iints-sdk-python35 1.5.4

A pre-clinical Edge-AI SDK for diabetes management validation.

IINTS-AF SDK

"Code shouldn't be a secret when it's managing a life."

Insulin pumps make hundreds of autonomous decisions about drug delivery every day.
The algorithms behind those decisions are proprietary, unauditable, and difficult to inspect or improve, even by the patients whose lives depend on them.

IINTS-AF is an open-source research platform that changes that.

Core Research Question

Can open-source simulation and deterministic safety supervision make insulin delivery algorithm development safer and more transparent for researchers and patients?

---

What It Does

Simulate: Run virtual patients through thousands of scenarios before any algorithm reaches a real device. A deterministic safety supervisor audits every AI decision. The AI may suggest. The supervisor decides.

Certify: Every dataset is fingerprinted and graded before it touches a study workflow. The goal is to keep benchmark inputs traceable, reviewable, and reproducible.

Understand: Generate audit-ready reports, visual posters, and local AI summaries from the same study bundle. IINTS-AF can use local models such as Ministral for explanation workflows on your own hardware.

---

Research Results

Final locked benchmark: 3600 simulation runs, 6 profiles, 4 scenario families, 5 algorithms, 10 fixed seeds.

Metric	ExampleAlgorithm	PID Baseline	Delta
Time in Range	87.16%	83.72%	+3.44%
Time < 70 mg/dL	1.28%	5.25%	-3.97%
Supervisor interventions	99	177	-78

Additional benchmark result:

• clean certified conditions showed +17.64 Time-in-Range points versus corrupted uncertified conditions

For the full scientific write-up, see:

• research/EUCYS_REPORT.md
• research/EUCYS_REPORT.pdf
• results/eucys_2026/EUCYS_RESULTS/EUCYS_SUMMARY.md

---

Install

python3 -m venv .venv && source .venv/bin/activate
pip install -U "iints-sdk-python35[full,mdmp]"
iints doctor --smoke-run

Edge devices (Raspberry Pi 5, Arduino UNO Q):

pip install -U "iints-sdk-python35[edge,mdmp]"

---

Quick Start

iints quickstart --project-name my_study
cd my_study
iints presets run --name baseline_t1d --algo algorithms/example_algorithm.py
iints data certify contracts/clinical_mdmp_contract.yaml data/demo/diabetes_cgm.csv --output-json audit/certification.json
iints ai report results/<run_id>

---

Final Benchmark Workflow

tools/research/run_eucys_final.sh \
  --algo algorithms/example_algorithm.py \
  --output-dir results/eucys_2026 \
  --seeds 1,2,3,4,5,6,7,8,9,10

Then render the report:

tools/research/render_eucys_report_pdf.sh

Main final artifacts:

• results/eucys_2026/
• results/eucys_2026/EUCYS_RESULTS/EUCYS_MAIN_FIGURE.png
• results/eucys_2026/EUCYS_RESULTS/EUCYS_RESULTS_TABLE.csv
• research/EUCYS_REPORT.md
• research/EUCYS_REPORT.pdf

---

Live Digital Patient (Raspberry Pi)

iints patient start \
  --algo algorithms/example_algorithm.py \
  --scenario-profile expo_hot_start \
  --mode demo-time --speed 60x

Open http://127.0.0.1:8765/dashboard. You will see a virtual patient running continuously, reacting to meals, exercise, and sleep in real time.

---

Documentation

Getting started	Open guide
Edge hardware	Open guide
Raspberry Pi setup	Open guide
Data certification	Open guide
Full manual	Open PDF manual
Research report	Open report source

---

IINTS-AF is research software. Not a medical device.
No clinical dosing advice. MIT Licensed.

---

Built by a 17-year-old with type 1 diabetes who wanted to understand the device managing his life.