How an Agent Built a 3D Paris Gallery by Chaining Two Hugging Face Spaces

Source: https://huggingface.co/blog/mishig/spaces-agents-md Back to Articles

• The building-block economy comes for multimedia
• Every Space is a building block, viaagents\.md
• The worked example: Paris monuments → splats
• Why this matters
• Try it yourself

An agent built a 3D Paris gallery from two Hugging Face Spaces.

I asked a coding agent to build a beautiful website showcasing the monuments of Paris as 3D Gaussian splats. I never opened an image generator. I never touched a 3D reconstruction tool. The agent produced every asset (the imagesandthe 3D splats) by calling two Hugging Face Spaces directly, then wired them into a cinematic viewer.

Here’s the result, live as a static Space:

👉mishig/monuments-de-paris

This post is abouthowthat’s possible now, and why I think it’s a preview of how a lot of multimedia software gets built from here on.

https://huggingface.co/blog/mishig/spaces-agents-md#the-building-block-economy-comes-for-multimediaThe building-block economy comes for multimedia

Mitchell Hashimoto recently described a shift he calls thebuilding block economy: the most effective path to software is no longer a polished monolith, but small, well-documented components that others (increasinglyagents) can assemble. His key observation: AI is okay at building everything from scratch, but it isreally good at gluing togetherproven pieces.

That thesis has mostly been told withcodelibraries. But the same forces are hittingmultimedia AI. The hard part of using a state-of-the-art image model, a video model, a TTS model, or a 3D reconstruction model was never the model. It was the integration: SDKs, weights, GPUs, input formats, polling. If each model were instead a documented, callable block, an agent could glue them together the same way it globs together npm packages.

That’s exactly what Hugging Face Spaces have quietly become.

https://huggingface.co/blog/mishig/spaces-agents-md#every-space-is-a-building-block-via-agentsmdEvery Space is a building block, viaagents\.md

The Hub hosts thousands of state-of-the-art models (a huge share of themopen-weights), and most are deployed as interactiveSpaces. As of now, every Gradio Space also exposes a plain-textagents\.mdthat tells an agentexactlyhow to call it:

curl https://huggingface.co/spaces/VAST-AI/TripoSplat/agents.md

returns everything needed in one shot: the schema URL, the call and poll templates, how to upload files, and the auth hint:

API schema:   GET  .../gradio_api/info
Call endpoint: POST .../gradio_api/call/v2/{endpoint} {"param_name": value, ...}
Poll result:  GET  .../gradio_api/call/{endpoint}/{event_id}
File inputs:  POST .../gradio_api/upload -F "[email protected]"
Auth:         Bearer $HF_TOKEN

No client library. No hardcoded integration. An agent reads that, and it can drive the Space end to end. Set anHF\_TOKENand you’re going.

The real unlock ischaining: the output of one Space becomes the input to the next. Prompt → image → 3D. That’s the whole pipeline behind this gallery.

https://huggingface.co/blog/mishig/spaces-agents-md#the-worked-example-paris-monuments-%E2%86%92-splatsThe worked example: Paris monuments → splats

The agent chained two Spaces:

1. Image:ideogram\-ai/ideogram4turned each monument into a clean, dark-background “specimen” shot (and the Eiffel Tower into a little diorama on a plinth). Prompt in, image out.
2. Splat:VAST\-AI/TripoSplatreconstructed a 3D Gaussian splat (\.ply) from each single image. Image in, 3D out.

Generated image

Reconstructed splat

The six source images the agent generated, all isolated on black, ready for single-image 3D reconstruction:

From there the agent did the “glue” work too. It noticed TripoSplat outputs are Y-down and flipped them upright, auto-framed each monument, compressed the\.plyfiles to\.ksplat(~3× smaller, so they load fast), built a Three.js viewer with a scroll-to-switch and drag-to-rotate UI, and deployed the whole thing as a static Space. The only human inputs were taste-level: “make it zoomed out,” “replace the obelisk with something better for splatting,” “the transition lingers too long.”

Several of those steps werethe agent reacting to reality. A wide glass pyramid splats poorly. A thin obelisk is dull. A single-view reconstruction infers the back. That is exactly the “outsourced R&D, fast iteration” loop the building-block economy predicts, except the R&D was a conversation.

https://huggingface.co/blog/mishig/spaces-agents-md#why-this-mattersWhy this matters

• **Models become composable.**A SOTA splat model and a SOTA image model, from different orgs, chained with zero integration code. The Hub’s open-weights catalog turns into a library of callable multimedia primitives.
• Agents prefer what’s documented and reachable.agents\.mdmakes a Space trivially reachable, so an agent will pick it over a model it has to set up by hand. That is the same dynamic Hashimoto flags for open-source libraries.
• The barrier was integration, and it’s largely gone.“Turn a prompt into a rotating 3D monument” used to be a project. Here it was a step in a pipeline.

https://huggingface.co/blog/mishig/spaces-agents-md#try-it-yourselfTry it yourself

Point your own agent at a Space’sagents\.mdand let it cook:

# image generation
curl https://huggingface.co/spaces/ideogram-ai/ideogram4/agents.md
# single-image to 3D gaussian splat
curl https://huggingface.co/spaces/VAST-AI/TripoSplat/agents.md

Paste either link into your coding agent (Claude Code, etc.), set yourHF\_TOKEN, and ask it to build something. The full, reproducible pipeline for this gallery, the scripts that hit those twoagents\.mdendpoints, lives in theSpace repo.

The building blocks are sitting right there on the Hub. The agents already know how to glue.