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
# Make-It-Animatable (jasongzy, MIT) runtime set, mirrored from its
# requirements.txt at the commit pinned in the Dockerfile, minus training-only
# extras (tensorboard, ipykernel, ipympl, fake-bpy-module).
#
# torch/torchvision install in their own Dockerfile layer BEFORE this file so
# anything that imports torch at build time already has it. torch-cluster is
# pinned to the exact pyg wheel build: the bare "torch-cluster" spec let pip
# pick the PyPI sdist, whose isolated build env has no torch and fails.
-f https://data.pyg.org/whl/torch-2.1.2+cu121.html
torch-cluster==1.6.3+pt21cu121
--extra-index-url https://miropsota.github.io/torch_packages_builder
pytorch3d==0.7.7+pt2.1.2cu121
einops
timm
PyMCubes
bpy==4.3.0
gradio>=5.25.0
matplotlib
numpy~=1.26.4
plyfile
potpourri3d
shapely
spaces
tqdm
# Held below 5 on purpose. The running revision was built on trimesh 4.x and a
# bare >=4.4.0 now resolves to 5.0.0, so the next rebuild would quietly ship a
# major-version jump in the loader both MIA and rig_glb parse meshes with.
# Raising it is a deliberate change to verify on the GPU, not a side effect of
# redeploying. The local suites pass on 4.12.2 and 5.0.0 alike.
trimesh>=4.4.0,<5
# Rig authoring + blendshape transfer (rig_glb.py / blendshapes.py)
pygltflib==1.16.2
scipy>=1.11.0
# Server. fastapi rides gradio's requirement (a hard ==0.115.0 pin conflicts
# with gradio>=5.25); the API surface this service uses is stable across the
# 0.115+ line.
fastapi>=0.115,<1
uvicorn[standard]>=0.32,<1
pydantic>=2.0.0
httpx>=0.27.0
# GCP
google-cloud-storage==2.18.0