iints-sdk-python35 1.5.32__tar.gz → 1.5.33__tar.gz

{iints_sdk_python35-1.5.32 → iints_sdk_python35-1.5.33}/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: "1.5.33"
+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_python35-1.5.33}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: iints-sdk-python35
-Version: 1.5.32
+Version: 1.5.33
 Summary: A pre-clinical Edge-AI SDK for diabetes management validation.
 Author-email: Rune Bobbaers <rune.bobbaers@gmail.com>
 License-Expression: Apache-2.0
@@ -22,7 +22,7 @@ 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: numpy<3.0.0,>=1.24.0
@@ -77,7 +77,7 @@ 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"

iints_sdk_python35-1.5.33/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.

{iints_sdk_python35-1.5.32 → iints_sdk_python35-1.5.33}/docs/API_REFERENCE.md

@@ -636,7 +636,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 ### Public Functions
 
 - `compute_metrics(results_df: pd.DataFrame) -> Dict[str, float]`
-- `run_baseline_comparison(patient_params: Dict[str, Any], stress_event_payloads: List[Dict[str, Any]], duration: int, time_step: int, primary_label: str, primary_results: pd.DataFrame, primary_safety: Dict[str, Any], compare_standard_pump: bool = True, seed: Optional[int] = None) -> Dict[str, Any]`
+- `run_baseline_comparison(patient_params: Dict[str, Any], stress_event_payloads: List[Dict[str, Any]], duration: int, time_step: int, primary_label: str, primary_results: pd.DataFrame, primary_safety: Dict[str, Any], compare_standard_pump: bool = True, seed: Optional[int] = None, patient_model_type: str = 'auto') -> Dict[str, Any]`
 - `write_baseline_comparison(comparison: Dict[str, Any], output_dir: Path) -> Dict[str, str]`
 
 ## `iints.analysis.booth_demo`
@@ -749,13 +749,13 @@ No public classes, functions, or all-caps constants are declared directly in thi
 ## `iints.analysis.edge_performance_monitor`
 
 - Source: `src/iints/analysis/edge_performance_monitor.py`
-- Summary: Edge AI Performance Monitor - IINTS-AF Jetson Nano performance validation for medical device standards
+- Summary: Edge AI Performance Monitor - IINTS-AF Jetson Nano research benchmarking for latency and resource use
 
 ### Public Classes
 
 | Class | Signature | Summary |
 | --- | --- | --- |
-| `EdgeAIPerformanceMonitor` | `EdgeAIPerformanceMonitor` | Monitor Jetson Nano performance for medical device validation |
+| `EdgeAIPerformanceMonitor` | `EdgeAIPerformanceMonitor` | Monitor edge performance; results are not medical-device validation. |
 
