Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
RawResponse -> StructuredResponse
Extract and separate model reasoning text wrapped in XML-style tags (such as `<think>`) from the final generated output.
Problem it solves
Reasoning traces in thinking models contaminate the final answer string, breaking downstream exact-match and parsing evaluations.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.