Sorting LiteLLM GitHub issues by difficulty to implement. Higher = harder.

{
The left issue spans pass-through response handling, guardrail enforcement semantics, raw-to-normalized response translation, and likely multiple response formats, creating broader regression and testing risk. The right issue is comparatively localized to preserving the configured alias through the streaming callback path.
}
https://github.com/berriai/litellm/issues/32201 4:1 https://github.com/berriai/litellm/issues/25628