IINTS-AF SDK¶
IINTS-AF is an open-source research SDK for building reproducible diabetes-technology experiments. It combines virtual patients, glucose and insulin scenarios, candidate algorithms, deterministic safety checks, data-quality tools, and research reports in one workflow.
It is designed for simulation, education, benchmarking, and pre-clinical software research.
Research boundary
IINTS-AF is not a medical device. It must not be used for insulin dosing, diagnosis, treatment decisions, or real-time patient care.
Start In The Right Place¶
| Your goal | Start here | You will learn to |
|---|---|---|
| Learn the SDK from the beginning | Learning Path | install, run, inspect, and validate an experiment |
| Get one result quickly | First Run | verify the installation and create a demo bundle |
| Use the desktop interface | Desktop App | run workflows and inspect outputs without memorising CLI commands |
| Work with data or local AI | Workflow Hub | choose the correct data, AI, study, or reporting route |
| Understand the scientific assumptions | Scientific Workflow | separate implementation, evidence, calibration, and limitations |
| Contribute code | Developer Portal | navigate the architecture and run the required checks |
How The SDK Works¶
flowchart LR
A["Patient and scenario"] --> B["Simulation engine"]
B --> C["Candidate algorithm"]
C --> D["Deterministic safety checks"]
D --> E["Run bundle"]
E --> F["Validation, reports, and optional AI review"]
The important separation is:
- The simulator computes the virtual physiological state.
- The candidate algorithm proposes an experimental action.
- The safety layer applies fixed, reviewable limits.
- The evidence layer records inputs, outputs, versions, and checks.
- The optional AI layer explains validated results; it is not the source of numerical truth.
Read Core Concepts for the vocabulary used throughout the documentation.
First Installation¶
IINTS-AF supports Python 3.10 through 3.14.
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install "iints-sdk-python35[full,mdmp]"
iints doctor --smoke-run
iints demo quick --output-dir results/first_run
On Windows PowerShell, activate the environment with:
.venv\Scripts\Activate.ps1
See Installation for source installs, optional research dependencies, and platform-specific help.
What You Can Build¶
- deterministic virtual-patient simulations
- repeatable scenario and algorithm comparisons
- glucose-data quality and MDMP certification artifacts
- AGP-style research reports, study summaries, and evidence bundles
- local Ollama-assisted explanations of completed runs
- glucose forecasting experiments with explicit evaluation gates
- bench-only Raspberry Pi, Pico, UNO Q, Jetson, and FPGA workflows
- interactive structural-biology and genomics research demonstrations
Each advanced feature has its own limitations. The documentation distinguishes between implemented behavior, scientific inspiration, empirical calibration, and clinical validation. Those terms are not interchangeable.
A Good First Session¶
- Complete the First Run.
- Read Core Concepts.
- Learn how to Understand A Run.
- Complete the First Workflow.
- Choose a specialised route from the Workflow Hub.
For the project website, visit iints.org. For source code and issues, use GitHub.