Pablan, as it stands

Self-hosted knowledge management for SMEs: a split-screen Markdown editor
whose sections an LLM refines while you write, and RAG question answering
over the documents that result. FastAPI + Postgres/pgvector on the back,
SvelteKit on the front, everything OpenAI-compatible and self-hostable.

Squashed into a single commit; the development history stays local.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CA43ZJda8Rbp2hKXNy8f6b

frontend/src/lib/components/StreamingText.svelte

@@ -0,0 +1,39 @@
+<script lang="ts">
+	// Renders streamed model output token-by-token with a fade-in, the way the
+	// Claude web UI does: each arriving token materialises instead of snapping
+	// in. Every token is its own keyed span, so Svelte mounts only the newest
+	// one per update and its fade plays exactly once — re-rendering the whole
+	// Markdown string on each token (what {@html} does) would replay every
+	// element's animation and flicker.
+	//
+	// This is plain text on purpose: it is the transient streaming view. The
+	// caller swaps to the sanitizing <Markdown> renderer once the turn settles,
+	// so formatting still arrives — just at the end, without mid-stream reflow.
+	let { tokens }: { tokens: string[] } = $props();
+</script>
+
+<div class="streaming-text text-sm break-words whitespace-pre-wrap">
+	{#each tokens as token, i (i)}<span class="tok">{token}</span>{/each}
+</div>
+
+<style>
+	@keyframes pb-token-in {
+		from {
+			opacity: 0;
+			filter: blur(3px);
+		}
+		to {
+			opacity: 1;
+			filter: blur(0);
+		}
+	}
+	.tok {
+		animation: pb-token-in 0.55s ease-out both;
+	}
+	/* Motion is decorative; the settled text is the same either way. */
+	@media (prefers-reduced-motion: reduce) {
+		.tok {
+			animation: none;
+		}
+	}
+</style>