 #### `EdgeAIPerformanceMonitor` methods
 
@@ -1423,7 +1423,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 - `research_glucose_model_init(output_dir: Annotated[Path, typer.Option(help='Output directory for the glucose model starter files.')] = Path('models/iints-glucose-forecast-v0'), profile: Annotated[str, typer.Option(help='Training profile: smoke, quick, long, or paper.')] = 'long', history_minutes: Annotated[int, typer.Option(help='Model history window in minutes.')] = 360, horizon_minutes: Annotated[int, typer.Option(help='Prediction horizon in minutes.')] = 120, time_step_minutes: Annotated[int, typer.Option(help='Expected CGM sample interval in minutes.')] = 5) -> None`
 - `research_glucose_model_train(data: Annotated[Path, typer.Option(help='Normalized glucose training dataset CSV/Parquet.')], output_dir: Annotated[Path, typer.Option(help='Output directory for predictor.pt and training_report.json.')] = Path('models/iints-glucose-forecast-v0'), config: Annotated[Optional[Path], typer.Option(help='Config YAML. If omitted, one is generated.')] = None, profile: Annotated[str, typer.Option(help='Generated config profile: smoke, quick, long, or paper.')] = 'long', epochs: Annotated[Optional[int], typer.Option(help='Override training.epochs for long local runs.')] = None, batch_size: Annotated[Optional[int], typer.Option(help='Override training.batch_size.')] = None, learning_rate: Annotated[Optional[float], typer.Option(help='Override training.learning_rate.')] = None, warm_start: Annotated[Optional[Path], typer.Option(help='Optional predictor.pt warm-start checkpoint.')] = None, export_hf: Annotated[bool, typer.Option('--export-hf/--no-export-hf', help='Build a Hugging Face-ready folder after training.')] = True, repo_id: Annotated[Optional[str], typer.Option(help='Optional Hugging Face repo id for model card hints.')] = None, dataset_manifest: Annotated[Optional[Path], typer.Option(help='Optional dataset manifest for HF public metadata.')] = None, comparison_dir: Annotated[Optional[Path], typer.Option(help='Optional glucose-model compare output directory to bundle with the Hugging Face export.')] = None) -> None`
 - `research_glucose_model_export_hf(model_dir: Annotated[Path, typer.Option(help='Directory containing predictor.pt and training_report.json.')] = Path('models/iints-glucose-forecast-v0'), output_dir: Annotated[Path, typer.Option(help='Output directory for the Hugging Face-ready bundle.')] = Path('models/iints-glucose-forecast-v0/huggingface'), repo_id: Annotated[Optional[str], typer.Option(help='Optional Hugging Face repo id, e.g. user/iints-glucose-forecast-v0.')] = None, dataset_manifest: Annotated[Optional[Path], typer.Option(help='Optional private manifest to redact into public metadata.')] = None, comparison_dir: Annotated[Optional[Path], typer.Option(help='Optional glucose-model compare output directory to include comparison metrics and reports.')] = None) -> None`
-- `research_glucose_model_compare(data: Annotated[Path, typer.Option(help='Normalized glucose dataset CSV/Parquet to evaluate on.')], output_dir: Annotated[Path, typer.Option(help='Output directory for comparison reports.')] = Path('results/glucose_model_comparison'), model_specs: Annotated[Optional[List[str]], typer.Option('--model', '-m', help='Model checkpoint as label=path/to/predictor.pt. Repeat for MSE/Band/PINN models.')] = None, config: Annotated[Optional[Path], typer.Option(help='Comparison config YAML. Defaults to glucose-model quick config.')] = None, include_baselines: Annotated[bool, typer.Option('--include-baselines/--no-baselines', help='Compare transparent LastValue/LinearTrend/Physiology baselines.')] = True, mc_samples: Annotated[int, typer.Option(help='MC dropout samples for checkpoint uncertainty. Use 0 to disable.')] = 0, max_roc_mgdl_min: Annotated[float, typer.Option(help='Maximum plausible predicted glucose rate-of-change in mg/dL/min.')] = 10.0) -> None`
+- `research_glucose_model_compare(data: Annotated[Path, typer.Option(help='Normalized glucose dataset CSV/Parquet to evaluate on.')], output_dir: Annotated[Path, typer.Option(help='Output directory for comparison reports.')] = Path('results/glucose_model_comparison'), model_specs: Annotated[Optional[List[str]], typer.Option('--model', '-m', help='Model checkpoint as label=path/to/predictor.pt. Repeat for MSE, band-weighted, or physiology-regularized models (legacy name: PINN).')] = None, config: Annotated[Optional[Path], typer.Option(help='Comparison config YAML. Defaults to glucose-model quick config.')] = None, include_baselines: Annotated[bool, typer.Option('--include-baselines/--no-baselines', help='Compare transparent LastValue/LinearTrend/Physiology baselines.')] = True, mc_samples: Annotated[int, typer.Option(help='MC dropout samples for checkpoint uncertainty. Use 0 to disable.')] = 0, max_roc_mgdl_min: Annotated[float, typer.Option(help='Maximum plausible predicted glucose rate-of-change in mg/dL/min.')] = 3.0) -> None`
