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

{
The right-hand task requires tracing provider-response normalization, preserving nested usage metadata across typed models and serialization layers, and adding regression coverage. The left-hand item appears primarily diagnostic or configuration-related, with little evidence of a required code change.
}
https://github.com/berriai/litellm/issues/33967 3:1 https://github.com/berriai/litellm/issues/32923