Swap OpenAI Code Interpreter for E2B/OpenSandbox
The OpenAI Responses and Chat Completions APIs let you declare a code_interpreter tool and the model runs Python inside an OpenAI-hosted container. That container is opaque, billed by OpenAI, and the code (often customer data) leaves your perimeter. LiteLLM now let's you intercept that tool call and runs it in a sandbox you control. The client request is unchanged.
Available starting LiteLLM v1.91.0.dev1. Check here for releases.