iints-sdk-python35 1.5.32__tar.gz → 1.5.34__tar.gz

{iints_sdk_python35-1.5.32 → iints_sdk_python35-1.5.34}/CITATION.cff

@@ -6,8 +6,8 @@ authors:
   - family-names: "Bobbaers"
     given-names: "Rune"
     email: "rune.bobbaers@gmail.com"
-version: "1.5.32"
-date-released: "2026-07-23"
+version: "1.5.34"
+date-released: "2026-08-09"
 repository-code: "https://github.com/python35/IINTS-SDK"
 url: "https://iints.org"
 license: "Apache-2.0"

{iints_sdk_python35-1.5.32 → iints_sdk_python35-1.5.34}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: iints-sdk-python35
-Version: 1.5.32
+Version: 1.5.34
 Summary: A pre-clinical Edge-AI SDK for diabetes management validation.
 Author-email: Rune Bobbaers <rune.bobbaers@gmail.com>
 License-Expression: Apache-2.0
@@ -22,10 +22,11 @@ License-File: LICENSE
 License-File: NOTICE
 License-File: LICENSE-MIT-IINTS-LEGACY
 Requires-Dist: certifi>=2024.2.2
-Requires-Dist: cryptography<50.0.0,>=48.0.1
+Requires-Dist: cryptography<51.0.0,>=50.0.0
 Requires-Dist: defusedxml<1.0.0,>=0.7.1
-Requires-Dist: fastapi!=0.136.3,<0.139.1,>=0.115.0
+Requires-Dist: fastapi!=0.136.3,<0.141.2,>=0.115.0
 Requires-Dist: numpy<3.0.0,>=1.24.0
+Requires-Dist: packaging<27.0,>=23.0
 Requires-Dist: pandas<3.0.0,>=2.0.0
 Requires-Dist: pydantic<3.0.0,>=2.0.0
 Requires-Dist: PyYAML<7.0.0,>=6.0.0
@@ -67,7 +68,7 @@ Requires-Dist: aiohttp<4.0.0,>=3.14.1; extra == "nightscout"
 Requires-Dist: py-nightscout; extra == "nightscout"
 Provides-Extra: research
 Requires-Dist: torch<3.0.0,>=2.13.0; extra == "research"
-Requires-Dist: pyarrow<25.0.0,>=12.0.0; extra == "research"
+Requires-Dist: pyarrow<26.0.0,>=12.0.0; extra == "research"
 Requires-Dist: h5py<4.0.0,>=3.10.0; extra == "research"
 Requires-Dist: onnx<2.0.0,>=1.22.0; extra == "research"
 Requires-Dist: onnxscript<1.0.0,>=0.1.0; extra == "research"
@@ -77,7 +78,7 @@ Requires-Dist: libroadrunner<3.0.0,>=2.9.2; extra == "mechanistic"
 Provides-Extra: fmi
 Requires-Dist: FMPy<0.4.0,>=0.3.30; extra == "fmi"
 Provides-Extra: mdmp
-Requires-Dist: cryptography<50.0.0,>=48.0.1; extra == "mdmp"
+Requires-Dist: cryptography<51.0.0,>=50.0.0; extra == "mdmp"
 Provides-Extra: desktop
 Requires-Dist: pyinstaller<7.0.0,>=6.11.0; extra == "desktop"
 Requires-Dist: setuptools<84.0.0,>=83.0.0; extra == "desktop"
@@ -104,7 +105,7 @@ Requires-Dist: onnxscript<1.0.0,>=0.1.0; extra == "desktop-all"
 Requires-Dist: openpyxl<4.0.0,>=3.0.0; extra == "desktop-all"
 Requires-Dist: pillow<13.0.0,>=12.3.0; extra == "desktop-all"
 Requires-Dist: plotly<7.0.0,>=5.18.0; extra == "desktop-all"
-Requires-Dist: pyarrow<25.0.0,>=12.0.0; extra == "desktop-all"
+Requires-Dist: pyarrow<26.0.0,>=12.0.0; extra == "desktop-all"
 Requires-Dist: pyinstaller<7.0.0,>=6.11.0; extra == "desktop-all"
 Requires-Dist: pyobjc-framework-Cocoa<13.0,>=10.0; platform_system == "Darwin" and extra == "desktop-all"
 Requires-Dist: pyserial<4.0,>=3.5; extra == "desktop-all"
@@ -121,57 +122,37 @@ Dynamic: license-file
 [![CI](https://github.com/python35/IINTS-SDK/actions/workflows/python-package.yml/badge.svg)](https://github.com/python35/IINTS-SDK/actions/workflows/python-package.yml)
 [![Docs](https://img.shields.io/badge/docs-IINTS--AF-0a66c2?style=flat)](https://python35.github.io/IINTS-SDK/)
 
-IINTS-AF is an open-source research SDK for diabetes technology simulation.
+IINTS-AF is an open-source research SDK for diabetes-technology simulation. It provides virtual-patient scenarios, algorithm experiments, deterministic safety checks, data-assurance tools, glucose-forecasting research, and reproducible reports.
 