 - `research_glucose_model_jetson_train_hf(dataset: Annotated[Path, typer.Option(help='Normalized glucose training dataset CSV/Parquet built by glucose-model build-dataset.')] = Path('models/iints-glucose-forecast-v0/dataset/glucose_training_dataset.csv'), base_hf_repo: Annotated[Optional[str], typer.Option('--base-hf-repo', '--repo-id', help='External Hugging Face model to pull from (e.g. username/GlucoFM). If empty, pulls from target_hf_repo. --repo-id is kept as a compatibility alias.')] = None, target_hf_repo: Annotated[Optional[str], typer.Option('--target-hf-repo', help='Your Hugging Face model repo id to push to, e.g. username/iints-glucose-forecast-v0.')] = 'IINTS/iints-glucose-forecast-v0', local_base_dir: Annotated[Optional[Path], typer.Option(help='Use an already downloaded base model folder instead of downloading from Hugging Face.')] = None, work_dir: Annotated[Path, typer.Option(help='Jetson training workspace for downloads, trials, champion, and leaderboard.')] = Path('models/jetson_hf_training'), revision: Annotated[Optional[str], typer.Option(help='Optional Hugging Face revision/tag/branch to download.')] = None, profile: Annotated[str, typer.Option(help='Fallback config profile when the HF repo has no glucose_model_config.yaml.')] = 'quick', max_trials: Annotated[int, typer.Option(help='Number of trials. Use 0 to keep training until Ctrl+C.')] = 1, epochs: Annotated[int, typer.Option(help='Fine-tune epochs per trial.')] = 8, batch_size: Annotated[int, typer.Option(help='Batch size per trial; keep modest on Jetson Nano.')] = 64, timeout_minutes: Annotated[float, typer.Option(help='Timeout for each train/compare subprocess.')] = 45.0, cooldown_seconds: Annotated[float, typer.Option(help='Pause between trials to keep Jetson thermals stable.')] = 10.0, min_lr: Annotated[float, typer.Option(help='Minimum sampled learning rate for fine-tuning.')] = 1e-05, max_lr: Annotated[float, typer.Option(help='Maximum sampled learning rate for fine-tuning.')] = 0.0005, min_pinn_lambda: Annotated[float, typer.Option(help='Minimum sampled PINN loss weight.')] = 0.05, max_pinn_lambda: Annotated[float, typer.Option(help='Maximum sampled PINN loss weight.')] = 0.8, min_score_improvement: Annotated[float, typer.Option(help='Required composite-score improvement before replacing the local champion.')] = 0.0, physiology_weight: Annotated[float, typer.Option(help='Composite score penalty weight for physiological violations.')] = 0.1, hypo_weight: Annotated[float, typer.Option(help='Composite score penalty weight for missed/false hypo behavior.')] = 0.2, seed: Annotated[int, typer.Option(help='Random seed for reproducible trial configs.')] = 42, dataset_manifest: Annotated[Optional[Path], typer.Option(help='Optional dataset manifest to redact into the HF export bundle.')] = None, upload_mode: Annotated[str, typer.Option(help='Upload behavior: none, pr, or direct. Default is safe local-only.')] = 'none', private_upload: Annotated[bool, typer.Option('--private-upload/--public-upload', help='Mark HF upload private when upload is enabled.')] = True, force_download: Annotated[bool, typer.Option('--force-download/--reuse-download', help='Re-download the HF base model even if cached locally.')] = False, hf_home: Annotated[Optional[Path], typer.Option(help='Optional HF_HOME cache directory. Defaults inside the Jetson work dir.')] = None) -> None`
 - `research_parity_check(model: Annotated[Path, typer.Option(help='Predictor checkpoint (.pt)')], onnx: Annotated[Path, typer.Option(help='Exported ONNX model path')], samples: Annotated[int, typer.Option(help='Random sample count for parity check')] = 64, tolerance: Annotated[float, typer.Option(help='Maximum allowed absolute error')] = 0.001, seed: Annotated[int, typer.Option(help='Random seed')] = 42, output_json: Annotated[Optional[Path], typer.Option(help='Write parity report JSON')] = None)`
 - `research_registry_list(registry: Annotated[Path, typer.Option(help='Path to model registry JSON')] = Path('models/registry.json'), stage: Annotated[Optional[str], typer.Option(help='Optional stage filter (candidate/validated/production/archived)')] = None, limit: Annotated[int, typer.Option(help='Max rows to print')] = 30)`
@@ -1698,7 +1698,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 | Class | Signature | Summary |
 | --- | --- | --- |
-| `MPCController` | `MPCController(InsulinAlgorithm)` | Agentic Physics-Informed MPC. Simulates the biological ODEs into the future to find the safest dose. |
+| `MPCController` | `MPCController(InsulinAlgorithm)` | Research nonlinear MPC prototype using an adapted Bergman model. |
 
