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

{
Transcription parsing spans shared response-model validation, provider-specific payload normalization, and both SDK and proxy execution paths, creating broader compatibility and regression-testing risk. The callback issue is more localized to proxy metadata representation and callback isolation, with a clearer architectural fix.
}
https://github.com/berriai/litellm/issues/33764 3:2 https://github.com/berriai/litellm/issues/30825