Issue with Showing Thoughts [Thinking] for OpenAI and Gemini Requests via LiteLLM Custom Endpoints in LibreChat

[kishore-rajendran-zoomrx]

Description:

I am using LibreChat v0.7.9-rc1 (commit f4d97e1) with LiteLLM v1.74.* and custom endpoints to send OpenAI and Gemini requests. I expect to see the thoughts (or thinking) tokens for models that support them, but while the web search functionality via the custom endpoints works fine, I'm unable to get the thoughts to display for these models.

Steps to Reproduce:

1. Setup: Use LibreChat v0.7.9-rc1 and LiteLLM v1.74.* with custom endpoints configured for OpenAI and Gemini.
2. Input Example: Send a request that would trigger the model’s reasoning process (i.e., a request where the model’s "thinking" or "thoughts" should be exposed).
3. Expected Behavior: The model’s thought process or thinking should appear as part of the response, similar to other tokens that the model sends.
4. Actual Behavior: The thoughts are not showing up, though web search functionality works as expected.

Environment:

• LibreChat Version: v0.7.9-rc1 (commit f4d97e1)
• LiteLLM Version: v1.74.*
• Model Versions: OpenAI and Gemini via LiteLLM
• Custom Endpoints Setup: [Provide additional details on the custom endpoints setup if possible]

Additional Information:

• I’ve confirmed that web search is working as intended via the custom endpoints, so the issue appears to be specifically related to the thoughts or thinking tokens.
• Is there an additional configuration required to enable thought display for these models? Are there any specific settings or flags that need to be enabled?

Request for Help:

If anyone has encountered this issue or has insights on how to get the model’s thinking/logic to be visible through LiteLLM in LibreChat, I would greatly appreciate any guidance or suggestions.