Nice to see others who think that Raku is a good fit for LLM ... I have had some success integrating LLM::DWIM (a raku command line LLM client built on LLM::Functions etc) with a DSL approach to make a command line calculator based on Raku Grammars.

  > crag
  > ?^<elephant mass in kg> / ?^<mouse mass in kg>    #300000①
  > ?^<speed of a flying swallow in mph>              #30mph

PS. Raku has Inline::Python where you need a lib from the Python ecosystem (which I am sure you know, but in case others are curious)