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.

1.3 KB
Raw Blame
# garment-forge is a CPU orchestrator: no torch, no CUDA. It calls the
# deployed GPU workers (mesh generation, rigging) over HTTP and does pure
# glTF/geometry work locally. Everything pinned or bounded, matching the
# convention set after the 2026-07-16 unpinned-transformers outage.

# ── Geometry / glTF ──────────────────────────────────────────────────────────
numpy==1.26.4
trimesh==4.5.3
pygltflib==1.16.2
# trimesh scene-graph operations (Scene.add_geometry / export) use networkx.
networkx==3.3
# GLB export path in trimesh imports scipy for spatial ops on some meshes.
scipy==1.13.1

# ── Imaging (reference-image decode, WEBP thumbnails) ────────────────────────
Pillow>=10.3.0

# ── Server ────────────────────────────────────────────────────────────────────
fastapi==0.115.0
uvicorn[standard]==0.32.0
pydantic>=2.0.0
httpx>=0.27.0

# ── GCP (storage for assets/jobs, ADC token for the Vertex image lane) ───────
google-cloud-storage==2.18.0
google-auth>=2.30.0