Skip to content

OpenMat v0.01

Latest

Choose a tag to compare

@dhve dhve released this 02 Aug 23:07

The first packaged build of OpenMat: the open-source, local-first, AI-native alternative to Mathematica. A Rust symbolic kernel that runs entirely on your machine, a Mathematica-style notebook, and natural language input built in. Website: https://openmat.tools

Install (macOS, Apple Silicon)

Download the dmg and drag OpenMat to Applications. The build is unsigned, so the first launch needs right-click on OpenMat.app, then Open, then Open again in the dialog. macOS remembers after that.

What works

  • Mathematica-style notebook: 2D math input, typeset output, In/Out numbering, right-edge cell brackets (click to collapse), Title/Section/Text cells, save and open
  • Symbolic kernel written in Rust: arithmetic and simplification, D, Integrate (indefinite and definite, with trig rules and a numeric fallback), Solve, Expand, Factor, Simplify, Table
  • Persistent kernel session: a = 5 and f[x_] := x^2 stay defined across cells until Clear, and resolve inside plots and solvers
  • Plot and ListPlot with adaptive sampling, legends, and discontinuity handling; NDSolve for first and second order ODEs; Manipulate sliders that re-solve live as you drag
  • Natural language input: type = at the start of any empty cell, or use the input bar docked at the bottom. One sentence ("make me a notebook exploring the driven pendulum with sliders") generates a full notebook of titles, prose, equations, and interactive plots, evaluated in place. Bring your own Anthropic API key or run a local Ollama model; configure either under Settings
  • Math symbol keyboards: every math cell has a full virtual keyboard, and the language bar has a symbol palette
  • One engine everywhere: the exact same kernel runs in-process on desktop and compiles to WebAssembly for the browser build

Known limits

  • ODEs are scalar first/second order (no coupled systems yet)
  • Solve covers polynomial equations through quadratics
  • The dmg is unsigned and built for Apple Silicon only

Dual-licensed under MIT or Apache-2.0.