Back
SiTech Professional Insights
Flint: Microsoft Research's Visualization Language for AI Agents
SiTech Team2 წთ. საკითხავი

Flint: Microsoft Research's Visualization Language for AI Agents

Microsoft Research has released Flint, an open-source visualization intermediate language that lets AI agents turn compact, human-editable chart specs into polished charts across five rendering backends.

Microsoft Research has published Flint, an open-source visualization intermediate language aimed at a new and specific problem: letting AI agents produce expressive, good-looking charts without hand-tuning dozens of low-level parameters.

A compact spec instead of hundreds of parameters

In Flint a chart is described by a short, human-editable specification with three parts: the data, a set of semantic types, and the chart spec itself. Semantic types — Rank, YearMonth, Delta, Temperature, Profit and others — capture what a field means, not only its format. The compiler uses them to infer parsing, scales, axis formatting, color schemes and the midpoint of diverging scales, so the author never writes those explicitly.

From that input Flint compiles a complete backend-native spec. The project page shows a heatmap of new-user gains by game and month turning into a full Vega-Lite definition, with temporal scales, band sizing and a red-blue diverging scheme centered on zero.

Layout that adapts itself

Flint handles layout too. An elastic layout model built on "banking" principles from visualization research manages sizing, spacing and arrangement so a chart fits the canvas. A grouped bar chart with many series is stretched and its bands narrowed automatically instead of overflowing. Changing a design means switching the chart type and rebinding visual encodings; the compiler cascades those choices into the low-level settings. The page illustrates this with a faceted bar chart of the 2000 U.S. Census population distribution by gender and age turning into a population pyramid.

Five backends, fifty chart types, one interface

Flint supports 50 chart types across Vega-Lite, ECharts, Chart.js and Plotly, with 123 backend-specific examples in its gallery, and can emit native editable Excel charts through Office.js. Each backend keeps its native strengths: ECharts for hierarchical sunbursts, Plotly for statistical traces, Excel for charts embedded in a workbook.

Versions and themes

Version 0.4.0 added 38 Plotly chart types and 18 native, editable Excel chart templates. Version 0.5.0 introduced formal visual themes with ten presets, custom ThemeSpec authoring and preset inheritance; 0.5.1 extended the themes to the Plotly backend and added a Vega-Lite calendar heatmap. Flint is built by Microsoft Research in collaboration with the IDEAS Lab at Renmin University of China, and is open source on GitHub.

SSiTech

SiTech — AI-powered web development

We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.