Anthropic: You're killing yourselves by not supporting structured responses. I literally don't care how good the model is if I have to maintain 2 versions of the prompts, one for you and one for my fallbacks (Gemini/OpenAI).
Get on and support proper pydantic schemas/JSON objects instead of XML.
Anthropic: You're killing yourselves by not supporting structured responses. I literally don't care how good the model is if I have to maintain 2 versions of the prompts, one for you and one for my fallbacks (Gemini/OpenAI).
Get on and support proper pydantic schemas/JSON objects instead of XML.