseDocs: Float64Array[];
}
export interface BuildIndexOptions {
    chunks: Array<{
        heading: string;
        content: string;
    }>;
    vocab: number;
    landmarks: number;
    headingW: number;
    useStem: boolean;
    kernel: Kernel;
    sigma: number;
}
/**
 * Build vocabulary and IDF from chunks
 */
export declare function buildVocabAndIdf(chunks: Array<{
    heading: string;
    content: string;
}>, vocabSize: number, useStem: boolean): {
    vocabMap: Map<string, number>;
    idf: number[];
};
/**
 * Build TF-IDF vectors for all chunks
 */
export declare function buildTfidfDocs(chunks: Array<{
    heading: string;
    content: string;
}>, vocabMap: Map<string, number>, idf: number[], headingW: number, useStem: boolean): SparseVec[];
/**
 * Build Nyström landmarks from TF-IDF documents
 */
export declare function buildLandmarks(tfidfDocs: SparseVec[], vocabSize: number, numLandmarks: number): {
    landmarksIdx: number[];
    landmarkMat: Float64Array[];
};
/**
 * Build dense projections for all TF-IDF documents
 */
export declare function buildDenseDocs(tfidfDocs: SparseVec[], vocabSize: number, landmarkMat: Float64Array[], kernel: Kernel, sigma: number): Float64Array[];
/**
 * Build complete index from chunks
 */
export declare function buildIndex(opts: BuildIndexOptions): IndexState;
//# sourceMappingURL=index-builder.d.ts.map