When should I use Reciprocal Rank Fusion (RRF)?

RRF is best used when you need to combine ranked results from multiple search engines without needing to manually normalize their different internal scoring scales.

Does this skill support PostgreSQL implementations?

Yes, it includes specific SQL and Python templates for combining PostgreSQL pgvector results with native tsvector full-text search capabilities.

How does cross-encoder re-ranking work here?

The skill provides a pattern where hybrid search retrieves an initial set of candidates, which are then passed through a neural cross-encoder for final high-precision ranking.

What is hybrid search in LLM applications?

Hybrid search combines semantic vector search (understanding meaning) with traditional keyword search (matching exact terms) to provide the most relevant results for RAG systems.

Hybrid Search Implementation

Name: Hybrid Search Implementation
Author: wshobson

bywshobson

•

23,194

データサイエンスとML

Combines vector similarity and keyword-based search to improve retrieval accuracy and recall in RAG systems.

概要

The hybrid-search-implementation skill provides standardized patterns for merging semantic vector search with traditional lexical matching. It is specifically designed for developers building Retrieval-Augmented Generation (RAG) systems or search engines where neither method alone provides sufficient recall. By utilizing fusion techniques like Reciprocal Rank Fusion (RRF) and linear weighting, this skill ensures Claude can implement robust search architectures that handle both conceptual queries and specific technical terminology, including specialized templates for PostgreSQL and Elasticsearch.

主な機能

Elasticsearch hybrid query configurations with dense vector support
Cross-encoder re-ranking patterns for high-precision result filtering
Normalization strategies for combining disparate search scoring systems
23,194 GitHub stars
Reciprocal Rank Fusion (RRF) and linear combination scoring templates
PostgreSQL integration utilizing pgvector and full-text search indexes

ユースケース

Building production-grade RAG systems that require high recall for technical terms
Implementing efficient 'search-then-rerank' pipelines for large-scale datasets
Improving search relevance for domain-specific vocabulary and exact-match codes

概要

主な機能

Elasticsearch hybrid query configurations with dense vector support
Cross-encoder re-ranking patterns for high-precision result filtering
Normalization strategies for combining disparate search scoring systems
23,194 GitHub stars
Reciprocal Rank Fusion (RRF) and linear combination scoring templates
PostgreSQL integration utilizing pgvector and full-text search indexes

ユースケース

Building production-grade RAG systems that require high recall for technical terms
Implementing efficient 'search-then-rerank' pipelines for large-scale datasets
Improving search relevance for domain-specific vocabulary and exact-match codes