Collected sources and patterns will appear here. Add from search or the patterns library.
RawChatResponse -> NormalizedChatResponse
Normalize non-standard reasoning fields and thinking tokens across disparate LLM providers to preserve conversation context in multi-turn tool loops.
Problem it solves
Different backend servers (like vLLM, OpenAI Responses API) format thinking phases and reasoning blocks inconsistently, breaking standard agent tool-calling loops.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.