Maximum context token settings not observed

[dvejsada]

What happened?

The agent model settings allows to set maximum context token for each agent. With our main agent, we have currently set maximum context token for 200000, but LibreChat happily sends requests with e.g. 500000 input tokens (as we can see in request traces) to OpenRouter instead of truncating or providing an error.

Version Information

ghcr.io/danny-avila/librechat-dev latest 33a740ed2dce 5 days ago 1.95GB

Steps to Reproduce

1. Create Agent
2. Set maximum context tokens in agent model settings
3. Create request exceeding this limit
4. See no error

What browsers are you seeing the problem on?

No response

Relevant log output

N/A

Screenshots

No response

Code of Conduct

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

[danny-avila]

I'm seeing max context tokens loading in from an agent configuration as expected.

Logs from LibreChat side will help troubleshoot the discrepancy as well as what your current agent configuration looks like in full, plus the "request exceeding this limit".

On a side note, token accounting will be improving soon to give better estimates, as there is a variance on the main branch that we've since improved. This will come when summarization is merged.