v1.11 — CloudConvert upload pipeline + 5-second cold starts

Upload pipeline rebuilt around server-side rendering. Browser RAM usage dropped from ~2 GB to <300 MB on 500-page batches.

Upload pipeline rebuilt. Server now renders one ~3600px JPEG per page via CloudConvert. The browser does zero PDF rendering on the happy path. Cold starts down to ~5 seconds via IndexedDB hydration of recent sheets. Peak RAM stays under 300 MB even on 500-page plan sets.

Previously, the browser rendered PDF pages locally to produce preview and high-res images. This worked but consumed 2+ GB of RAM on large plan sets and could crash on resource-constrained machines.

The new pipeline ships the PDF to a CloudConvert job, which renders one high-quality JPEG per page server-side. The browser receives the rendered images via realtime updates and displays them immediately. Peak browser RAM is now under 300 MB regardless of plan set size.

The browser-side renderer remains as a fallback for the rare case where CloudConvert kickoff fails.

Machine-readable site summary for AI crawlers: /llms.txt