zoan37 / agentcube

agentcube

Experiments with AI agent avatars in the browser with window.ai.

Run locally

npm run dev

agentcube's Issues

Cap max tokens for generative agent

Right now, memory of generative agent keeps increasing without bound. Should limit memory so don't go over API token limit.

VRM loading doesn't work for Boomboxhead NFT

NFT: https://opensea.io/assets/ethereum/0xb67ff46dfde55ad2fe05881433e5687fd1000312/99
VRM: ipfs://bafybeigt556wa5s562z6glcjzmqza77yux7fb4z674f4nuzyyap2iopavm/36.vrm

I'm using @pixiv/three-vrm (https://github.com/pixiv/three-vrm ) to load VRM with Three.js. However, loading the Boomboxhead VRM results in the following error:

game.js?t=1682621429299:116 RangeError: Maximum call stack size exceeded
    at String.replace (<anonymous>)
    at PropertyBinding.sanitizeNodeName (three.module.js:45062:15)
    at GLTFParser.createUniqueName (GLTFLoader.js:3335:41)
    at GLTFParser.loadNode (GLTFLoader.js:3889:42)
    at GLTFLoader.js:2548:34
    at GLTFParser._invokeOne (GLTFLoader.js:2497:19)
    at GLTFParser.getDependency (GLTFLoader.js:2546:24)
    at GLTFParser.loadSkin (GLTFLoader.js:3615:23)
    at GLTFParser.getDependency (GLTFLoader.js:2594:24)
    at GLTFLoader.js:3938:14
(

I get the same error when using drag-and-drop VRM demos: https://pixiv.github.io/three-vrm/packages/three-vrm/examples/

Loading with @0b5vr/three-vrm-inspector (https://github.com/0b5vr/three-vrm-inspector ) works though (it also uses @pixiv/three-vrm). So need to investigate how three-vrm-inspector is able to get it working.

Edit: vrm-dance-viewer also works: https://github.com/JLChnToZ/vrm-dance-viewer