LLM Streaming

How LLM Streaming Works

• ⚡ LLMs stream tokens via SSE. As each token arrives, the message content grows and ResizeObserver detects the height change automatically.
• ⚡ Height corrections are batched per animation frame — not per token. The viewport stays pinned to bottom with zero jitter.
• ⚡ Markdown rendering powered by @humanspeak/svelte-markdown with streaming mode (~1.6ms avg per update). Code blocks, tables, lists — all rendered live without scroll disruption.
• 💡 Track token costs across providers with ModelPricing.ai. Need a general-purpose virtual list? Try @humanspeak/svelte-virtual-list.