Unexpected Usage of OpenRouter/OpenAI Models Despite Selecting Free Options

[eramax]

What happened?

While utilizing free models via OpenRouter, I observed that the application is invoking OpenAI models without explicit user consent or transparency regarding the selection mechanism. There is no apparent configuration or setting available to control or disable this behavior.

Version Information

the demo version on the website

Steps to Reproduce

check description

What browsers are you seeing the problem on?

Firefox

Relevant log output

Screenshots

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

It’s the generations for titles, which by default use a specific system defined model by provider if not configured (gpt-4o-mini for OpenAI, for example).

you can customize this for each endpoint, and even using “current_model” as the value of you want the system to use the current chat model:

https://www.librechat.ai/docs/configuration/librechat_yaml/object_structure/shared_endpoint_settings#titlemodel

[eramax]

I understand but It should be configurable on your demo app, so I don't have to self host to set the title model.

[fdterr]

I am also seeing this. I am getting unexplained usage of o1-pro coming from LibreChat per OpenRouter. I cannot find any chat logs corresponding to that and the usage is way too high to be from title generation

[danny-avila]

Are you hosting LibreChat yourself? There is not enough information to help you determine where this usage is coming from.

If you are hosting yourself, you can review the transaction collection for any relevant documents and determine the context (message, conversation, user ID's):

If reviewing debug logs, as they are separated by date, you would have to search through all possible dates, and determine the context.

[danny-avila]

In the absolute worst case, someone has unauthorized access to your instance, but even so, it would show up in your logs.

See this previous comment for more info:
#6967 (comment)