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

{
The right issue is harder because it requires a cross-cutting, backward-compatible exception API change, reliable normalization across provider-specific error formats, and validation through direct and wrapped error paths. The left issue is comparatively localized to model-specific parameter capability detection and regression tests.
}
https://github.com/berriai/litellm/issues/26070 3:1 https://github.com/berriai/litellm/issues/26444