-The SDK lets researchers, students, and developers test diabetes-algorithm ideas in a virtual environment. It can simulate digital patients, meals, insulin delivery, CGM/sensor behavior, safety checks, glucose-prediction experiments, and generated reports.
-
-The purpose of IINTS-AF is to make algorithm behavior easier to inspect and discuss before anything is connected to a real person.
+Its purpose is to make diabetes-technology behavior easier to inspect, test, and discuss before anything is connected to a real person.
 
 IINTS-AF is not a medical device. It must not be used for diagnosis, insulin dosing, treatment decisions, or real-time patient care.
 
-## Links
-
-- Website: [iints.org](https://iints.org)
-- Documentation: [python35.github.io/IINTS-SDK](https://python35.github.io/IINTS-SDK/)
-- Rust research workbench: [latest beta](https://github.com/python35/IINTS-SDK/releases/tag/tauri-beta-latest)
-- Classic desktop app: [latest Qt beta](https://github.com/python35/IINTS-SDK/releases/tag/desktop-beta-latest)
+## Start
 
-## Rust Research Workbench
+```bash
+python -m pip install -U "iints-sdk-python35[full,mdmp,research,edge]"
+iints demo
+iints menu
+iints version --refresh
+```
 
-The new native research workbench uses a Rust/Tauri security boundary around the Python scientific engine. It provides guided simulation workflows, result inspection, reproducibility records, local Ollama review, MDMP evidence tools, and audited research connectors.
+## Desktop App
 
-Current Rust/Tauri beta downloads:
+The optional Rust/Tauri research workbench offers guided runs, result inspection, local AI review, and data-assurance tools.
 
-| Platform | Download |
+| Platform | Latest beta |
 | --- | --- |
 | Windows | [`.exe` installer](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-windows-x64-setup.exe) |
 | macOS | [`.dmg`](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-macos.dmg) |
 | Linux | [`.AppImage`](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-linux-x64.AppImage) |
 
-The beta delegates scientific operations to the IINTS-AF Python engine. On first launch, choose
-**Install Python engine** if the bridge is unavailable; the app creates a private engine and keeps
-its setup visible in a terminal.
-
-See the [Tauri workbench documentation](https://python35.github.io/IINTS-SDK/TAURI_DESKTOP/) for architecture, security boundaries, and development instructions.
-
-## Classic Desktop App
-
-IINTS-AF also has a native Qt desktop app for running demos, reviewing results, certifying data, and using the research workbench without memorising terminal commands. The packaged Windows, macOS, and Linux betas include their Python runtime and supported Python-side research engines.
-
-Current beta downloads:
-
-| Platform | Download |
-| --- | --- |
-| Windows | [`.exe`](https://github.com/python35/IINTS-SDK/releases/download/desktop-beta-latest/IINTS-AF-Desktop-Beta-windows-x64.exe) |
-| macOS | [`.dmg`](https://github.com/python35/IINTS-SDK/releases/download/desktop-beta-latest/IINTS-AF-Desktop-Beta-macos.dmg) |
-| Linux | [executable](https://github.com/python35/IINTS-SDK/releases/download/desktop-beta-latest/IINTS-AF-Desktop-Beta-linux-x64) |
-
-Python install, including the PySide6 desktop runtime:
+## Links
 
-```bash
-python -m pip install -U "iints-sdk-python35[desktop-all]"
-iints-desktop
-```
+- Website: [iints.org](https://iints.org)
+- Documentation: [python35.github.io/IINTS-SDK](https://python35.github.io/IINTS-SDK/)
+- Source: [github.com/python35/IINTS-SDK](https://github.com/python35/IINTS-SDK)
+- Desktop guide: [Tauri research workbench](https://python35.github.io/IINTS-SDK/TAURI_DESKTOP/)
 
