@gpx/estimates

Monte Carlo project duration forecasting with a local web UI.

Edit tasks with optimistic, most-likely, and pessimistic estimates, dependencies, and team resources. Set project phase and historical bias, then view probabilistic forecasts (P50/P85/P95 and a CDF chart).

The engine uses Beta-PERT sampling, Reference Class Forecasting calibration, and Cone of Uncertainty enforcement.

How to run

The preferred way to use estimates is via npx — no install or clone required:

npx @gpx/estimates                  # starts UI, seeds ./tasks.json if missing
npx @gpx/estimates my-project.json  # use a specific project file
npx @gpx/estimates --no-open        # skip auto-opening browser

For repeated use, you can optionally install globally:

npm install -g @gpx/estimates
estimates

CLI options

| Flag | Description | |------|-------------| | [project-file] | Path to project JSON (default: ./tasks.json) | | --port <N> | Server port (default: 3847) | | --no-open | Do not open browser automatically | | --iterations, -n <N> | Monte Carlo runs (default: 20000) | | --help, -h | Show usage |

Requirements

• Node.js >=24.18.0
• The server binds to 127.0.0.1 only (local tool, no network exposure)

Project file

Projects are stored as JSON. On first run, if the project file does not exist, an empty project is created with phase Concept, historical bias 80%, start date today, and zero engineers and designers.

Developing from source

git clone [email protected]:Gpx/estimates.git
cd estimates
npm install
npm run dev              # run via tsx, no build step
npm run build && npm start   # or compile then run
npm test

Releasing

From a clean working tree, with npm logged in to registry.npmjs.org:

npm run release

The release script prompts for a patch (bug fix), minor, or major bump, runs tests, updates package.json, creates a git commit and tag, pushes to GitHub, and publishes to npm.

First-time publish only:

npm login --registry https://registry.npmjs.org/
npm run release

License

MIT