A lab/playground for experimenting with LLMs and appying different patterns such as Chain of Throught, Emotional Intelligence, Styles of memory, short term and longterm and even dreaming.
I'm still working on the structure and the content. I'm also working on the code to make it as modular as possible. I'm inspired my the process of modular synthesis and the idea of creating a system that can be patched together in different ways to create different results. I'm also inspired by the idea of developing artful and beautiful UIs and try to break out of the typical Chatbot UI's that are so common in the field of AI.
My job is to provide creative direction and to keep the code patterns decoupled, modular, and pipeline based. The big challenge of this is clearly articulating my ideas to the AI in a way that it can understand and then generate the code and then slowly refining it, testing and logging the results as I go, while keeping the context relevant for the AI so it doesn't lose track of what we're trying to do.
- Each module should only know about the things it absolutely has to, in order to do its job.
- Each module should have a generic standard interface so that it can be swapped out for another module in the pipeline.