Very similar to AutoRAG, but less focused on using the community as a testbed, more for feedback. Their tool allows people to use ready-to-use templates of RAG pipelines, run experiments against their own data, evaluate performance, and tweak as needed. Really cool project, and they point out that “building a reliable RAG pipeline seems increasingly like navigating a maze with so many techniques…super overwhelming to try out multiple chunkers, embeddings, retrievers—trying to figure out the best RAG setup for your data and use-case.”