For our Operations Research course at NMIMS Shirpur, my teammates and I wanted a tool that shows why each simplex step happens, not only the final answer. So I built OR/Studio: a free, open-source Operations Research workbench that runs in the browser and works offline.
What it does
There are 13 solver modules: simplex (including Big-M, two-phase and dual), integer programming, nonlinear programming, transportation, assignment, network models, project planning, queuing, inventory, game theory, Markov chains, simulation and job sequencing.
Every tableau and ratio test comes with a plain-language reason and the rule that was applied. In tutor mode you pick the entering variable and leaving row yourself and get told what went wrong if you're off. In diff mode you type your own working and it finds the first step where you diverged.
Exact arithmetic
Fractions are exact, so the numbers in a tableau match what you'd get by hand. Big-M and prohibited routes use a symbolic M, never a large number.
How I tested it
The solvers are pure TypeScript with no UI dependency, and I tested them against independent checks, not against themselves:
- Linear programs against vertex enumeration and strong-duality checks on random instances.
- Integer programs and assignment against brute force.
- Transportation, max flow and games re-formulated as LPs and compared.
- Shortest-path and spanning-tree algorithms that must agree with each other.
The repo has 452 tests. The hardest part wasn't the algorithms. It was the edge cases: unbounded problems, cycling, degenerate pivots and negative cycles.
What it doesn't do yet
Separable programming isn't implemented, "Save as PDF" goes through the browser's print dialog, and the nonlinear module works in floating point. I've opened these as issues.
Try it
- Live: https://or-studio.antideploy.app
- Code (MIT): https://github.com/Jay-Naik2526/or-studio
I'd like feedback from anyone taking or teaching an OR course, especially bugs and missing methods.
