We improved the ShuttleAI API for better speed and stability, removed unused models, introduced a smarter shuttleai/auto router, refreshed the UI, and expanded context support to 128K tokens.