v1.5.31¶
Release date: 2026-07-21
Summary¶
This release turns the desktop beta into a more complete academic workbench. Windows, macOS, and Linux now build the same Qt interface, and the new desktop-all profile installs the supported Python-side GUI, reporting, MDMP, research, edge, SBML, and FMI dependencies together.
The release also adds independent cross-scale reference workflows. These tools let researchers inspect or compare external mechanistic and evidence sources without silently replacing, tuning, or validating the IINTS digital-patient equations.
Complete Desktop Runtime¶
- Added
iints-sdk-python35[desktop-all]as the complete Python desktop install. - Included PySide6, Plotly, Matplotlib/reporting, MDMP cryptography, serial/edge support, research/AI libraries, libRoadRunner, and FMPy.
- Updated both the Python app updater and Rust/Tauri updater to use the same package profile.
- Changed all three desktop beta platforms to the rich Qt workbench.
- Added explicit PyInstaller collection for lazy Plotly, libRoadRunner, and FMPy imports.
- Kept direct
.exe,.dmg, and Linux executable downloads under the stabledesktop-beta-latestprerelease.
COPASI, OpenCOR, and Ollama remain separate system applications. Their connectors and readiness checks are included, but their binaries and AI model files are not silently downloaded. This preserves visible licensing, native-code trust, storage, and reproducibility decisions.
Academic Reproducibility¶
- Added an academic bundle exporter with RO-Crate metadata, checksums, run/software metadata, source references, and unresolved review checks.
- Added machine-readable software citation metadata through
CITATION.cff. - Added a clearer publication workflow and documentation hub for external researchers.
Mechanistic And Cross-Scale Labs¶
- SBML: safe XML inspection and optional independent libRoadRunner time-course simulation.
- COPASI: task inspection and explicitly confirmed
CopasiSEexecution. - CellML: safe model inspection and OpenCOR standards validation.
- FMI: bounded FMU archive inspection and explicitly trusted FMPy execution for device-physics models.
- BindingDB: read-only affinity evidence retrieval by UniProt accession.
- Tauri/Rust: validated command boundaries, fixed argument schemas, HTTPS allowlists, output provenance, and visible readiness states.
Scientific Boundaries¶
- External models are independent references; they do not automatically alter IINTS physiology parameters.
- AlphaFold confidence is structural confidence, not pathogenicity or physiological effect size.
- ClinVar labels are preserved as condition-specific classification context and never converted into receptor-function scalars.
- Variant simulation remains blocked without an explicit quantitative functional scalar and provenance; a missing record is not treated as benign.
- Binding affinity, expression, variants, and pathways are context for research interpretation, not treatment logic.
- Native FMU code and configured COPASI tasks require explicit execution consent.
- Generated evidence does not establish clinical validation, regulatory compliance, or medical-device safety.
Verification¶
Local release verification completed with:
- full Python suite:
763 passed, 2 skipped - mypy:
Success: no issues found in 229 source files - Rust/Tauri unit tests:
3 passed - strict MkDocs build and Python wheel/sdist build
- Tauri frontend static checks
- frozen macOS Qt app
--smoke-full, including Plotly, libRoadRunner, FMPy/SUNDIALS, and Torch imports - a verified 517 MB macOS DMG with a generated SHA-256 checksum
Platform beta artifacts are rebuilt and smoke-tested again inside GitHub Actions before upload.
Safety¶
IINTS-AF remains pre-clinical research and educational software. It is not a medical device and must not be used for diagnosis, insulin dosing, glucagon dosing, treatment decisions, or real-time patient care.