groovy-web/rag-system-pgvector

Combines semantic and keyword search with intelligent reranking, supports multiple embedding providers (OpenAI, Cohere, local models), and includes streaming responses with conversation memory. The system uses pgvector's IVFFlat indexing for efficient similarity queries, integrates result reranking to improve relevance, and provides TypeScript interfaces with comprehensive monitoring and caching layers. Built for horizontal scaling with connection pooling and Redis support, it targets Node.js applications requiring production-grade RAG capabilities.