Module rerank 

Reranking and result fusion strategies

Provides RRF (Reciprocal Rank Fusion) for combining vector and FTS results, and neural reranking using cross-encoder models for improved accuracy.

Structs

FusedResult

Fused search result combining vector and FTS scores

NeuralReranker

Neural reranker using cross-encoder model

Constants

DEFAULT_RRF_K

Default RRF k parameter (per osgrep reference)

EXACT_MATCH_RRF_K

RRF k parameter for exact matches (lower = stronger boost)

Functions

rrf_fusion

rrf_fusion_with_exact

Reciprocal Rank Fusion with exact match boosting

vector_only

Simple vector-only pass-through (no fusion)