@juspay/neurolink/dist/lib/rag/index.d.ts

Universal AI Development Platform with working MCP integration, multi-provider support, voice (TTS/STT/realtime), and professional CLI. 58+ external MCP servers discoverable, multimodal file processing, RAG pipelines. Build, test, and deploy AI applicatio

/**
 * RAG Document Processing Module
 *
 * Provides comprehensive RAG (Retrieval-Augmented Generation) capabilities:
 * - Document loading (text, markdown, HTML, JSON, CSV, PDF, web)
 * - MDocument class for fluent document processing
 * - 10 chunking strategies (character, recursive, sentence, token, markdown, html, json, latex, semantic, semantic-markdown)
 * - LLM-powered metadata extraction (title, summary, keywords, Q&A)
 * - Vector query tools with metadata filtering and reranking
 * - Hybrid search (BM25 + vector fusion)
 * - Graph RAG for knowledge graph-based retrieval
 * - RAG pipeline orchestration
 * - Context assembly and formatting
 * - ChunkerFactory and ChunkerRegistry patterns for extensibility
 * - Error handling and resilience (CircuitBreaker, RetryHandler)
 *
 * @example
 * ```typescript
 * import {
 *   MDocument,
 *   loadDocument,
 *   RAGPipeline,
 *   ChunkerRegistry,
 *   ChunkerFactory,
 *   CircuitBreaker
 * } from '@juspay/neurolink';
 *
 * // Load and process a document
 * const doc = await loadDocument('/path/to/document.md');
 * await doc.chunk({ strategy: 'markdown', config: { maxSize: 1000 } });
 * await doc.embed('openai', 'text-embedding-3-small');
 *
 * // Or use the full RAG pipeline
 * const pipeline = new RAGPipeline({
 *   embeddingModel: { provider: 'openai', modelName: 'text-embedding-3-small' },
 *   generationModel: { provider: 'openai', modelName: 'gpt-4o-mini' }
 * });
 * await pipeline.ingest(['/path/to/docs/*.md']);
 * const response = await pipeline.query('What are the key features?');
 *
 * // Use factory pattern for chunker creation
 * const chunker = await ChunkerFactory.createChunker('semantic', { maxSize: 500 });
 * const chunks = await chunker.chunk(text);
 * ```
 */
export { ChunkerFactory, chunkerFactory, createChunker, getAvailableStrategies as getFactoryStrategies, getDefaultConfig as getFactoryDefaultConfig, } from "./ChunkerFactory.js";
export { ChunkerRegistry as ChunkerRegistryV2, chunkerRegistry, getAvailableChunkers, getChunker, getChunkerMetadata, } from "./ChunkerRegistry.js";
export * from "./chunkers/index.js";
export { CharacterChunker, ChunkerRegistry, chunkText, HTMLChunker, JSONChunker, LaTeXChunker, MarkdownChunker, RecursiveChunker, SemanticChunker, SentenceChunker, TokenChunker, } from "./chunking/index.js";
export { CSVLoader, HTMLLoader, JSONLoader, loadDocument, loadDocuments, MarkdownLoader, MDocument, PDFLoader, TextLoader, WebLoader, } from "./document/index.js";
export * from "./errors/index.js";
export { GraphRAG } from "./graphRag/index.js";
export { createMetadataExtractor, extractMetadata, getAvailableExtractors, getAvailableExtractorTypes, getExtractor, getExtractorDefaultConfig, getExtractorMetadata, getRegisteredExtractorMetadata, LLMMetadataExtractor, MetadataExtractorFactory, MetadataExtractorRegistry, metadataExtractorFactory, metadataExtractorRegistry, } from "./metadata/index.js";
export { assembleContext, createContextWindow, createRAGPipeline, extractKeySentences, formatContextWithCitations, orderByDocumentStructure, RAGPipeline, summarizeContext, } from "./pipeline/index.js";
export { prepareRAGTool } from "./ragIntegration.js";
export { batchRerank, CohereRelevanceScorer, CrossEncoderReranker, createReranker, getAvailableRerankers, getAvailableRerankerTypes, getRegisteredRerankerMetadata, getReranker, getRerankerDefaultConfig, getRerankerMetadata, RerankerFactory, RerankerRegistry, rerank, rerankerFactory, rerankerRegistry, simpleRerank, } from "./reranker/index.js";
export * from "./resilience/index.js";
export { createHybridSearch, createVectorQueryTool, InMemoryBM25Index, InMemoryVectorStore, linearCombination, reciprocalRankFusion, } from "./retrieval/index.js";
export * from "../types/index.js";
import type { Chunk, ChunkingStrategy, ExtractParams } from "../types/index.js";
/**
 * Process a document through the full RAG pipeline
 *
 * @param text - Document text to process
 * @param options - Processing options
 * @returns Processed chunks with optional metadata
 */
export declare function processDocument(text: string, options?: {
    /** Chunking strategy (default: recursive) */
    strategy?: ChunkingStrategy;
    /** Maximum chunk size */
    maxSize?: number;
    /** Chunk overlap */
    overlap?: number;
    /** Metadata extraction options */
    extract?: ExtractParams;
    /** Provider for metadata extraction */
    provider?: string;
    /** Model for metadata extraction */
    model?: string;
    /** Custom metadata to add */
    metadata?: Record<string, unknown>;
}): Promise<Chunk[]>;
/**
 * Get recommended chunking strategy based on content type
 *
 * @param contentType - MIME type or file extension
 * @returns Recommended chunking strategy
 */
export declare function getRecommendedStrategy(contentType: string): ChunkingStrategy;
/**
 * Get available chunking strategies
 *
 * @returns Array of available strategy names
 */
export declare function getAvailableStrategies(): ChunkingStrategy[];
/**
 * Get default configuration for a chunking strategy
 *
 * @param strategy - Chunking strategy name
 * @returns Default configuration object
 */
export declare function getDefaultChunkerConfig(strategy: ChunkingStrategy): Record<string, unknown>;