[Feature Request]: Provide access to the content of the reasoning when using reasoning LLMs, Ollama and structured output

[MHaurel]

Feature Description

When requesting a structured output from a reasoning LLM using the code below, only the json is returned by the LLM. We would like to have access to the reasoning of the LLM (generally encapsuled in between tags).

llm = Ollama(model="deepseek-r1:1.5b")

structured_llm = llm.as_structured_llm(Invoice)
response = structured_llm.complete(input_text)

Thanks for your work !

Reason

No response

Value of Feature

No response

[logan-markewich]

Generally speaking there is no reasoning -- Ollama, OpenAI, Gemin, etc. hide structured outputs as tool calls within their API

LLM classes that do not support function calling use json prompting, but tbh exposing this will kind of complicate this API design or even break usage for other users 🤔 I'll think a bit more about this