Implements efficient similarity search and vector database patterns for semantic retrieval and RAG systems.
This skill provides standardized implementation patterns for building high-performance similarity search systems using leading vector databases like Pinecone, Qdrant, and pgvector. It guides developers through selecting the optimal distance metrics and index types—such as HNSW for speed or IVF+PQ for scale—while providing production-ready templates for upserting embeddings, performing hybrid searches, and implementing re-ranking logic. It is an essential resource for developers building LLM-powered applications that require fast, accurate, and scalable data retrieval.
Key Features
010 GitHub stars
02Expert guidance on distance metrics like Cosine, Euclidean, and Dot Product
03Hybrid search patterns combining dense vectors with keyword filtering
04Advanced re-ranking logic implementation using cross-encoders
05Ready-to-use templates for Pinecone, Qdrant, and pgvector integrations
06Optimized index configurations including HNSW and Scalar Quantization
Use Cases
01Building semantic search engines for multi-million document repositories
02Creating recommendation systems based on high-dimensional vector similarity
03Developing Retrieval-Augmented Generation (RAG) pipelines for AI agents
