How does it help with RAG performance?

It provides templates for over-fetching and reranking results with cross-encoders, and offers guidance on choosing the right index types like HNSW for low-latency retrieval.

Which vector databases does this skill support?

The skill includes patterns for popular vector stores including Pinecone, Qdrant, and PostgreSQL via the pgvector extension.

Which distance metrics are covered?

The skill covers Cosine similarity for normalized embeddings, Euclidean (L2) for raw embeddings, and Dot Product for cases where vector magnitude is important.

Does it support hybrid search?

Yes, it includes implementation patterns for combining dense semantic search with sparse keyword search to improve retrieval accuracy.

What is the Similarity Search Patterns skill?

It is a specialized Claude Code skill that provides production-ready implementation templates and best practices for building vector-based search systems.

Similarity Search Patterns

Name: Similarity Search Patterns
Author: as4584

byas4584

Data Science & ML

Implements efficient semantic search and vector database patterns for production-grade retrieval systems.

About

This skill provides standardized patterns for implementing similarity search across leading vector databases including Pinecone, Qdrant, and pgvector. It equips Claude with the expertise to select appropriate distance metrics such as Cosine or Euclidean, choose optimal indexing strategies like HNSW or IVF+PQ, and implement advanced retrieval techniques including cross-encoder reranking and hybrid search. Whether you are building a RAG pipeline, a recommendation engine, or a large-scale semantic search tool, this skill ensures scalable, low-latency implementations optimized for both performance and recall.

Key Features

Multi-provider support for Pinecone, Qdrant, and PostgreSQL (pgvector)
Implementation patterns for major distance metrics and indexing types
Hybrid search patterns combining dense and sparse vector lookups
0 GitHub stars
Advanced retrieval strategies including cross-encoder reranking
Optimization guidance for HNSW and quantized indexing strategies

Use Cases

Building high-performance Retrieval Augmented Generation (RAG) systems
Implementing personalized recommendation systems using vector embeddings
Developing semantic search engines for large-scale document collections

About

Key Features

Multi-provider support for Pinecone, Qdrant, and PostgreSQL (pgvector)
Implementation patterns for major distance metrics and indexing types
Hybrid search patterns combining dense and sparse vector lookups
0 GitHub stars
Advanced retrieval strategies including cross-encoder reranking
Optimization guidance for HNSW and quantized indexing strategies

Use Cases

Building high-performance Retrieval Augmented Generation (RAG) systems
Implementing personalized recommendation systems using vector embeddings
Developing semantic search engines for large-scale document collections