Showing Reasoning Texts to user in Assistant API

API
,
1

Hi everyone,

I would like to use the o1 and o3-mini models in our internal project. While testing these models in the platform’s playground for Assistants, I noticed that in the Logs, I can see “Run” and “Run Steps.” However, I’m not able to view the “Reasoning texts” in the responses. My goal is to use these reasoning texts and demonstrate them to our users in the Chat UI, similar to how ChatGPT shows its chain-of-thought.

Below is a scenario to better illustrate my question:

Question: {USER MEETING TEXT}
“Explain the text above and summarize, provide meeting minutes with agenda and descriptions.”

Answer:
{reasoning part 1}
Extracting important topics from the user’s text…
{/reasoning part 1}

{reasoning part 2}
Creating Agenda Titles…
{/reasoning part 2}

{reasoning part 3}
Creating Descriptions of Agendas…
{/reasoning part 3}

{answer}
[ANSWER]
{/answer}

Is there a way to retrieve and display these “Reasoning texts” (or chain-of-thought) in a similar format for our internal users? If so, could you provide guidance on how to enable this feature in the platform playground or through the API?

Thank you!

2

The reasoning generation is internal and is not returned, but in fact is protected from examination.

It also may not be as simple as just some additional prompts and additional text that can be understood and presented to you.

I am hopeful we would at least get the “summary” of thinking at some point, like ChatGPT, where we have some dialog being returned to keep us amused about the progress.