The Capella iQ reference architecture integrates advanced database solutions with generative AI models, utilizing Couchbase NoSQL database for efficient data handling and retrieval. The architecture leverages multiple storage solutions, including Docs DB, Object Storage, and Vector DB, to handle different types of data efficiently. It also includes an orchestrator that manages the flow of data and operations within the system, embeddings that transform textual data into numerical vectors, context and prompt builders that prepare contextual prompts for AI models, sensitive data filters that ensure compliance with data privacy regulations, LLM backend proxies that interact with large language models, moderation and rate limiting components that enforce guidelines and prevent overloading, metrics and logging systems that monitor and optimize system efficiency, public and private LLMs that provide flexibility in choosing the best model for specific use cases. The architecture's workflow begins with an NLU query input from the user interface and ends with generation of content, following a generic step-by-step process that involves prompt building, context building, sensitive data filtering, model selection and proxy, moderation and rate limiting, response generation, response building, and response delivery.