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

{
The token-counting issue is harder because it crosses multimodal content normalization and tokenizer behavior, requiring careful handling of media payloads across models and regression coverage. The dashboard issue is narrower, likely involving stale relationship state and refresh/invalidation logic in one UI workflow.
}
https://github.com/berriai/litellm/issues/28071 3:1 https://github.com/berriai/litellm/issues/26785