 #### `MPCController` methods
 
@@ -1864,7 +1864,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 | Class | Signature | Summary |
 | --- | --- | --- |
-| `DigitalTwinCalibrator` | `DigitalTwinCalibrator` | AI-driven Calibration Engine for creating a personalized Digital Twin. It takes historical data (CGM, carbs, insulin) and uses scipy optimization (L-BFGS-B) to find the metabolic "Big 5" parameters that minimize the RMSE between the simulation and real patient data. |
+| `DigitalTwinCalibrator` | `DigitalTwinCalibrator` | Research parameter-calibration engine for a Hovorka simulation profile. |
 
 #### `DigitalTwinCalibrator` methods
 
@@ -1934,7 +1934,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 ## `iints.core.patient.bergman_model`
 
 - Source: `src/iints/core/patient/bergman_model.py`
-- Summary: Bergman Minimal Model — IINTS-AF ================================== ODE-based patient model inspired by the Bergman Minimal Model with an additional gut absorption compartment for realistic carbohydrate dynamics.
+- Summary: Bergman Minimal Model — IINTS-AF ================================== ODE-based patient model inspired by the Bergman Minimal Model with an adapted gut absorption chain for delayed carbohydrate appearance.
 
 ### Public Classes
 
@@ -1962,7 +1962,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 ## `iints.core.patient.hovorka_model`
 
 - Source: `src/iints/core/patient/hovorka_model.py`
-- Summary: Improved Hovorka Model - IINTS-AF ================================== Based on standard Hovorka artificial pancreas equations and extended to match the IINTS simulator's interface.
+- Summary: Adapted Hovorka Research Model - IINTS-AF ========================================== Based on published Hovorka artificial-pancreas equations and extended with explicit research stressors to match the IINTS simulator interface. The extensions are not part of the canonical Hovorka model and are not clinically validated patient physiology.
 
 ### Public Classes
 
@@ -1996,6 +1996,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 | Class | Signature | Summary |
 | --- | --- | --- |
+| `PatientModelDomainError` | `PatientModelDomainError(RuntimeError)` | The model left its declared numerical/physiological validity domain. |
 | `CustomPatientModel` | `CustomPatientModel` | A simplified patient model for simulating blood glucose dynamics. This model is intended for educational and stress-testing purposes, not for clinical accuracy. |
 
 #### `CustomPatientModel` methods
@@ -2040,6 +2041,8 @@ No public classes, functions, or all-caps constants are declared directly in thi
 - `carbs_on_board(self)`
 - `trigger_event(self, event_type, value)`
 - `get_patient_state(self)`
+- `get_ratio_state(self) -> Dict[str, float]`
+- `set_ratio_state(self, isf: Optional[float] = None, icr: Optional[float] = None, basal_rate: Optional[float] = None, dia_minutes: Optional[float] = None) -> None`
 
 ## `iints.core.patient.physiology`
 
@@ -2048,9 +2051,20 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 ### Public Functions
 
