VectorInstitute/retrieval-augmented-generation

Implements five distinct RAG pipeline examples—web search, document QA, SQL databases, cloud storage, and biomedical literature—using LangChain and LlamaIndex to demonstrate core techniques including chunking, embeddings, sparse/dense retrieval, and reranking. The PubMed implementation provides a complete end-to-end workflow with evaluation metrics via the Ragas framework, while the evaluation module focuses on assessing RAG pipeline quality through test set construction.