Module fusion 

RRF (Reciprocal Rank Fusion) utilities shared between hybrid-search and deep-research.

The formula used is the canonical RRF score:

score(d) = sum_over_lists { weight * 1 / (rrf_k + rank(d)) }

where rank is 1-indexed position in each ordered list. The map returned by rrf_fuse contains un-normalised scores; callers that need a [0,1] range should divide by the theoretical maximum:

max_possible = sum_over_lists { weight * 1 / (rrf_k + 1) }

Functions

rrf_fuse

Fuse multiple ranked lists of integer IDs via Reciprocal Rank Fusion.

rrf_max_possible

Compute the theoretical maximum RRF score for a given set of weights and rrf_k.