As AI services and the data they consume and create become more important and prevalent in various applications and processes, so do the platforms and architectures they are built upon. As usual, there is no “one size fits all.” However, what is briefly presented here is an optimized approach to such data-driven AI application architectures. All of the source code mentioned and more can be found here and a free “Develop with Oracle AI and Database Services: Gen, Vision, Speech, Language, and OML” workshop (where all of the use cases are based on the U.N.’s 17 Sustainable Development Goals) giving many more examples can be found here.
Often multiple network calls must be made in a given AI app, entailing calls to the AI services as well as calls to retrieve and persist the content (whether it be text, audio, images, video, etc.) that is the input or output. The persistent information is then often processed and analyzed further, and additional calls, AI or otherwise, are made in reaction.