Repository navigation
Releases: MAGA2010/data-master
Releases · MAGA2010/data-master
Release list
data-master v0.3.0
data-master v0.3.0
This release makes reports easier to share and expands the public evidence for
ambiguous and survey-shaped data.
Highlights
- Every full and profile-only analysis now includes a self-contained HTML report.
- A date-versus-number benchmark proves ambiguous values are held instead of guessed.
- A synthetic survey case study covers labelled categories, missing-value codes and zero-padded identifiers.
- Rank-based omnibus tests now include a bootstrap interval for epsilon squared.
- Report language is derived from the observed results and no longer upgrades null findings.
- Distribution validation now includes macOS with Python 3.13.
- The full install, analysis and verification path was independently exercised in ZCode 3.14.4 on Windows.
Quick start
npm install -g github:MAGA2010/data-master#v0.3.0
data-master setup
data-master skill install
data-master demo --out outputs/first-demoThe release includes a Python wheel, npm tarball, portable .skill archive and
SHA-256 checksums. See the repository README for the evidence contract and
supported inputs.
data-master v0.2.0
data-master v0.2.0
The first public release turns a raw dataset into a reproducible Markdown and
PDF report with an inspectable evidence trail.
Highlights
- Unit-aware group comparisons avoid counting repeated rows as independent samples.
- Cleaning plans preserve raw inputs, record every conversion and replay byte-for-byte.
- Sensitivity analysis and an eight-gate audit keep uncertainty and unresolved decisions visible.
- Resume verifies artifact hashes and rebuilds stale downstream stages.
- Windows and Linux CI cover Python 3.11–3.13.
- Install the CLI from GitHub through npm or a Python wheel.
- Install the Codex skill with
data-master skill install.
Quick start
npm install -g github:MAGA2010/data-master#v0.2.0
data-master setup
data-master skill install
data-master demo --out outputs/first-demoThe release includes a Python wheel, an npm tarball, a portable .skill
archive and SHA-256 checksums. See the repository README for supported inputs,
current limitations and the full evidence contract.