A Windows-first scientific research workstation and Performance OS, built with Python and native Qt Widgets.
DynamisScience brings two local workflows into one desktop application:
- Research: project and study authoring, dataset admission and quality control, cohort resolution, frozen analysis plans, qualified runs, diagnostics, evidence, lineage and versioned reports.
- Performance OS: Squad Command, Athlete 360, planning, monitoring, performance tests, match/pose inspection, research links and practitioner reports.
The five research workspaces are Overview & Protocol, Dataset & Cohort, Exploration & Signals, Analysis Plan & Runs, and Evidence & Report. The Performance OS follows Today → Plan → Monitor → Test → Research → Report.
The interface uses PySide6/Qt Widgets, pyqtgraph for interactive signals, and Matplotlib for scientific figures and PNG/PDF/SVG exports. Scientific methods and persistent application logic run directly in Python. The desktop application does not require Node.js, a browser runtime, or a local HTTP server.
Additional native evidence: Research workstation, study authoring, Athlete 360, and Match & Pose.
Install Python 3.12+ and uv, then:
uv sync --locked --directory services/science
$env:DYNAMISSCIENCE_DATA_ROOT = 'D:\datasets\DynamisScience'
uv run --directory services/science python -m dynamis_science.desktopThe private data root must be an absolute directory outside the repository. On first launch, the application initializes the SQLite WAL research store there or prompts you to choose a root. Admitted research snapshots and user records remain outside Git.
To build the Windows one-directory application and per-user installer:
powershell -NoProfile -ExecutionPolicy Bypass -File services/science/scripts/build_windows.ps1
powershell -NoProfile -ExecutionPolicy Bypass -File services/science/scripts/test_windows_installer.ps1The installer includes a deterministic synthetic fixture only. It does not include private Mendeley or MoveSmart records. Installation and uninstall do not remove the configured data root. Qt and Python dependency notices are available under Help → Open-source licenses and in the licensing notice.
- Mendeley and MoveSmart snapshots retain their checksums, source rights and limitations. Displayed licences come from each admitted source manifest, never from hardcoded labels. The MoveSmart manifest records CC0-1.0 (Zenodo
license.id = cc-zero, re-checked 2026-11-10); DataCite also lists a "© 2025 The Authors" rights statement, so owner review is required before any commercial reliance. Pseudonymous participant metadata stays private regardless of licence. - Synthetic
DS-SYNTH-001data are labeled as demonstration data and are never empirical evidence. - Unsupported models remain marked as prototypes or unavailable. The app does not produce athlete predictions or injury-risk estimates from unsupported methods.
- Studies, cohorts, plans, runs, evidence, figures, reports and audit history persist in SQLite. Scientific tables and artifacts use Parquet and the configured private filesystem.
- Database integrity checks, verified online backups, isolated restore rehearsals and independent real-data qualification are available from the native Tools menu. Qualification uses the qualified Python methods and an optional local R/lme4 reference; without R, the application records an honest blocked state.
The current packaging measurements are recorded in G4 evidence. The migration status and remaining verification are tracked in RES-604 progress.
$env:QT_QPA_PLATFORM = 'offscreen'
uv run --directory services/science pytest -qCopyright © 2026 Julio Rodriguez. All rights reserved. This repository is public for portfolio viewing only. No permission is granted to copy, distribute, modify, deploy or otherwise use any part of it without prior written permission. See LICENSE for the full terms.
