We could could also output some additional info the user can use for prompting, example:
When rendering endpoints, let's dump useful LLM information into a text file that may be used for prompting (manual).
Ideally for each rendered endpoints we have a path fragment of openAPI spec that contains the following
only GET operation
parameters but dereferenced
actual response that we picked
in the response the schema should be de-referenced (but maybe only to the 2 nesting level)
examples should be kept
let's render using yaml. this is what LLM will understand
if we detected parent endpoint let's give path to it.
We could could also output some additional info the user can use for prompting, example:
When rendering endpoints, let's dump useful LLM information into a text file that may be used for prompting (manual). Ideally for each rendered endpoints we have a path fragment of openAPI spec that contains the following
let's render using yaml. this is what LLM will understand
if we detected parent endpoint let's give path to it.