vector-db

Vector database and RAG patterns — embedding models (OpenAI text-embedding-3, Cohere, Voyage AI, sentence-transformers, Ollama local models, chunking strategies, token counting), pgvector (PostgreSQL extension setup, HNSW/IVFFlat indexes, cosine/L2/IP similarity search, hybrid search with full-text, metadata filtering, Drizzle/Prisma/node-postgres usage), vector search fundamentals (distance metrics, ANN algorithms, filtering strategies, cross-encoder re-ranking, Cohere rerank, Reciprocal Rank Fusion, evaluation metrics), RAG patterns (retrieval-augmented generation pipelines, query rewriting, HyDE, multi-step retrieval, parent document retrieval, context assembly, citation, ingestion pipelines, LLM-as-judge evaluation), Pinecone (serverless/pod indexes, batch upsert, metadata filtering, namespaces, hybrid search, index management), and ChromaDB/Weaviate (local embedded DB, collections, custom embedding functions, Weaviate schema, hybrid BM25+vector search, multi-tenancy). 6 skills, 3 commands, 1 agent. No dependencies.