+- `validated_snapshot_scalar(value: Any, *, name: str, minimum: float | None = None, maximum: float | None = None) -> float`
+- `validated_snapshot_bool(value: Any, *, name: str) -> bool`
+- `validated_activity_events(value: Any, *, name: str, age_key: str) -> list[dict[str, float]]`
+- `glucagon_mg_to_pg(dose_mg: float) -> float`
+- `dawn_glucose_rate_mgdl_min(current_time_minutes: float, *, peak_strength_mgdl_per_hour: float, start_hour: float, end_hour: float) -> float`
+- `antecedent_hypoglycemia_memory_derivative(glucose_mgdl: float, memory: float, *, awareness_threshold_mgdl: float = 70.0, severe_threshold_mgdl: float = 54.0, build_time_constant_minutes: float = 360.0, recovery_time_constant_minutes: float = 4320.0) -> float`
+- `counterregulatory_rescue_multiplier(glucose_mgdl: float, memory: float, *, threshold_mgdl: float = 70.0, half_activation_mgdl: float = 16.0, maximum_fractional_increase: float = 1.0) -> float`
 - `smooth_threshold_excess(value: float, *, threshold: float, splay: float = 10.0) -> float`
 - `renal_glucose_clearance_concentration(glucose_mgdl: float, *, threshold_mgdl: float = 180.0, gain: float = 0.05, splay_mgdl: float = 10.0) -> float`
 
+### Public Constants
+
+- `PICOGRAMS_PER_MILLIGRAM`
+
 ## `iints.core.patient.profile`
 
 - Source: `src/iints/core/patient/profile.py`
@@ -2235,7 +2249,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 #### `IndependentSupervisor` methods
 
-- `evaluate_safety(self, current_glucose: float, proposed_insulin: float, current_time: float, current_iob: float = 0.0, predicted_glucose_30min: Optional[float] = None, basal_insulin_units: Optional[float] = None, basal_limit_units: Optional[float] = None) -> Dict[str, Any]`
+- `evaluate_safety(self, current_glucose: float, proposed_insulin: float, current_time: float, current_iob: float = 0.0, predicted_glucose_30min: Optional[float] = None, basal_insulin_units: Optional[float] = None, basal_limit_units: Optional[float] = None, meal_bolus_units: Optional[float] = None) -> Dict[str, Any]`
 - `get_safety_report(self) -> Dict[str, Any]`
 - `reset(self) -> None`
 - `get_state(self) -> Dict[str, Any]`
@@ -3559,7 +3573,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 ## `iints.population.generator`
 
 - Source: `src/iints/population/generator.py`
-- Summary: Population Generator — IINTS-AF ================================ Generates a virtual population of N patients with physiological variation around a base patient profile. Each parameter is drawn from a configurable distribution (truncated normal or log-normal) whose bounds respect the clinically valid ranges defined in the SDK validation schemas.
+- Summary: Population Generator — IINTS-AF ================================ Generates a virtual population of N patients with physiological variation around a base patient profile. Each parameter is drawn from a configurable distribution (truncated normal or log-normal) whose bounds respect the configured research ranges defined in the SDK schemas. These bounds are not population-validation or clinical-validity claims.
 
 ### Public Classes
 
@@ -4051,7 +4065,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 ### Public Functions
 
