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
/**
* AgentAvatar × sustained mood → gesture-slot bias — unit tests.
*
* Exercises Stage 3.5 of `_tickEmotion()`: the sustained mood layer (valence ×
* arousal, set via `setMood()`) biasing which gesture *slot* the avatar plays,
* on top of the pre-existing transient-emotion gesture triggers (Stage 3) and
* the facial/postural mood layer (`_applyMoodLayer`). No Three.js render loop
* or real animation clips are needed — `_tickEmotion()` is called directly
* against a minimal fake viewer, mirroring the approach in
* tests/agent-avatar-lipsync.test.js.
*/
import { describe, it, expect, beforeEach } from 'vitest';
let AgentAvatar;
// ── Fakes ────────────────────────────────────────────────────────────────────
// Object3D-shaped root with a no-op traverse — enough for `_findHeadBone()`
// and `_trackBodyToCamera()` to no-op safely without a real skeleton.
function makeRoot() {
return {
isObject3D: true,
children: [],
traverse(fn) {
fn(this);
},
};
}
/**
* Fake AnimationManager. `loadedNames` is the whitelist of clip names that
* "exist" on this rig (isLoaded → true); everything else is unregistered.
* Records every clip name actually handed to play()/crossfadeTo() so tests
* can assert the gesture bias never reaches for a name outside the whitelist.
*/
function makeFakeAnimationManager(loadedNames = []) {
const loaded = new Set(loadedNames);
const played = [];
return {
currentName: null,
isLoaded: (name) => loaded.has(name),
getAnimationDefs: () => [],
play: (name) => {
played.push(name);
return Promise.resolve(true);
},
crossfadeTo: (name) => {
played.push(name);
},
played,
};
}
function makeFakeViewer(am) {
return {
content: makeRoot(),
animationManager: am,
state: {},
};
}
function makeAvatar(am) {
const viewer = makeFakeViewer(am);
const protocol = {}; // unused outside attach()
const identity = { id: 'test' };
return new AgentAvatar(viewer, protocol, identity);
}
// Ticks with a large dt so `_moodApplied` (lerp factor `dt * 1.2`, clamped to
// 1) fully converges to the mood target within a single tick, and a second
// tick is enough for Stage 3.5 (which reads the *previous* frame's applied
// mood, since it runs before Stage 4's lerp) to see the converged value.
function settleMood(avatar, times = 3) {
for (let i = 0; i < times; i++) avatar._tickEmotion(1);
}
beforeEach(async () => {
if (!AgentAvatar) {
const mod = await import('../src/agent-avatar.js');
AgentAvatar = mod.AgentAvatar;
}
});
describe('AgentAvatar — sustained mood → gesture-slot bias', () => {
it('a sustained high-arousal-positive mood biases toward the energetic (celebrate) gesture', () => {
const am = makeFakeAnimationManager(['celebrate', 'defeated']);
const avatar = makeAvatar(am);
avatar.setMood(0.6, 0.8); // clearly up + energetic
settleMood(avatar);
expect(am.played).toContain('celebrate');
expect(am.played).not.toContain('defeated');
});
it('a sustained low-arousal-negative mood biases away from the energetic gesture', () => {
const am = makeFakeAnimationManager(['celebrate', 'defeated']);
const avatar = makeAvatar(am);
avatar.setMood(-0.6, 0.1); // clearly down + subdued
settleMood(avatar);
expect(am.played).not.toContain('celebrate');
// The concern slot resolves to the "defeated" library clip by default —
// confirms the bias picked the opposite-valence gesture, not silence.
expect(am.played).toContain('defeated');
});
it('a neutral resting mood (defaults, setMood never called) triggers no gesture bias', () => {
const am = makeFakeAnimationManager(['celebrate', 'defeated']);
const avatar = makeAvatar(am);
settleMood(avatar); // _mood.active stays false — no setMood() call
expect(am.played).toEqual([]);
});
it('never selects a clip that is not registered on the attached rig', () => {
// Nothing is loaded and no defs are registered — every slot resolves to
// an unregistered clip name. The bias must not hand an unknown name to
// play()/crossfadeTo(); it degrades to the embedded-clip fallback (a
// silent no-op here, since the fake viewer has no `mixer`/`clips`).
const am = makeFakeAnimationManager([]);
const avatar = makeAvatar(am);
avatar.setMood(0.9, 0.95); // maximally energetic + positive
expect(() => settleMood(avatar)).not.toThrow();
expect(am.played).toEqual([]);
});
it('composes with meta.edits.animations — an overridden celebrate slot plays the custom clip', () => {
const am = makeFakeAnimationManager(['victory-dance-custom']);
const avatar = makeAvatar(am);
avatar.setAnimationMap({ celebrate: 'victory-dance-custom' });
avatar.setMood(0.6, 0.8);
settleMood(avatar);
expect(am.played).toContain('victory-dance-custom');
expect(am.played).not.toContain('celebrate');
});
it('a live transient emotion spike still takes priority over the sustained mood bias', () => {
const am = makeFakeAnimationManager(['celebrate', 'defeated']);
const avatar = makeAvatar(am);
// Sustained mood says "down + subdued" (would bias toward `concern`)...
avatar.setMood(-0.6, 0.1);
// ...but a momentary celebration spike (e.g. a skill just succeeded) is
// injected via the public emote path, well above the Stage-3 threshold.
avatar._injectStimulus('celebration', 0.9);
// Only 2 ticks: the transient spike decays below the 0.6 threshold after
// ~2s (DECAY.celebration = 0.18/s), at which point Stage 3 correctly
// falls through to the sustained mood bias below it — that hand-off is
// the intended behaviour, not something this test covers. Within the
// window the spike is still live, mood must not preempt it.
settleMood(avatar, 2);
expect(am.played).toContain('celebrate');
expect(am.played).not.toContain('defeated');
});
});