runpod-workers / worker-vllm

The RunPod worker template for serving our large language model endpoints. Powered by vLLM.
MIT License
213 stars 82 forks source link

Best way to record data #64

Closed aodhan-domhnaill closed 3 months ago

aodhan-domhnaill commented 4 months ago

I am looking to record input and output to the vLLM. I could put an HTTP proxy in front and capture the traffic, or modify your handler.

Rather than make changes to the code, I was wondering if you might have a better way to do this?

alpayariyak commented 3 months ago

We currently do not have a way to capture token usage across all workers, but it is a future item we will implement