The ai-framework provides a lot of useful functions and abstractions to manage the ai providers, give context to the llm and manage functions.
We would like to explore how could this be exposed as a service to xblocks so xblock developers don't need to cross xblock boundaries to import orchestrators or processors.