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.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
// One GLTF loader for the whole studio, wired for the compression the real world
// ships: Draco (most Sketchfab / pipeline exports) and EXT_meshopt_compression
// (what most server-side GLB bakers emit, three.ws avatars included). A studio
// that loads arbitrary caller-supplied GLBs has to handle both or it fails on
// perfectly valid files.
//
// Decoder binaries are fetched from a CDN by default because they are WASM blobs
// that do not belong in an npm package's install footprint. Self-host them and
// point `dracoPath` at your own copy when you would rather not reach out.
import { GLTFLoader } from 'three/addons/loaders/GLTFLoader.js';
import { DRACOLoader } from 'three/addons/loaders/DRACOLoader.js';
/** Where the Draco decoder (wasm + js) is fetched from when nothing else is set. */
export const DEFAULT_DRACO_PATH = 'https://cdn.jsdelivr.net/npm/[email protected]/examples/jsm/libs/draco/gltf/';
let _loader = null;
let _dracoPath = DEFAULT_DRACO_PATH;
let _meshoptPromise = null;
/** Point the Draco decoder at a different host. Call before the first load. */
export function setDracoPath(path) {
if (typeof path === 'string' && path) _dracoPath = path.endsWith('/') ? path : `${path}/`;
}
function meshoptDecoder() {
if (!_meshoptPromise) {
_meshoptPromise = import('three/addons/libs/meshopt_decoder.module.js')
.then((m) => m.MeshoptDecoder)
.catch(() => null);
}
return _meshoptPromise;
}
/**
* The shared loader. Draco is attached synchronously; meshopt resolves async and
* attaches as soon as it is ready, so plain and Draco GLBs never wait on it.
* @returns {GLTFLoader}
*/
export function sharedGLTFLoader() {
if (_loader) return _loader;
_loader = new GLTFLoader();
const draco = new DRACOLoader();
draco.setDecoderPath(_dracoPath);
_loader.setDRACOLoader(draco);
meshoptDecoder().then((d) => {
if (d) _loader.setMeshoptDecoder(d);
});
return _loader;
}
/**
* The shared loader, once every optional decoder that can attach has attached.
*
* `sharedGLTFLoader()` returns before meshopt resolves, which is right for the
* queue (a plain GLB should not wait on a WASM download it will never use) and
* wrong for a one-shot parse: hand a meshopt-compressed GLB to a loader whose
* decoder is still in flight and it throws "no DRACOLoader instance provided"'s
* meshopt equivalent on a file that is perfectly valid. Anything that parses a
* single caller-supplied GLB should await this instead.
*
* @returns {Promise<GLTFLoader>}
*/
export async function sharedGLTFLoaderReady() {
const loader = sharedGLTFLoader();
await meshoptDecoder();
return loader;
}
/** Drop the cached loader (tests, or after changing the Draco path mid-session). */
export function resetLoader() {
_loader = null;
}