Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

feat(model-lib): derive a readable card title from a spec-shaped prompt

nirholas7a26dff13f7a
diff --git a/src/model-lib.js b/src/model-lib.js
index 3b64829d0..a1a946f3b 100644
--- a/src/model-lib.js
+++ b/src/model-lib.js
@@ -21,6 +21,31 @@ export function titleFromPrompt(prompt) {
return t.charAt(0).toUpperCase() + t.slice(1);
}
+// A card label, not a heading. Prompts that came out of a refiner or an
+// image-to-3D pass are structured specs hundreds of characters long ("1.
+// Geometry and pose: Torso: volumetric, teardrop shaped, …"), so a grid of them
+// reads as a wall of identical text with the one distinguishing word off the
+// end. Keep the first clause of the first line: enough to tell two models
+// apart, short enough to sit on two lines of a card. The full prompt still
+// belongs in the tooltip/aria-label of whatever renders this.
+export function cardTitleFromPrompt(prompt, maxLen = 48) {
+ const line = String(prompt || '').trim().split('\n')[0].trim();
+ if (!line) return 'Untitled model';
+ // Drop list scaffolding ("1.", "2)", "-", "•") the spec formats start with.
+ const unnumbered = line.replace(/^\s*(?:[-*•]|\d+\s*[.)])\s+/, '').trim();
+ // First clause: a colon/semicolon, the "1." that opens the spec's first
+ // numbered section mid-line, or a comma/period followed by a space.
+ // Requiring the space keeps "3.5 inch" and "1,200" whole.
+ const clause = unnumbered.split(/[:;]|\s\d+\s*[.)]\s|,\s|\.\s|\.$/)[0].trim() || unnumbered;
+ const head = clause.length >= 3 ? clause : unnumbered;
+ const trimmed = head.replace(/[\s,;:.-]+$/, '');
+ const t =
+ trimmed.length <= maxLen
+ ? trimmed
+ : `${trimmed.slice(0, maxLen).replace(/\s+\S*$/, '')}…`;
+ return t ? t.charAt(0).toUpperCase() + t.slice(1) : 'Untitled model';
+}
+
// 1234 → "1.2k", 1200000 → "1.2M"; below 1000 verbatim.
export function formatCount(n) {
const v = Number(n) || 0;