Having a consistent, well-defined, coherent, human readable format for applications to interact with LLM's is quite critical. Having a way to specifiy data sources would be useful - RAG sources, S3 raw data files?. What about an evals section?
This spec is an attempt to will a sensible, infrastructure-independent, declarative abstraction for LLM application patterns into existence
I'd love critical feedback and for more folks to get involved
For me, one of the most important parts of the spec is formulation of the DAG of tasks. Whether that be calling other LLMs or some retrieval mechanism.
What does everyone think about the pros/cons of a formal DAG specification vs using natural language? E.g. defining the DAG in yaml vs. something more natural language like writing logic and making calls in the prompt text?