A real-time collaborative editor for web-native scientific articles. Co-write with your team, drop in interactive figures, math and citations, and publish with one click - straight from a Hugging Face Space.
Three co-authors writing an article on Transformer attention - live cursors, block math, agent edits and a real publish, in under a minute.
Built for science: not a generic doc editor with math bolted on.
Co-write with visible cursors and per-user colors, powered by Y.js CRDTs. No merge conflicts, ever.
Self-contained D3.js embeds instead of static PNGs. Readers explore your data, not just look at it.
Describe a chart, get a working D3 embed, iterate on it in a chat loop. Powered by HF Inference Providers.
KaTeX equations, BibTeX bibliographies, numbered figures, sidenotes and glossary terms out of the box.
Static HTML page, print-ready PDF and social thumbnail, generated in one click while others keep editing.
Anchor a comment on any selection and resolve threads inline, like a doc review that ships.
Every article ships a Markdown twin at /llms.txt, advertised in robots.txt, so agents can read your work.
Automatic dark mode, and a hue slider that retints the whole article - banner included - in one drag.
Deploys as a Space, signs in with HF OAuth, persists to HF datasets. Duplicate and you own the stack.
No API keys, no servers to run. Duplicate the Space and invite your co-authors.
The editor runs entirely on your Hugging Face Space. Sign-in uses HF OAuth, documents persist to your own HF dataset, and the AI features (chat agent, embed studio) run on each user's own Inference Providers quota - nothing to configure.
The embed studio turns a plain-language prompt - or a pasted dataset - into a self-contained D3 visualization, live in your article. Iterate in a chat loop until it looks right: the banner and the chart in this video were both made that way, without writing a line of code.
Real figures from the template's demo, running live below. Pick a brand color: every palette re-derives itself from it, dark mode included.
The same rendering engine is available as an Astro + MDX template. Write chapters in MDX, craft every D3 embed by hand, version your paper with git and export PDF or LaTeX - the editor and the template produce the same output.
HuggingFaceTB
HuggingFaceM4
transformers-community
npx create-research-article my-paper Real research articles from the Hugging Face community, all built on this engine.
The secrets to building world-class LLMs
Open data is all you need
1M python loc, 400+ models