Why use hybrid search instead of just vector search?

Vector search is great for semantic meaning but often fails on exact matches like product IDs, names, or specific technical terms. Hybrid search combines both to ensure high recall across all query types.

What is Reciprocal Rank Fusion (RRF)?

RRF is a method for combining multiple ranked result lists into a single ranking without needing to normalize different scoring scales (like BM25 vs. Cosine Similarity).

How does re-ranking fit into hybrid search?

Hybrid search identifies a broad set of candidates, while a cross-encoder re-ranker (included in the templates) can be used to perform a final, more expensive pass to ensure the absolute most relevant results are at the top.

Can I use this with PostgreSQL?

Yes, this skill includes specific templates for implementing hybrid search using the pgvector extension alongside PostgreSQL's native full-text search (TSVector).

Hybrid Search Implementation

Name: Hybrid Search Implementation
Author: sla-te

bysla-te

0•

Data Science & ML

Combines vector similarity and keyword-based search to optimize retrieval accuracy for RAG systems and search engines.

This skill provides standardized patterns and implementation templates for hybrid search, a critical architecture for high-performance Retrieval-Augmented Generation (RAG) applications. It enables developers to bridge the gap between semantic understanding (vector search) and exact keyword matching (BM25/FTS), ensuring high recall for domain-specific vocabulary, technical codes, and names. By providing ready-to-use logic for Reciprocal Rank Fusion (RRF), linear scoring, and database-specific implementations for PostgreSQL and Elasticsearch, this skill helps developers build retrieval systems that are significantly more robust than single-method approaches.

Key Features

01Reciprocal Rank Fusion (RRF) and linear scoring algorithms for result blending

02Elasticsearch patterns for combining dense vectors with BM25 keyword matching

03Normalized scoring methods for balancing semantic and lexical search weights

04Cross-encoder re-ranking templates for secondary high-precision filtering

05PostgreSQL integration templates using pgvector and full-text search (FTS)

060 GitHub stars

Use Cases

01Enhancing RAG systems to handle technical jargon and exact product codes

02Building enterprise search engines that require both semantic and keyword precision

03Optimizing document retrieval performance in PostgreSQL and Elasticsearch backends

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sla-te/copilot-converter hybrid-search-implementation

For use in Claude.ai and ChatGPT

Download Skill