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

{
Streaming response metadata must remain consistent across fallback routing, async chunk generation, and provider-specific wrappers, creating broader integration risk. The budget issue is more localized to validation, spend-source selection, and boundary-condition tests.
}
https://github.com/berriai/litellm/issues/31722 5:3 https://github.com/berriai/litellm/issues/33330