Chartifact – Declarative, interactive data documents (microsoft.github.io)

Chartifact is a new low-code document format for building declarative, interactive data documents — reports, dashboards, and presentations — that “travel like a document and work like a mini app.” It’s designed to be used alongside LLMs so analytic conversations can be captured as shareable, remixable artifacts. The project includes a document schema and ecosystem (VS Code extension, web editor, examples, and HTML export) that make it easy to author, preview, and distribute interactive pages without custom JavaScript or heavy engineering. That makes it valuable to AI/ML teams wanting reproducible, conversational workflows, rapid prototyping, and lighter-weight sharing of analyses. Technically, Chartifact exposes a set of composable components (Markdown text with dynamic placeholders, inputs, editable tables, Vega/Vega-Lite charts, Mermaid diagrams, dynamic images, and presets) wired by reactive variables and a signal bus so every element stays in sync. Documents can be authored as human-readable Markdown with embedded JSON blocks or as strict JSON for programmatic generation. Dynamic bindings let variables flow into chart specs, text, URLs and REST calls; Vega transforms and REST integration handle reshaping and fetching data. Security is built in: runtime rendering occurs in sandboxed iframes, no custom JS or raw HTML is allowed, and CSS parsing is XSS-defensive. Overall, Chartifact bridges LLM-enabled authoring and secure, interactive analytics in a portable format.