HomeDemo
Live demo
Streaming Markdown Demo
Stream the Naive Bayes article into the renderer to inspect incremental layout behavior, scheduler sensitivity, and patch convergence on the exact surface used by the docs and benchmark pages.
Interactive controlsScheduler-aware renderingAutomation hooks for regression
What to look for
Focus on visible-first latency, list/table/code stability during streaming, and whether the final DOM settles without post-finalize structural drift.
The same demo surface feeds the reliability harness. If this page looks wrong under real streaming conditions, that is a product defect rather than a cosmetic edge case.
Streaming lab
Same surface used by the reliability harness
Use it for
Visual-first latency and convergence checks
Operator focus
Control rail, metrics strip, and final DOM stability
Markdown Example
Streaming0 / 22,728 chars streamed
First flush
—
Latency p95
—
Patch p95
—
Long tasks
—
Shiki ms
—
Coalescing
—
Metrics
Run the stream to populate batch metrics.
Coalescing
—
input 0 → output 0 (0 merged)
duration total 0.00 ms
append lines 0
Normal budgetduration p95 — (0 samples)
Reduction—
No samples yet
Accumulator duration—
No samples yet
Recent batchestarget reduction ≥ 5% · duration ≤ 5ms
Run the stream to populate batch metrics.