An AI agent chained two Hugging Face Spaces to build a 3D Paris monument gallery by calling documented APIs without manual integration.

A coding agent generated images of Paris monuments using ideogram-ai/ideogram4, then reconstructed each as a 3D Gaussian splat (.ply file) using VAST-AI/TripoSplat, and deployed the result as a static interactive Space with a Three.js viewer.

Every Gradio Space now exposes a plain-text agents.md file that tells an agent the exact API schema, call templates, file-upload method, and auth requirements—eliminating the need for client libraries or hardcoded integrations. The agent compressed the output splats to .ksplat (~3× smaller) and auto-framed each monument.

The barrier to chaining multimedia models (image generation, 3D reconstruction) from different organizations is largely removed; an agent can now assemble documented, callable blocks from the Hub the same way it combines code libraries, with zero integration code required.