Module search 

Search module for hybrid retrieval

Provides combined dense (vector) and sparse (BM25) search capabilities with Reciprocal Rank Fusion for optimal result quality.

Also includes cross-encoder reranking for improved precision, context compression for token-efficient output, query understanding for intelligent search preprocessing, and persistent index management.

Structs

ExtractedEntity

Extracted entity from query

IndexConfig

Configuration for the index manager

IndexDocument

Document to be indexed

IndexManager

Index manager for persistent skill indexing

IndexMetadata

Index metadata stored on disk

IndexStats

Statistics about the index

PipelineHealth

Health status of the pipeline

PipelineIndexStats

Statistics about indexed documents

PipelineSearchResult

Result from a search operation

ProcessedQuery

Processed query with intent, entities, and expansions

ProviderStatus

Status of an individual component

QueryExpansion

Query expansion with synonyms and related terms

QueryProcessor

Query processor for intelligent search preprocessing

SearchPipeline

Unified search pipeline that orchestrates all RAG components

SkillChecksum

Checksum for a skill to detect changes

SyncResult

Result of a sync operation

Enums

EntityType

Entity type for extracted entities

FusionMethod

Fusion method to combine search results

QueryIntent

Query intent classification

Functions

reciprocal_rank_fusion

Reciprocal Rank Fusion (RRF)

weighted_sum_fusion

Weighted sum fusion