Why does Cascaide exist?

Note: This will be the only verbose part of this documentation.

AI orchestration is somewhat of a new domain. It’s the wild west out here.

Large Language Models (LLMs) burst into the scene towards the end of 2022 with the release of ChatGPT. Since then things have moved fast, so fast that we consider papers than came out in 2023 ancient. So much attention, talent and money has flowed into AI, very much like the early days of the dot com rush. Soon, we realised that robust software frameworks are required to actually put these smart models to work, birthing the first of the AI orchestration frameworks. Everyone started building AI agents. So did I.

While this is all well and good, this downpour also resulted in hype. Copious amounts of it. With hype came the idea that orchestration is supposed to be an extremely complex endeavour requiring frameworks that have massive codebases and impenetrable abstractions.

Well, after struggling around for a while, I decided to build something to make my life easier. Eventually, I realised others might find it useful too.

My claim is simple: most AI orchestration use cases don’t need that complexity. The abstractions should be light enough to grasp quickly and powerful enough to build whatever you need.

Cascaide is my shot at this goal. Do help, maybe we can build something good.