Auto Mode Routing

The mode option defaults to "auto": the library picks the optimal path based on total row count and the runtime environment, so business code never needs to react to data size.

Routing rules

Browser

Rows	Path	Notes
< 20,000	main	Synchronous main-thread build; ~120ms at 10k×6 columns
20,000 – 49,999	worker + Workbook	Main thread only does one structured clone (~94ms at 100k rows); WASM runs in the Worker
>= 50,000	worker + Fast stream	Custom fflate writes avoid the toBuffer cliff

Node / SSR (no Web Worker)

Rows	Path	Notes
< 50,000	main	Main-thread Workbook build, full styling
>= 50,000	stream	Fast stream, ~0.8s at 100k rows

Explicit modes

await exportExcel({ ..., mode: "stream" });  // force streaming
await exportExcel({ ..., mode: "worker" });  // force Worker (browser)
await exportExcel({ ..., mode: "main" });    // force main thread

mode: "worker" does not error in Node/SSR: it falls back to the main-thread path (stream above 50k rows), preserving style semantics instead of silently degrading to the style-less stream.

Why 20,000 / 50,000

• 20,000 rows: synchronous main-thread work is acceptable below this level; avoids unnecessary Worker startup;
• 50,000 rows: Workbook.toBuffer() shows a superlinear cliff beyond ~55k rows (~17.5s at 100k), while Fast stream stays at ~0.8s. STREAM_THRESHOLD = 50_000 keeps a safety margin.

Trade-off: the stream path supports multi-row headers and cell merges, but not cell styles, header styles or layout features (width/freeze/filter). For fully-styled exports stay under 50k rows or split into multiple sheets.