Edge Hardware Profiles¶
IINTS supports two installation styles:
- a
fullworkstation install for laptops and desktops - a lighter
edgeinstall for always-on single-board computers and hybrid Linux boards
Use this page to choose a supported board, the right install profile, and a practical deployment pattern for edge use.
SBC Support Matrix¶
| Board | Status | Best fit | Notes |
|---|---|---|---|
Pi 5 (8 GB) |
Official | Demo kiosk, digital patient | Best all-round edge choice. |
Pi 5 (16 GB) |
Official | Edge studies, local AI | More headroom for longer runs. |
Pi 5 (4 GB) |
Supported | Lighter live runtime | Best with the edge profile only. |
UNO Q (4 GB) |
Supported | Hybrid demo rig | Good if the MCU drives alerts. |
UNO Q (2 GB) |
Experimental | Minimal runtime tests | Very limited headroom. |
Pi Zero 2 W |
Unsupported | None | Too little memory for a reliable runtime. |
UNO R4 / UNO |
Unsupported | None | No Linux/Python runtime. |
Status meanings:
Recommended default in the docs.
Expected to work, but not the main target.
Needs manual tuning and validation.
Not suitable for a full SDK deployment.
Edge Architecture¶
This diagram shows the default edge deployment layout:
flowchart TD
A["Raspberry Pi 5"] --> B["IINTS Patient Daemon"]
B --> C["State Store (SQLite/JSONL)"]
B --> D["Local FastAPI Service"]
D --> E["Dashboard in Browser on Pi"]
F["Raspberry Pi Connect"] --> G["Remote Shell"]
F --> H["Screen Sharing"]
G --> B
H --> E
What the pieces do:
IINTS Patient Daemon: advances the virtual patient and algorithm state.State Store: keeps persistent runtime state for status, replay, and audit.Local FastAPI Service: exposes the live API and control surface.Dashboard in Browser on Pi: gives the live glucose view on the device itself.Raspberry Pi Connect: lets you present and control the Pi from a laptop without extra SSH/VNC setup.
Install Profiles¶
Full workstation install¶
Use this on laptops or desktops where you want the whole SDK:
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -U "iints-sdk-python35[full,mdmp]"
This includes:
- simulator
- certification
- AI review
- plotting
- PDF reporting
- posters
- CareLink visual workbench
Edge install¶
Use this on Raspberry Pi or Linux-capable edge boards where the live patient runtime matters most:
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -U "iints-sdk-python35[edge,mdmp]"
This profile is designed around:
- core simulator
- SQLite state store
- FastAPI dashboard
- CLI control
- optional local AI review if Ollama is present
It intentionally avoids the heavier reporting stack by default.
One-Line Edge Workflow¶
The new edge namespace ties the SBC story together:
iints edge setup --output-dir iints_edge_demo --board raspberry_pi
cd iints_edge_demo
iints edge up --project-dir .
iints edge status --project-dir .
iints edge bundle --project-dir . --output results/edge_runtime_bundle.zip
That gives you:
- a generated edge project scaffold
- a persistent patient runtime
- a kiosk-ready dashboard
- a ZIP bundle you can move back to a laptop for deeper analysis
Raspberry Pi¶
Recommended starting setup:
Raspberry Pi 58 GBRAM- active cooling
- Raspberry Pi OS Desktop
- Raspberry Pi Connect
Typical live runtime flow:
iints edge setup --output-dir iints_pi_demo --board raspberry_pi
cd iints_pi_demo
iints edge up --project-dir .
iints edge status --project-dir .
iints edge kiosk --project-dir .
For most users, the 8 GB Pi 5 is the default recommendation.
Arduino UNO Q¶
UNO Q combines:
- a Linux-capable side for the IINTS runtime
- an STM32 side for simple physical feedback
The recommended path is:
- install the
edgeprofile on the Linux side - generate a UNO Q scaffold with
iints edge setup --board uno_q - start the Linux-side digital patient
- flash the generated bridge sketch onto the STM32 side
- run
iints edge bridge-test --port ... - run
iints edge bridge-run --project-dir . --port ...
Fastest scaffold command:
iints edge setup --output-dir iints_uno_q_demo --board uno_q
If you only need the bridge sketch:
iints edge hardware-bridge --board uno_q --output-dir uno_q_bridge
Use the dedicated guide for the full step-by-step flow:
UNO Q is still a secondary target. It is a good fit when you want visible hardware feedback on top of the Linux-side runtime, but it is not the default first install for most SDK users.
Auto-Start With systemd¶
After starting the patient once, export a service unit:
iints edge service --project-dir .
Then install it on the device:
sudo cp patient_runtime/iints-digital-patient.service /etc/systemd/system/iints-digital-patient.service
sudo systemctl daemon-reload
sudo systemctl enable iints-digital-patient.service
sudo systemctl start iints-digital-patient.service
If you want the update script and service file scaffolded automatically, start with:
iints edge setup --output-dir iints_edge_demo --board raspberry_pi
The generated folder already contains:
run_edge_patient.shlaunch_kiosk.shupdate_edge_runtime.shpatient_runtime/iints-digital-patient.servicepatient_runtime/iints-digital-patient.INSTALL.txt
Live Kiosk And Status¶
For the presentation layer:
iints edge kiosk --project-dir .
iints edge status --project-dir .
The kiosk view highlights:
- live glucose
- scenario profile
- active algorithm
- certification status
- realism review status
- one-click scenario reset buttons
Hardware Benchmark¶
To gather technical numbers for a deployment note or hardware comparison:
iints edge-benchmark \
--algo algorithms/example_algorithm.py \
--platform auto \
--output-json results/edge_benchmark.json
This reports:
- steps per second
- mean step latency
- peak process memory
- dashboard response time
- API status response time
Keep the resulting JSON with your deployment notes. It gives you concrete numbers for:
- throughput
- memory footprint
- dashboard responsiveness
Runtime Export Back To A Laptop¶
To move a live SBC run back to a workstation:
iints edge bundle --workspace patient_runtime --output results/edge_runtime_bundle.zip
That archive contains:
patient_state.dbpatient_runtime_config.jsonlive_bundle/results.csv- manifests and audit files
- certification artifacts if present
- realism review markdown if present
Scope¶
Edge deployments are still:
- research use only
- not medical devices
- not clinical treatment systems