Hi,

I’d like to ask whether WebRTC-based models (e.g., Azure OpenAI gpt-realtime) are currently supported for cost tracking in LiteLLM.

I tested LiteLLM (v1.82.3) with Azure gpt-realtime using WebRTC, and the inference itself works correctly (audio streaming, responses, etc.). However, I noticed that:

No cost-related logs are generated The backend logs show the following error:

_Cost tracking failed for model=gpt-realtime_