Skip to content

v0.6.0

Choose a tag to compare

View all tags
@alterago alterago released this 26 May 20:55
v0.6.0
36357e7

AI Craftery Bord v0.6.0

Two features land together: Bord now talks to Google's Antigravity CLI as a first-class provider, and the whole UI runs through proper internationalization with ten translated languages.

Antigravity provider

Open a chat in Bord, pick Antigravity from the provider list, and you get the same chat panel that Claude / Codex / Gemini / Echo show — input box, history, model and permission dropdowns. The five models agy 1.0.2 actually accepts are exposed:

  • Gemini 3.5 Flash — Low / Medium / High effort
  • Gemini 3.1 Pro — High / Low effort
  • Claude Sonnet 4.6 — Thinking
  • Claude Opus 4.6 — Thinking
  • GPT-OSS-120b — Medium

The effort dropdown updates the moment you change the model. Switch model between turns without restarting the conversation — Bord re-targets agy per turn, the next reply comes from the new model.

Permission modes match Claude / Codex / Gemini in shape: Ask (default), Sandbox, YOLO. Right-click on a session in the sidebar for Rename / Move (change working directory) / Delete. Multi-select + delete works for batch cleanup.

No additional API keys, no Python install, no Node packages. Bord uses agy's existing OAuth. One-time setup in the terminal: brew install --cask antigravity-cli and agy auth login.

Ten translated languages

Settings → Language has eleven entries: ten languages plus Auto.

  • English, Русский, Українська, Беларуская, ქართული, Deutsch, Español, Français, Türkçe, Italiano

Every name appears in its own script. Switching is instant — no app restart. Tab titles, settings labels, input placeholders, tooltips, chat headers, the task launcher all repaint live. Open panels are not lost, conversation context is not lost.

What gets translated and what stays raw follows a simple rule: text the UI shows to you is translated, values that flow into a model or CLI as a flag stay verbatim. The model dropdown shows claude-sonnet-4-5 exactly as the CLI expects it; the dropdown label "Model" translates.

Quality of life

  • Session history rendering is correct on resumed antigravity conversations — every turn shows its own reply once, no more concatenated wall of prior assistant text.
  • Orphan brain directories that agy leaves behind after failed initialization get swept on first launch of this version. Shaves disk space silently.
  • Sandbox / YOLO permission choices stick for antigravity assistants in Settings (they previously reverted to Ask on save).
Assets 16
Loading