Can I use this with SQLAlchemy?

Yes, the skill provides specific asynchronous SQLAlchemy patterns for executing hybrid search queries, including subqueries for RRF and rank calculations.

Does this skill support metadata filtering?

Yes, it includes implementation patterns for filtered search, allowing you to narrow down vector and keyword results by content type, categories, or other relational metadata.

How does Reciprocal Rank Fusion (RRF) improve search results?

RRF combines the rankings from different search methods (vector similarity and keyword matching) without needing to normalize scores, effectively surfacing results that perform well across both retrieval types.

What is the benefit of HNSW over IVFFlat in PGVector?

HNSW offers significantly faster query speeds—up to 17x faster in some benchmarks—and higher recall for datasets exceeding 100k vectors, making it the preferred choice for production environments.

PGVector Hybrid Search

Name: PGVector Hybrid Search
Author: yonatangross

byyonatangross

•

Database Management

Implements production-grade hybrid search by combining PGVector HNSW semantic retrieval with BM25 keyword matching using Reciprocal Rank Fusion.

About

This skill provides a comprehensive framework for building advanced search systems within PostgreSQL using the pgvector extension. It offers production-ready patterns for hybrid retrieval, combining the conceptual understanding of semantic vector search with the precision of keyword-based BM25. Developers can leverage optimized SQL schemas, high-performance HNSW indexing, and Reciprocal Rank Fusion (RRF) algorithms to merge results for maximum relevance. It is an essential toolkit for building RAG applications, knowledge bases, or recommendation engines that need to scale to millions of vectors while maintaining high recall and low latency.

Key Features

Optimized HNSW index configuration for millions of vectors
Pre-computed tsvector patterns for 10x faster keyword matching
29 GitHub stars
Hybrid retrieval combining semantic and keyword search
Reciprocal Rank Fusion (RRF) implementation for rank-based merging
Production-ready SQLAlchemy and raw SQL implementation examples

Use Cases

Building high-accuracy RAG applications with semantic and phrase matching
Scaling vector search performance in PostgreSQL for large document collections
Implementing faceted and filtered search with vector similarity thresholds

About

Key Features

Optimized HNSW index configuration for millions of vectors
Pre-computed tsvector patterns for 10x faster keyword matching
29 GitHub stars
Hybrid retrieval combining semantic and keyword search
Reciprocal Rank Fusion (RRF) implementation for rank-based merging
Production-ready SQLAlchemy and raw SQL implementation examples

Use Cases

Building high-accuracy RAG applications with semantic and phrase matching
Scaling vector search performance in PostgreSQL for large document collections
Implementing faceted and filtered search with vector similarity thresholds