I plan to register the package in the first quarter of 2026.
The low-level base models, such as arima(), auto_arima(), and ets() are now stable from an API perspective and will not undergo any breaking changes. I’ll continue to refactor and optimize their internals, but the external interfaces will remain consistent.
At the moment, I’m introducing a formula interface grammar and building parallelization capabilities to support large-scale forecasting problems efficiently.