14_types_of_RAG

RAG Retrieval Lab: 15 retrieval-augmented generation strategies (hybrid RRF, HyDE, corrective, agentic, reranking) behind one interface, with pgvector, eval metrics and tracing; runs offline.

A learning project that puts 15 retrieval-augmented generation (RAG) strategies behind one interface so they can be compared on the same corpus, the same questions and the same metrics. Everything runs offline by default: a deterministic hashing embedder and an extractive FakeLLM replace paid APIs, so tests, CI and the eval suite are reproducible and free. Real backends plug in through adapters: Postgres + pgvector (HNSW) with full-text search (tsvector), sentence-transformers, any OpenAI-compatible endpoint or Ollama.