 ## License
 

iints_sdk_python35-1.5.34/README.md

@@ -0,0 +1,42 @@
+# IINTS-AF SDK
+
+[![EUCYS 2026](https://img.shields.io/badge/EUCYS-2026%20Selected-gold?style=flat)](https://www.uni-kiel.de/en/eucys2026)
+[![PyPI version](https://badge.fury.io/py/iints-sdk-python35.svg)](https://badge.fury.io/py/iints-sdk-python35)
+[![CI](https://github.com/python35/IINTS-SDK/actions/workflows/python-package.yml/badge.svg)](https://github.com/python35/IINTS-SDK/actions/workflows/python-package.yml)
+[![Docs](https://img.shields.io/badge/docs-IINTS--AF-0a66c2?style=flat)](https://python35.github.io/IINTS-SDK/)
+
+IINTS-AF is an open-source research SDK for diabetes-technology simulation. It provides virtual-patient scenarios, algorithm experiments, deterministic safety checks, data-assurance tools, glucose-forecasting research, and reproducible reports.
+
+Its purpose is to make diabetes-technology behavior easier to inspect, test, and discuss before anything is connected to a real person.
+
+IINTS-AF is not a medical device. It must not be used for diagnosis, insulin dosing, treatment decisions, or real-time patient care.
+
+## Start
+
+```bash
+python -m pip install -U "iints-sdk-python35[full,mdmp,research,edge]"
+iints demo
+iints menu
+iints version --refresh
+```
+
+## Desktop App
+
+The optional Rust/Tauri research workbench offers guided runs, result inspection, local AI review, and data-assurance tools.
+
+| Platform | Latest beta |
+| --- | --- |
+| Windows | [`.exe` installer](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-windows-x64-setup.exe) |
+| macOS | [`.dmg`](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-macos.dmg) |
+| Linux | [`.AppImage`](https://github.com/python35/IINTS-SDK/releases/download/tauri-beta-latest/IINTS-AF-Research-Workbench-linux-x64.AppImage) |
+
+## Links
+
+- Website: [iints.org](https://iints.org)
+- Documentation: [python35.github.io/IINTS-SDK](https://python35.github.io/IINTS-SDK/)
+- Source: [github.com/python35/IINTS-SDK](https://github.com/python35/IINTS-SDK)
+- Desktop guide: [Tauri research workbench](https://python35.github.io/IINTS-SDK/TAURI_DESKTOP/)
+
+## License
+
+Apache-2.0 licensed, with legacy MIT notices where applicable.

