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

{
The right issue spans request-mode propagation across synchronous and asynchronous paths, provider-specific translation, and response normalization, creating broader regression risk. The left issue is more localized to usage extraction and cost accounting with a narrower behavioral surface.
}
https://github.com/berriai/litellm/issues/34094 3:1 https://github.com/berriai/litellm/issues/22984