Ampersand

Ampersand 0.4.0

Clearer local AI startup, a simpler inference runtime, and a consistent Editor width.

Improved

  • Clear local AI startup — Ampersand now shows when it is downloading, verifying, or loading a model, including download progress. Startup errors include a clear message and a retry action.
  • Simpler local AI runtime — Ampersand now talks directly to its bundled llama.cpp server. Removing the extra service cuts a process and keeps model setup, writing context, and AI state in the Mac app.
  • Safer model setup — Model downloads can resume after an interruption, and Ampersand verifies each download before adding it to the local cache.

Fixed

  • Consistent Editor width — Returning from Markdown view always restores the Editor’s centered writing width, even when the window finishes resizing after the view switch.
  • Clear startup conflicts — If another local AI process is using Ampersand’s port, the app reports the conflict instead of appearing to load forever.