v2.1.1

Fixed

Startup latency — models, history, config, init now start in ~0.1s instead of ~3s. litellm was being imported at startup even for commands that don't need it. Now lazy-loaded inside ask only.

Added

Loading spinner — ask shows a dim grey spinner with a random thematic message while the engine loads, so there's immediate feedback before the debate begins.

Upgrade

uv tool upgrade dissenter