mirror of
https://github.com/excalidraw/excalidraw.git
synced 2026-11-21 21:44:59 +03:00
* perf(editor): bump @excalidraw/mermaid-to-excalidraw to 3.1.0 3.1.0 loads Pele's code per diagram type, on first use (`pele/lazy`), instead of all 33 diagram types at once: a first flowchart downloads ~66 KiB (brotli) instead of ~217 KiB. The per-diagram chunks are only reachable through the lib's dynamic import, so `lazyChunksPlugin` keeps them out of the PWA precache. Claude Opus 5.5 (xhigh) * chore(editor): name Pele's diagram chunks in TTD chunk-load comments Since mermaid-to-excalidraw v3, it's Pele (not Mermaid) that lazy-loads a chunk per diagram type, as of 3.1.0. Claude Opus 5.5 (xhigh) * feat(app): cache all Mermaid chunks for offline use once the TTD dialog opens mermaid-to-excalidraw 3.1 splits Pele into a chunk per diagram type, which aren't precached. With 3.0, the single Pele chunk cached on first use covered every diagram type offline; now only the types converted while online would. `lazyChunksPlugin` emits the list of mermaid chunks as a hashed JSON asset, which index.html points to with a `<meta>` (+33 B br). Once the TTD dialog is opened, the app fetches the list and the chunks (after 5 s, so the dialog's own chunks aren't fetched twice, at low priority) for the service worker to cache in `lazy-chunks` (~200 kB br on top of what the dialog loads). App only: package hosts (e.g. E+) are unaffected. Claude Opus 5.5 (xhigh)