Back
GoogleAntigravity
Model API retry resilience, headless print optimizations, and detailed tool approval prompts
AI summary
Written by AI from the official notes. Check them for exact details.The update enhances error resilience, speeds up sign-in, and improves headless execution and tool approval prompts.
- Improved resilience to transient model API errors with extended exponential backoff.
- Accelerated sign-in and startup by using cached credentials.
- Streamlined headless execution to exit promptly after delivering answers.
- Enhanced tool approval prompts with specific action requests and reasons.
- Fixed multiple issues related to plugin reinstalls and memory leaks.
Why it matters: Users relying on headless execution and tool integrations should care for improved performance and reliability.
Full release notes16 changes
Improves resilience to transient model API errors with extended exponential backoff, accelerates startup and sign-in by reading cached credentials, streamlines headless execution and error reporting, enhances tool approval prompts with contextual action descriptions and reasons, and fixes issues across plugin reinstalls, plugin MCP working directories, subagent lifecycle states, and command memory leaks.
- Improved resilience to transient model API errors by retrying errors such as
503 Unavailablewith extended exponential backoff to prevent brief hiccups from aborting sessions. - Improved sign-in and startup speed by reading signed-in identity details from stored credentials instead of making network requests on every launch.
- Improved headless (
-p) runs to exit promptly upon delivering final answers, waiting for non-daemon background tasks and scheduled timers within--print-timeoutwhile preserving running background daemons. - Improved failure reporting in headless (
-p) runs by printing fatal errors to stderr with stableerror:markers and adding truncation notifications. - Improved headless (
-p) latency by eliminating up to 200 ms of idle delay per turn and skipping unused conversation title generation model calls. - Improved tool approval prompts to display specific action requests (e.g.
Run this command?,Allow access to this URL?, orAllow calling this tool?), restricting command editing to commands and including aReason:explanation when triggered by hooks or cross-project files. - Improved model selection auditability by logging alias resolutions,
--effortvariant mappings, and deprecated model replacements incli.log. - Changed
--print-timeoutexpiration behavior in headless mode to return partial output and exit successfully with a warning on stderr instead of raising a timeout failure. - Changed default URL fetching permissions to prompt for approval by default before reading external URLs unless access has been pre-granted.
- Fixed plugin reinstalls leaving behind files deleted from the source by replacing managed plugin directories cleanly, preventing self-directory installation corruption, and cleaning up disabled state entries from
config.jsonupon uninstall. - Fixed plugin-defined MCP servers resolving relative working directories incorrectly by scoping them directly to the plugin's root directory.
- Fixed a startup race condition where subagents could fail to discover tools from MCP servers that were still initializing.
- Fixed subagents occasionally appearing stuck in running states in large conversations, resolving delays delivering queued idle messages.
- Fixed a memory leak where finished terminal command executions retained internal state throughout long-running sessions.
- Fixed headless (
-p) runs hanging indefinitely on interactive implementation-plan approval prompts by proceeding through plan review automatically in non-interactive mode. - Fixed sign-in failing with project access errors after switching from a business account with a selected project to a personal account by resetting cached project and license data.