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

{
The launch-agent feature is substantially harder: it requires a new CLI workflow, remote authentication and model discovery, subprocess/environment integration, and compatibility across multiple external agent tools and platforms. The cancellation bug is narrower, though it still needs careful async exception, fallback, cooldown, and regression-test handling.
}
https://github.com/berriai/litellm/issues/30421 5:1 https://github.com/berriai/litellm/issues/35329