Retrieval-Augmented Generation (RAG)

Retrieval-Augmented Generation (RAG) is a process that integrates real-time data access from Unified APIs to enhance the capability of applications in generating contextually relevant responses using data from connected SaaS platforms. The RAG pipeline involves subscribing to webhooks for updates on objects, retrieving full content from source APIs, chunking and embedding the content, and storing these embeddings in a vector database. This method allows applications to retrieve and use customer records as context rather than relying solely on pre-trained model data. While Unified facilitates real-time retrieval and normalization of data across different providers, it does not store embeddings or maintain vector indexes; these responsibilities fall on the user's infrastructure. The process involves multiple steps, including connecting data sources, subscribing to object updates, chunking and embedding content, storing data in a vector database, and retrieving and generating responses based on user queries. RAG is particularly useful for applications such as enterprise search, AI responses grounded in customer data, and CRM assistants, and it requires maintaining data freshness through real-time updates and re-embedding when necessary.