Create 3 types of RAG apps (Naive, Advanced, Modular)

[dnielsen]

• Naive RAG is a basic technique where the system retrieves relevant data based on a query, which is then simply fed into a language model (such as GPT) to generate the final answer. Basic AllyCat uses this technique.

• Advanced RAG technique provides enhancements in accuracy, adaptability and performance by using sophisticated algorithms for better retrieval and augmentation. AllyCat with GraphRAG falls under this category.

• Modular RAG is the most advanced form of RAG that breaks the RAG components into modules that work together in a pipeline-like architecture. Its key feature is that you can adapt, debug, and optimize each component of the RAG pipeline independently which provides customizability and scalability