LLM code output can do a variety of things like print to stdout, stderr, draw graphs, render inline tables as html, have errors...
Presenting these to the user has a likely order of importance for what is most meaningful; e.g. a graph diagram will be more important than the textual representation (like how matplotlib specifies dimensions)
This PR adds groundwork for icon selection and having a human readable interface for those types.
This PR adds groundwork for icon selection and having a human readable interface for those types.