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

{
Adding a new provider capability spans API routing, response normalization, media encoding, compatibility with existing image-generation abstractions, and integration tests. The configuration precedence fix is comparatively localized to request/model selection logic with targeted regression coverage, so the left issue carries substantially greater implementation risk and scope.
}
https://github.com/berriai/litellm/issues/28026 3:1 https://github.com/berriai/litellm/issues/21554