Module hybrid 

Hybrid Search with RRF Fusion (Task 4)

This module combines vector similarity search (ANN) with lexical search (BM25) using Reciprocal Rank Fusion (RRF) for score combination.

RRF Algorithm

RRF_score(d) = Σ weight_i / (k + rank_i(d))

Where:

• k is typically 60 (robust default)
• rank_i(d) is the rank of document d in result list i (1-indexed)
• weight_i is the weight for result list i

Pipeline

                   Query
                     │
          ┌──────────┴──────────┐
          │                     │
          ▼                     ▼
   ┌─────────────┐       ┌─────────────┐
   │   Vector    │       │   Lexical   │
   │   Search    │       │   Search    │
   │   (HNSW)    │       │   (BM25)    │
   └──────┬──────┘       └──────┬──────┘
          │                     │
          │  [(id, score), ...]│  [(id, score), ...]
          │                     │
          └──────────┬──────────┘
                     │
                     ▼
              ┌─────────────┐
              │  RRF Fusion │
              └──────┬──────┘
                     │
                     ▼
              ┌─────────────┐
              │   Filter    │
              │  (optional) │
              └──────┬──────┘
                     │
                     ▼
              ┌─────────────┐
              │   Top-K     │
              └─────────────┘

Structs

ComponentScores

Individual component scores

HybridSearchEngine

A combined vector + lexical search engine

RRFConfig

Configuration for Reciprocal Rank Fusion

RRFFusion

Reciprocal Rank Fusion combiner

SearchResult

Search result with score

Traits

LexicalSearchBackend

Trait for lexical search backends

VectorSearchBackend

Trait for vector search backends

Functions

filter_results

Post-filter results by metadata predicate

Type Aliases

DocId

Document ID type