GithubStarsManager v0.5.8
What's Changed
New Features
- Translation skip support: Added
data-translateattribute support to skip specific elements during AI translation. (PR #134 by @SummerRay160)
Bug Fixes
- SQL placeholder crash: Fixed releases sync INSERT statement missing one placeholder (
14 values for 15 columns), causing sync failures. (PR #140 by @iamvicliu) - Settings serialization: Fixed
better-sqlite3RangeError when saving AI/WebDAV config objects — complex values are now properly JSON-serialized. (PR #140 by @iamvicliu) - Gemini thinking model support: Raised
maxTokensto 2048 for Gemini connection tests, extended timeout to 30s for thinking models, and filtered outthought: trueparts from responses. (PR #140 by @iamvicliu) - Sync data loss on refresh: Fixed starred repos disappearing if the page was refreshed within 2 seconds of sync — data is now pushed to backend immediately after fetching. (PR #142 by @iamvicliu)
- Empty backend overwrites local cache: Guard against backend returning empty array silently wiping locally-cached repos — local data is now preserved when backend has nothing. (PR #142 by @iamvicliu)
- MiMo thinking chain: Disabled thinking chain (
thinking: { type: 'disabled' }) for Xiaomi MiMo models to prevent reasoning content from leaking into analysis results. Fixes #127. (PR #143) - Docker sync ERR_CONNECTION_RESET: Added automatic retry with exponential backoff (1s/2s/4s, up to 3 attempts) for transient network errors, increased frontend sync timeout from 30s to 120s, and bumped nginx proxy timeouts to 300s with
proxy_buffering off. Fixes #133. (PR #144)
Acknowledgements
Special thanks to external contributors:
- @SummerRay160 (PR #134) — for the translation skip feature
- @iamvicliu (PR #140, #142) — for the SQL fix, settings serialization, Gemini thinking model support, and sync data-loss guard
Full Changelog: v0.5.7...v0.5.8