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.

fix(avatars): load a stored avatar through the same-origin GLB proxy

nirholasb4d2f710eba6
diff --git a/src/avatar-embed.js b/src/avatar-embed.js
index 3e351dfda..50aa52700 100644
--- a/src/avatar-embed.js
+++ b/src/avatar-embed.js
@@ -954,9 +954,9 @@ async function resolveAvatar(params) {
const r = await fetch(`/api/avatars/${encodeURIComponent(id)}`, { credentials: 'include' });
if (!r.ok) throw new Error(`avatar ${id} not found`);
const { avatar } = await r.json();
- if (!avatar?.url) throw new Error('avatar has no model_url');
+ if (!avatar?.id) throw new Error(`avatar ${id} not found`);
return {
- modelUrl: optimizedUrl(avatar.url, params),
+ modelUrl: storedAvatarUrl(avatar, params),
name: avatar.name || null,
id: avatar.id,
handle: null,
@@ -997,14 +997,42 @@ async function resolveAvatar(params) {
return null;
}
-function optimizedUrl(baseUrl, params) {
+// A stored avatar always renders through the same-origin GLB proxy
+// (api/avatars/[id]/[action].js), never through the `url` the API hands back.
+// That `url` is a presigned r2.cloudflarestorage.com read for a private avatar
+// and the public r2.dev domain for a shared one: the first sends no
+// access-control-allow-origin at all, and the second only allows the three.ws
+// origin, so GLTFLoader is CORS-blocked in every other context this frame runs
+// in (a dev server, a preview host, an embedding page on another domain). The
+// proxy is same-origin, carries the session cookie through the permission gate,
+// and falls back to the public bucket when the signed read is refused.
+function storedAvatarUrl(avatar, params) {
+ const proxy = `/api/avatars/${encodeURIComponent(avatar.id)}/glb`;
+ if (!wantsOptimization(params)) return proxy;
+ // /api/avatar/optimize is unauthenticated and resolves an id only for a
+ // public or unlisted avatar (resolving a private one by id was an IDOR), so
+ // a private avatar stays on the unoptimized proxy rather than 404ing.
+ const shared = avatar.visibility === 'public' || avatar.visibility === 'unlisted';
+ if (!shared) return proxy;
+ return optimizedUrl(avatar.id, params);
+}
+
+function wantsOptimization(params) {
+ return !!(
+ params.get('lod') ||
+ params.get('textureSize') ||
+ params.get('morphs') ||
+ params.get('draco') === '1'
+ );
+}
+
+function optimizedUrl(avatarId, params) {
const lod = params.get('lod');
const textureSize = params.get('textureSize');
const morphs = params.get('morphs');
const draco = params.get('draco');
- if (!lod && !textureSize && !morphs && draco !== '1') return baseUrl;
const u = new URL('/api/avatar/optimize', location.origin);
- u.searchParams.set('src', baseUrl);
+ u.searchParams.set('id', avatarId);
if (lod) u.searchParams.set('lod', lod);
if (textureSize) u.searchParams.set('textureSize', textureSize);
if (morphs) u.searchParams.set('morphs', morphs);