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

{
The right issue is harder because it requires tracing a regression across message normalization, provider-specific identifier constraints, multi-turn state preservation, and compatibility tests, with risk of affecting existing Anthropic integrations. The left issue is comparatively localized: expose provider capability metadata through a small, stable API with focused tests.
}
https://github.com/berriai/litellm/issues/32214 5:1 https://github.com/berriai/litellm/issues/26724