Can I implement hybrid search with these patterns?

Yes, the skill includes templates for combining dense vector search with sparse vectors or keyword filtering for improved retrieval accuracy.

How do I choose the right distance metric?

The skill includes a guide explaining that Cosine is best for normalized embeddings, Euclidean for raw embeddings, and Dot Product when vector magnitude matters.

Which vector databases are supported by this skill?

The skill provides ready-to-use implementation templates for Pinecone, Qdrant, and PostgreSQL (via pgvector).

Does this skill help with RAG performance?

Yes, it includes advanced patterns for RAG, such as result reranking using cross-encoders and efficient metadata filtering.

Similarity Search Patterns

Name: Similarity Search Patterns
Author: rajasekarm

byrajasekarm

0•

Data Science & ML

Implements high-performance similarity search and vector database patterns for RAG, semantic search, and recommendation systems.

This skill provides standardized patterns and implementation templates for building efficient similarity search systems using popular vector databases like Pinecone, Qdrant, and PostgreSQL with pgvector. It guides developers through selecting the right distance metrics, choosing optimal index types like HNSW or IVF, and implementing advanced techniques such as hybrid search, reranking with cross-encoders, and metadata filtering. Whether you are scaling to millions of vectors or optimizing search latency for a RAG pipeline, these patterns provide the production-ready logic needed for robust retrieval.

Key Features

01Comprehensive guide for selecting distance metrics like Cosine and L2

020 GitHub stars

03Optimized indexing strategies including HNSW and quantization patterns

04Standardized templates for Pinecone, Qdrant, and pgvector integrations

05Advanced retrieval workflows featuring cross-encoder reranking

06Metadata filtering and hybrid search implementation patterns

Use Cases

01Creating scalable recommendation systems based on vector similarity

02Building Retrieval-Augmented Generation (RAG) pipelines for LLM applications

03Developing semantic search engines that understand context and intent

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rajasekarm/skills similarity-search-patterns

For use in Claude.ai and ChatGPT

Download Skill