Developing a Retrieval Augmented Generation (RAG) with Vector Database Functionality for Generative and Custom AI and ML using Oracle APEX Rapid Application Low Code Development

Abstract

The developer-facing user interface of the prototypical Generative AI system is called the “APEX Assistant” within the Oracle APEX application development platform. In the context of this research project, the APEX Assistant is enabled for every SQL code editor; wherever the developer would have previously been asked to write SQL code, the APEX Assistant can be used to generate SQL from a textual (or spoken) description of the desired result. Application of Retrieval Augmented Generation (RAG) aims to improve the quality of the response by using an external datastore (like an Oracle Database with vector database functionality) at inference time to build a richer prompt that includes some combination of context, history and relevant “knowledge”. The diagram below illustrates where RAG is used for the APEX Assistant. Evaluating the APEX Assistant Text to SQL functionality from the APEX code editor using a range of test cases. The tests were performed on an HR data set that can be installed from APEX SQL Workshop –> Utilities –> Data Sets. Overall, the potential results will be an open door for further development in generative AI with custom machine learning and software development to see a lot of value in text to SQL generation. Developers will continue to carefully verify and test the results for further enhancements. The advantage of APEX Assistant is something an experienced developer can use to quickly get started or as a programming peer one can ask from time to time.