About

Help-text fix plus a codebase-wide maintainability pass with no change to statistical results.

• Power sample-size help field no longer resolves to a YAML boolean, so its tooltip is reachable again
• API routes split one file per endpoint, with registry dispatch for trial designs and model families
• Help copy moved out of R source into YAML so wording edits no longer touch code
• End-to-end contract tests added for previously untested power and analysis paths

Containment for unsafe N-of-1 basket inference, dependency advisories, and guide corrections.

• Unsafe N-of-1 basket inference contained
• js-yaml and postcss raised past published advisories
• Crossover power guide text corrected to match the implementation

N-of-1 correctness fixes and dependency advisory updates.

• N-of-1 standard-error and analysis corrections
• brace-expansion and js-yaml advisories cleared

N-of-1 power and analysis available end to end.

• N-of-1 power and analysis wired through API and GUI
• Trial-design naming aligned across the platform

First stable release. Power and sample-size surfaces renamed design-first (breaking change).

• Power and sample-size functions and endpoints renamed design-first
• N-of-1 GUI added
• Basket MCMC defaults to at least two chains so R-hat is computable
• One- and two-sided options honored for N-of-1

Crossover and multi-arm statistical corrections from the review stream.

• Paired analytic and GEE paths reconciled for binary crossover
• Conjunctive power for multi-arm parallel designs
• Diagnostic generation exposed through API and GUI
• Dormant, unsupported crossover features rejected at the API instead of silently ignored

Umbrella control selection, per-substudy diagnostics, and recorded cloud deployments.

• Umbrella control-arm selector with per-substudy diagnostics
• Crossover non-inferiority margin guarded on every path against a zero margin
• Cloud Run deployments recorded against the repository

Statistical review remediation across power defaults and model reporting.

• One-sided tests default to alpha 0.025 rather than 0.05
• Family-aware QIC reporting for GEE
• Correlation-aware within-substudy umbrella power
• Per-stratum GLMM and GLM handling corrected

Crossover mixed-model and dropout handling fixes.

• Crossover LMM path corrected
• Post-dropout observation handling
• Zero-variance results no longer produce NaN

Simulation defaults raised and basket diagnostics added.

• Per-arm dropout supported
• Default simulation replicates raised from 1000 to 5000
• Posterior-predictive diagnostics for basket models

Intent disclaimer lifecycle and accessibility corrections.

• Disclaimer acknowledgment expires after 12 hours so users re-confirm
• Disclaimer recap stays mounted so its accessibility references always resolve

Disclaimer banner presentation fixes.

• Collapsed disclaimer with a tooltip on the blocked Next action
• Spacing corrections below the collapsed banner

Intent disclaimer and enforcement of supported calculation paths.

• Intent disclaimer gate added before analysis
• Method selector honored by the API rather than silently substituted

Binary power stability fix.

• Separated binary fits handled instead of failing

Privacy controls, security hardening, and accessibility work.

• Client-side privacy gate for column selection and anonymization
• Security headers, upload limits, and immediate session deletion
• Wall-clock compute guard and capped basket MCMC parameters
• Results announced to assistive technology, with reduced-motion scrolling respected

Release-readiness pass with synchronized versions, documentation, CI reporting, and maintainability updates.

• Version synchronized across the R package, frontend, and lockfile
• Root contributing guide and release-facing documentation cleanup
• Frontend unit and R package coverage summaries added to CI
• Wizard pages decomposed into smaller step components without changing behavior

Statistical guards, analysis fixes, accessibility improvements, and clearer example data.

• Defective basket-model combinations blocked pending author review
• Parallel analysis and post-hoc term selection fixes
• Accessible error, focus, tooltip, and result-announcement behavior

Umbrella analysis, statistical validation, privacy controls, and production hardening.

• Umbrella power and analysis supported for shared and substudy-specific controls
• Client-side privacy gate and immediate session deletion
• Rate limiting, compute timeout, production startup guard, and optional file session store

Privacy gate, security hardening, trial design fixes from clinical statistician review, and CI/CD improvements.

• Client-side privacy gate: column selection and subject ID anonymization before data leaves the browser
• Security headers, scoped upload limits, session ID validation, and immediate session deletion
• Trial design terminology corrections (Basket, Umbrella, N-of-1, Crossover) and icon overlap fixes
• CI pushes tested Docker image to both Docker Hub and GHCR instead of rebuilding
• Explicit carry column mapping for crossover designs, new unit tests

First release. Delivers TRL 4 (Docker image with full functionality) for the REPO4EU platform.

• Power Calculator for 5 trial designs with power curve grids for all 7 design/endpoint combos
• Analysis Wizard: CSV import, descriptive stats, diagnostic plots, GLMM/GEE fitting, post-hoc contrasts
• 9 interactive statistical guide visualizations with redesigned controls
• 136 tests (52 R unit, 67 Playwright E2E, 17 Docker smoke) and CI with 7 jobs
• Security audit, bearer token auth, Swagger UI via nginx, About page with BibTeX citation