iints_sdk_python35-1.5.34/docs/ALL_SCIENTIFIC_THEORIES.md

@@ -0,0 +1,146 @@
+# Scientific Mechanisms And Their Limits
+
+IINTS-AF is a **research and educational simulator**, not a clinically validated
+patient model or medical device. This page explains what the deterministic code
+implements and, equally importantly, what each mechanism does **not** prove.
+
+The authoritative equation list is the [Formula Registry](FORMULA_REGISTRY.md).
+Every entry is labelled as:
+
+- `canonical`: a direct published equation;
+- `adapted`: a published model modified for SDK integration;
+- `heuristic`: an explicit, testable research assumption.
+
+Published equations do not automatically validate the combined implementation,
+its parameters, or its use for an individual person.
+
+## 1. Bergman-Style Glucose Balance (`F01`)
+
+The Bergman mode uses a concentration-domain glucose balance with a remote
+insulin-action state. Meal appearance, renal loss, exercise, stress, dawn and
+glucagon terms are IINTS extensions. It is therefore **Bergman-style**, not an
+unmodified implementation of the 1979 protocol model.
+
+## 2. Remote Insulin Action (`F02`)
+
+The action state follows insulin above or below a basal reference:
+
+$$
+\frac{dX}{dt}=-p_2X+p_{3,\mathrm{eff}}(I-I_{\mathrm{ref}}).
+$$
+
+Here, \(I_{\mathrm{ref}}\) is the pump-supported fasting concentration derived
+from basal delivery, distribution volume and clearance. Allowing a negative
+deviation matters after interrupted basal delivery; clipping the deviation at
+zero would hide the loss of basal insulin action.
+
+## 3. Plasma Insulin And Optional Graft Secretion (`F03`)
+
+Plasma insulin is cleared by a first-order term and receives absorbed
+subcutaneous insulin. Optional beta-cell or graft secretion is an **experimental
+heuristic**, disabled in standard T1D profiles. It is not a transplantation
+outcome predictor.
+
+## 4. Subcutaneous Insulin PK (`F04`)
+
+Two serial depots delay pump delivery before plasma appearance. The structure is
+literature-based; insulin-type time constants are predefined research profiles,
+not product-specific bioequivalence claims.
+
+## 5. Meal Appearance (`F05`)
+
+Hovorka mode uses its published two-compartment meal chain. Bergman and advanced
+modes use an adapted stomach/gut chain. The latter must not be called an exact
+Dalla Man implementation: it does not reproduce the complete Dalla Man meal
+submodel or its nonlinear gastric-emptying function.
+
+## 6. Hovorka Glucose Mass Balance (`F06`)
+
+Accessible and non-accessible glucose are represented as mass compartments.
+The base balance follows Hovorka structure, while stress, exercise, circadian,
+renal, HAAF and glucagon terms are declared extensions. Mass is converted to
+concentration only through the configured glucose distribution volume.
+
+## 7. Hovorka Insulin-Action Channels (`F07`)
+
+Three action channels affect distribution, disposal and endogenous glucose
+production. Tissue and molecular-affinity scalars are scenario stressors, not
+quantitative consequences inferred from AlphaFold, GTEx or ClinVar. Structural
+confidence and gene-expression evidence remain contextual evidence only.
+
+## 8. Stress And Exercise States (`F08`)
+
+Stress and exercise are filtered scenario inputs that alter sensitivity and
+glucose fluxes. They are not measured cortisol, adrenaline, AMPK or lactate
+concentrations. Their coefficients require empirical calibration before a
+specific population claim can be made.
+
+## 9. Exercise/GLUT4 Abstraction (`F09`)
+
+The bounded `GLUT4_active` state adds insulin-independent glucose uptake during
+exercise. It captures a plausible direction of effect but is not a molecular
+translocation model and cannot be validated by a protein structure image alone.
+
+## 10. Circadian/Dawn Term (`F10`)
+
+A gated Fourier profile can perturb endogenous glucose production. It is off or
+weak in baseline profiles and is a configurable circadian stressor, not a model
+of measured growth hormone or cortisol secretion.
+
+## 11. Counter-Regulatory Rescue (`F11`)
+
+Low glucose activates a bounded increase in endogenous glucose production. The
+response is reduced by the antecedent-hypoglycaemia memory state. This tests the
+direction of counter-regulation; it does not estimate a patient's glucagon or
+epinephrine response.
+
+## 12. Antecedent-Hypoglycaemia Memory (`F12`)
+
+The HAAF-like state accumulates gradually during hypoglycaemia and recovers over
+days. It is deliberately bounded and labelled heuristic. It must never be
+reported as a diagnosis of impaired awareness or HAAF.
+
+## 13. Exogenous Glucagon PK/PD (`F13`)
+
+The two-state absorption and clearance equations use literature-informed ranges,
+with exact mass conversion (`1 mg = 10^9 pg`). The bounded concentration-effect
+coupling is an IINTS adaptation and has not been established as a dose-selection
+model for patient care.
+
+## 14. Renal Glucose Loss (`F14`)
+
+A smooth softplus approximation replaces a discontinuous renal threshold. This
+is numerically useful and represents threshold/splay behaviour qualitatively,
+but renal threshold varies between and within people. It is not an eGFR or renal
+disease model.
+
+## 15. CGM Observation Model (`F15`)
+
+Latent blood glucose passes through an explicit dead time and a first-order
+blood-to-interstitial compartment before bias, drift, dropout and seeded noise
+are applied. This separates physiology from measurement. It is a generic CGM
+observation model, not a Dexcom-, Libre- or Medtronic-equivalent sensor model.
+
+## Additional Advanced Stressors
+
+The advanced model also exposes FFA, ketone, protein, fat, illness, menstrual
+cycle, beta-cell and cannula-age states. These are hypothesis-generating stress
+tests. In particular:
+
+- ketone state is not a DKA diagnosis;
+- menstrual phase is not inferred from hormone measurements;
+- cannula ageing is not a device-specific failure probability;
+- protein-to-glucose conversion is a simplified delayed flux;
+- beta-cell/graft states are not transplant efficacy predictions.
+
+## AI, Calibration And Validation Boundary
+
+LLMs may summarize deterministic outputs but never calculate the ODEs, safety
+limits or formula values. The glucose predictor's physiology-aware loss is a
+regularizer, not a full ODE-residual PINN. Parameter calibration estimates a
+small bounded profile from CGM and event data; those parameters are generally
+not uniquely identifiable and require held-out and external validation.
+
+Scientific use therefore requires reporting the model version, parameter set,
+seed, input semantics, evidence class, numerical checks, held-out performance,
+failure cases and limitations with every result.