Function Calling for LLMs: RAG without a Vector Database

Explore the concept of extending Retrieval-Augmented Generation (RAG) with Function Calling to access structured and tabular data in this 39-minute conference talk from MLOps World: Machine Learning in Production. Delivered by Jim Dowling, CEO of Hopsworks, the presentation delves into techniques for enriching tables with metadata and examines the potential for performing expressive queries effectively. Learn about function calling in the context of queries to the Hopsworks feature store, which supports extensive metadata and statistics for columns and tables (feature groups) to enhance function calling performance. Compare the application of function calling in both cloud-hosted Large Language Models (LLMs) like GPT-4 and private LLMs such as Hermes-2, a fine-tuned version of the Mistral 7B model. Gain insights into implementing RAG without relying on a vector database, opening up new possibilities for leveraging LLMs in data-rich environments.