-- `standardize_glucose_forecast_frame(df: pd.DataFrame, *, source_label: str, time_step_minutes: int = 5, subject_prefix: Optional[str] = None, max_gap_multiplier: float = 2.5) -> pd.DataFrame`
+- `standardize_glucose_forecast_frame(df: pd.DataFrame, *, source_label: str, time_step_minutes: int = 5, subject_prefix: Optional[str] = None, max_gap_multiplier: float = 2.5, glucose_bounds_mgdl: tuple[float, float] = (20.0, 600.0)) -> pd.DataFrame`
 - `glucose_model_config_payload(*, profile: str = 'long', history_minutes: int = 360, horizon_minutes: int = 120, time_step_minutes: int = 5, feature_columns: Optional[Sequence[str]] = None) -> dict[str, Any]`
 - `write_glucose_model_config(path: Path, **kwargs: Any) -> dict[str, Any]`
 - `build_glucose_training_pack(input_paths: Sequence[Path], output_dir: Path, *, labels: Optional[Sequence[str]] = None, output_format: str = 'csv', profile: str = 'long', history_minutes: int = 360, horizon_minutes: int = 120, time_step_minutes: int = 5) -> GlucoseTrainingPack`
@@ -4066,8 +4080,8 @@ No public classes, functions, or all-caps constants are declared directly in thi
 - `render_hf_publishing_notes(*, repo_id: Optional[str]) -> str`
 - `parse_model_specs(values: Sequence[str]) -> list[GlucoseModelSpec]`
 - `horizon_error_rows(*, label: str, observed: np.ndarray, predicted: np.ndarray, time_step_minutes: int) -> list[dict[str, Any]]`
-- `physiological_violation_report(X: np.ndarray, predicted: np.ndarray, *, feature_columns: Sequence[str], time_step_minutes: int, absolute_low_mgdl: float = 20.0, absolute_high_mgdl: float = 600.0, display_low_mgdl: float = 35.0, display_high_mgdl: float = 450.0, max_roc_mgdl_min: float = 10.0, suspicious_roc_mgdl_min: float = 2.0) -> dict[str, Any]`
-- `compare_glucose_models(*, data_path: Path, output_dir: Path, model_specs: Sequence[GlucoseModelSpec] = (), config_path: Optional[Path] = None, include_baselines: bool = True, mc_samples: int = 0, max_roc_mgdl_min: float = 10.0) -> GlucoseModelComparisonBundle`
+- `physiological_violation_report(X: np.ndarray, predicted: np.ndarray, *, feature_columns: Sequence[str], time_step_minutes: int, absolute_low_mgdl: float = 20.0, absolute_high_mgdl: float = 600.0, display_low_mgdl: float = 35.0, display_high_mgdl: float = 450.0, max_roc_mgdl_min: float = 3.0, suspicious_roc_mgdl_min: float = 2.0) -> dict[str, Any]`
+- `compare_glucose_models(*, data_path: Path, output_dir: Path, model_specs: Sequence[GlucoseModelSpec] = (), config_path: Optional[Path] = None, include_baselines: bool = True, mc_samples: int = 0, max_roc_mgdl_min: float = 3.0) -> GlucoseModelComparisonBundle`
 
 ### Public Constants
 
@@ -4603,7 +4617,7 @@ No public classes, functions, or all-caps constants are declared directly in thi
 
 | Class | Signature | Summary |
 | --- | --- | --- |
-| `DigitalTwinCalibrator` | `DigitalTwinCalibrator` | Fits the AdvancedMetabolicModel parameters (p1, p2, p3, Gb) to real patient data (like the OhioT1DM dataset) to create a true Digital Twin. |
+| `DigitalTwinCalibrator` | `DigitalTwinCalibrator` | Fit a limited AdvancedMetabolicModel parameter set to research data. |
 
 #### `DigitalTwinCalibrator` methods
 

{iints_sdk_python35-1.5.32 → iints_sdk_python35-1.5.33}/docs/EU_AI_PACT_GOVERNANCE.md

@@ -54,12 +54,21 @@ result = review_eu_ai_pact_readiness({
 print(result.status)
 ```
 
+`compliance_score` accepts either a fraction (`0.99`) or the historical
+percentage representation (`99`). The SDK normalizes both to a fraction and
+stores the raw value in the evidence output. Values outside these two declared
+scales are rejected rather than silently reinterpreted.
+
 ## Interpretation
 
 - `research_ready`: the evidence bundle is complete enough for internal research review.
 - `needs_review`: no hard blocker, but the bundle still needs explanation or cleanup.
 - `blocked`: do not use the bundle for public AI claims until the missing controls are resolved.
 
+These statuses describe completion of an IINTS self-assessment evidence
+bundle. They are not an official EU AI Act classification, legal compliance
+decision, notified-body assessment, CE marking, or medical-device approval.
+
 ## Official Sources
 
 - European Commission: [AI Pact](https://digital-strategy.ec.europa.eu/en/policies/ai-pact)