Tab Radar – AI Tab Manager & Suspender 作者： Jonas Kopp
Ultra-light tab manager: AI grouping by project/topic, reliable tab suspender, 100% on-device.
1 个用户1 个用户
扩展元数据
屏幕截图
关于此扩展
When your browser is buckling under a pile of open tabs and eating up RAM, Tab Radar brings order back. It groups tabs by topic, suspends the ones you are not using to free memory, and preserves what you have typed — all on your own machine.
Smart topic grouping — One click sorts your tabs into meaningful groups: curated categories (News, Social Media, Video…) with no model required, TF-IDF keyword clustering for the rest, automatic group naming, and new tabs slotted into the right existing group.
Tab suspending to reclaim RAM — Inactive tabs are discarded so the browser reclaims memory, then reload on return. Custom inactivity threshold, real-time tabs (WhatsApp, Discord, Slack, Teams, ChatGPT, Zoom, Figma, Google Docs…) protected by default, one-click restore.
Page-state and form preservation (opt-in) — Form inputs and scroll position captured before suspension and restored on reload, SPA fallback included. Password/hidden/file fields never captured.
Privacy and offline AI — No account, no tracking, no data collection. The optional model (all-MiniLM-L6-v2) and runtime are bundled — no runtime downloads, no CDN, no remote code.
Lightweight by design — Embeddings off by default; when on, computed only on demand, cached, RAM-gated (~4 GB+), single-threaded, released when idle.
Languages — English and German, switchable in settings.
Made with ♥ in Hamburg, Germany, by Jonas Kopp. buy me a beer 🍺
Smart topic grouping — One click sorts your tabs into meaningful groups: curated categories (News, Social Media, Video…) with no model required, TF-IDF keyword clustering for the rest, automatic group naming, and new tabs slotted into the right existing group.
Tab suspending to reclaim RAM — Inactive tabs are discarded so the browser reclaims memory, then reload on return. Custom inactivity threshold, real-time tabs (WhatsApp, Discord, Slack, Teams, ChatGPT, Zoom, Figma, Google Docs…) protected by default, one-click restore.
Page-state and form preservation (opt-in) — Form inputs and scroll position captured before suspension and restored on reload, SPA fallback included. Password/hidden/file fields never captured.
Privacy and offline AI — No account, no tracking, no data collection. The optional model (all-MiniLM-L6-v2) and runtime are bundled — no runtime downloads, no CDN, no remote code.
Lightweight by design — Embeddings off by default; when on, computed only on demand, cached, RAM-gated (~4 GB+), single-threaded, released when idle.
Languages — English and German, switchable in settings.
Made with ♥ in Hamburg, Germany, by Jonas Kopp. buy me a beer 🍺
评分 0（1 位用户）
权限与数据
更多信息