RAG Pipeline Toolkit

A complete reference architecture and working code for building Retrieval-Augmented Generation pipelines over your own documents. Includes drop-in Python (pgvector + OpenAI + Anthropic) and TypeScript (Pinecone) implementations, chunking strategies that respect semantic boundaries, query-side improvements ranked by ROI (hybrid search, query rewriting, HyDE, re-ranking), an evaluation harness with recall@k and MRR, and a failure-mode table that maps symptoms to fixes. Skips the marketing fluff and gives you the patterns experienced teams actually use. Provider-agnostic — swap embedder, vector DB, or generator without rewriting the pipeline. Includes cost notes and an iteration checklist so you stop tuning blind.