Producing a schema from a (derived) codec seems quite useful to ensure consistency across non-Scala stack. It can also be used to Prompt LLMs for consistent response format.
The schema could be output either as String or as a JSON AST.
https://json-schema.org/draft/2020-12/json-schema-core