T3 Code Partitions
T3 Code stores per-partition preview caches under its Partitions directory so each isolated workspace or embedded webview can keep its own Chromium-style temporary data. These partitions accumulate transient browser artifacts such as HTTP cache entries, media cache files, GPU shader and GPUCache data, code cache bytecode, and other disposable preview resources created to speed up loading rendered docs, local web apps, and extension-driven previews. Kudu removes only those preview-partition cache files on Windows, leaving partition sessions, site storage, workspace state, accounts, settings, and saved project data untouched.
Why clean T3 Code Partitions?
- Stale HTTP cache inside a partition can keep serving an old preview build, so you keep seeing outdated CSS, JavaScript, or generated docs even after the source files changed.
- Invalidated code cache bytecode after a T3 Code or embedded runtime update can make the first preview launch slow or cause a partitioned webview to rebuild scripts before it becomes responsive.
- GPUCache and shader blobs from an older graphics driver can stop matching the current driver, leading to black preview panes, flickering canvases, or stutter the first time a WebGL-heavy preview opens.
- Corrupted media cache entries can break embedded audio or video previews, showing endless loading spinners, missing thumbnails, or playback that fails until the cache is recreated.
- Overgrown partition caches consume disk space across many workspaces, especially when each preview keeps its own copies of bundled assets, so AppData grows unexpectedly even though projects themselves have not.
- Partition-specific temporary webview files can become inconsistent after crashes, causing one preview tab to misbehave while others work normally; clearing that partition cache often restores normal rendering without touching its stored session or site data.
Cache paths Kudu targets
Windows
%AppData%/t3code/Partitions |
Common questions about T3 Code Partitions
Download Kudu and reclaim your disk space.
Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.