Antigravity 1.0.8: Smarter UI layout and slash command history support

Release notes

FEATURE

Smarter UI layout and slash command history support

The terminal interface is now much more practical and better looking. You can finally press the "Up" arrow key to recall previously entered slash commands (like /plan), saving you from retyping them every time. When typing in the autocomplete menu, it will no longer mistakenly "help" you by forcefully auto-completing /conv into /conv-switch and then abruptly closing the menu. The completely redesigned "Models & Quota" page is now live, and the status bar at the very bottom will always keep you informed of your current quota usage and execution mode. When viewing your conversation history or task lists, if your terminal window is too narrow, the system will automatically drop less important columns to prevent layout chaos, while ensuring times and statuses align perfectly. Like your favorite cafe getting a remodel: the register finally remembers your secret menu orders, and even if you sit at a tiny table, the waiter still manages to arrange your plates perfectly.

BUGFIX

Fixes performance on large text pastes, legacy CPUs, and clipboard issues

A slew of performance issues that could freeze or crash the program have been resolved. Previously, if you pasted a ridiculously long single line of text (like tens of thousands of characters), the input box would choke and freeze for ages; now, it smartly collapses it into a little placeholder icon, keeping performance smooth. When you use /btw to ask the agent for a quick favor, its replies are now faster and stream in word by word instead of getting prematurely truncated. If you're on a very old computer (with a CPU that lacks hardware acceleration), the program used to instantly crash on startup, but now it opens perfectly. In addition, clipboard reading on Windows and certain Linux setups (Wayland) has been fixed, and the system now checks the file size before reading the clipboard so pasting massive files won't devour all your RAM. Like installing a much stronger water pump: previously, dumping too much water at once would clog or burst the pipes, but now it automatically bundles and reroutes the flow so even the old plumbing in an aging house can handle it, while fixing a few leaks along the way.

FEATURE

Hooks save to the correct path, and denying permissions no longer causes hangs

Settings and state synchronization within the system are now much more precise. Previously, when you used /hooks to attach automated scripts, the system would foolishly write the config file into the wrong folder, meaning your backend wouldn't read anything despite all your setup work; it now correctly writes to the shared configuration directory. When you switch projects or add a new directory, any custom slash commands or skills you wrote will instantly appear in the autocomplete menu without you needing to restart the program. If saving a configuration fails, the status bar will now immediately flash a red warning alert. Additionally, if you denied an agent's permission request and then immediately sent it a new message, the backend used to hang because it hadn't finished cleaning up after the denial; this "slow reaction" bug is now cured. Like visiting a government office: the application forms you filled out at the counter used to be filed in the wrong cabinet, and if you just joined a new department, it took a whole day before anyone could find you in the directory; now everything is instantly synced, and if you turn away a salesperson, they no longer block the doorway preventing you from talking to the next person.

Original changelog

Original changelogSmarter UI layout and slash command history support

Added support for capturing slash command history, allowing users to use the up arrow to replay previously entered slash commands.
Redesigned the "Models & Quota" page (enabled by default, replacing the legacy usage page) to gracefully handle disabled quota buckets by displaying a dimmed "Disabled" status and omitting the progress bar.
Added display of quota usage and execution mode in the status line.
Redesigned the `/resume` conversation picker to align the workspace column and added adaptive column dropping (workspace, time, steps) to support narrow terminals.
Redesigned the `/tasks` list and detail views for better alignment and readability, placing start times on the left, right-aligning status, and capping the panel height.
Fixed an autocomplete bug where a command that is an exact prefix of another (e.g., `/conv` vs `/conv-switch`) would aggressively auto-complete and hide the suggestions menu.

Original changelogFixes performance on large text pastes, legacy CPUs, and clipboard issues

Improved `/btw` to be more token efficient and support streaming responses for a smoother user experience and fixed premature truncation.
Fixed a CPU compatibility issue (SIGILL on non-AES-NI CPUs), preventing immediate crashes on startup on older CPUs (like Intel Ivy Bridge) or VM environments that lack AES-NI support.
Added a per-line guard against extremely long single-line pastes in the TUI prompt editor to prevent performance lag, replacing them with an expandable placeholder.
Fixed a TUI hang in the artifact view during long sessions by optimizing the rendering complexity of large step histories.
Fixed potential OOM risks when reading large clipboard files by verifying file size before reading.
Fixed Windows and Wayland-only Linux distributions clipboard image and file reading.

Original changelogHooks save to the correct path, and denying permissions no longer causes hangs

Fixed a bug where the `/hooks` command wrote configurations to `~/.gemini/antigravity-cli/hooks.json` instead of the shared `~/.gemini/config/hooks.json`, ensuring hooks remain synchronized between the TUI and the backend.
Fixed dynamic reloading of custom skills and system slash commands, ensuring they are instantly discovered in autocomplete upon conversation switch or `/add-dir`.
Improved configuration saving by propagating write failures as transient error flashes on the statusline.
Improved settings inheritance by ensuring the CLI inherits the `use_ai_credits` setting from global user settings on startup.
Fixed a race condition where sending a message immediately after denying a permission request would fail due to incomplete backend cleanup.