v1.5.4¶

Release date: 2026-04-25

v1.5.4 is the edge deployment and long-horizon study release.

This release takes the Raspberry Pi and UNO Q story from a strong booth demo into a more complete edge research platform: one-command remote deployment, offline recovery bundles, Pi-local reproducible studies, and multi-day long-study workflows with resume and storage-safety support for real unattended runs.

Highlights¶

1. One-command remote Raspberry Pi deployment¶

The SDK can now provision and start a remote Raspberry Pi from a workstation with a single command:

iints edge deploy \
  --host raspberrypi.local \
  --user pi \
  --local-output-dir iints_pi_demo \
  --remote-dir ~/iints_pi_demo \
  --board raspberry_pi \
  --scenario-profile expo_hot_start

Key additions:

iints edge deploy
iints edge remote-status
iints edge remote-reset
iints edge remote-stop
SSH retry and timeout controls
dry-run and verbose deploy modes

The deploy flow is now much more practical for real devices:

the remote Pi generates its own edge project in-place
service files and scripts match the remote filesystem instead of baking in local paths
kiosk, autostart, and watchdog setup can be pushed in the same flow
UNO Q bridge support can be included during deploy

This makes the edge workflow much less fragile for Maker Faire, classrooms, and lab benches.

2. Raspberry Pi Connect is now the recommended remote access story¶

The SDK docs and remote flow now align around a safer default:

keep the live dashboard bound to 127.0.0.1
use Raspberry Pi Connect for remote screen and shell access
avoid casually exposing the patient dashboard on the LAN

This gives the Pi a cleaner “remote booth” story without weakening the security posture of the live patient service.

3. Offline install bundles for unreliable networks¶

You can now build a transportable edge bundle for expo or field use:

iints edge offline-bundle --output iints_offline.tar.gz

The bundle includes:

an offline wheelhouse/
an install helper script
an offline install guide
a prepared edge project scaffold

This is especially useful when:

venue Wi-Fi is unreliable
a Pi needs to be recovered quickly from USB media
the demo machine must be reinstallable without internet access

4. Pi-local reproducible studies¶

The edge runtime can now run reproducible multi-seed studies directly on a Raspberry Pi:

iints edge study \
  --algo algorithms/example_algorithm.py \
  --seeds 1,2,3,4,5 \
  --output-dir results/pi_study

Generated outputs include:

protocol and matrix artifacts
study summaries
edge runtime metadata about the device and run context

This strengthens the scientific story for EUCYS and similar work: the edge hardware is no longer only a live demo target, but also a reproducible experimental endpoint.

5. New long-horizon edge study engine¶

The SDK now supports multi-day and multi-week edge studies from a single configuration file:

iints edge long-study \
  --config edge_long_study.yaml \
  --project-dir .

New commands:

iints edge long-study
iints edge study-snapshot
iints edge study-export

The new long-study flow supports:

weekday schedules with different daily profiles
multiple algorithms and seeds
export-friendly file-based study storage
automatic study summaries and manifests
study archives that can be moved to another machine for analysis

This makes the Pi much closer to a real unattended research rig instead of a short demo-only runtime.

6. Safer long runs on Raspberry Pi storage¶

Long studies now have a much better storage model for real Pi hardware.

The default generated edge_long_study.yaml now uses:

USB SSD-ready output paths such as /media/pi/usb_ssd/results/long_study
scratch-first execution in /tmp/iints_edge_long_study

That means:

intermediate writes do not hammer the SD card directly
completed day outputs are synced into the durable study directory
the resulting study remains easy to export and inspect as normal files

For unattended edge studies, this is a major robustness improvement.

7. Resume support after reboot or interruption¶

Long studies can now continue cleanly after a restart:

iints edge long-study \
  --config edge_long_study.yaml \
  --project-dir . \
  --resume

The resume logic inspects long_study_index.csv, finds the last contiguous completed day, and resumes from the next one. This avoids wasting completed work after a reboot or power interruption.

8. New day profiles for more realistic weekly schedules¶

The live patient runtime now includes more day-scale profiles for long studies and booth demos:

school_day
relaxed_day

These complement existing profiles such as sport_day and bad_carb_count, making rolling weekly schedules more realistic and easier to describe in study configs.

9. Edge documentation is now much stronger¶

The edge documentation has been expanded with:

a dedicated remote deploy guide
clearer Raspberry Pi setup guidance
updated command reference coverage
improved Maker Faire notes for Pi and UNO Q paths

This release is noticeably easier to adopt if you are coming in fresh and trying to get from laptop to working Pi without reverse-engineering the repo.

Validation¶

Before release, the updated SDK passed:

python3 -m pytest tests/ -q
342 passed, 4 skipped
flake8 .
mypy src/iints/
mkdocs build --strict
python3 -m build

That validation included the new edge deploy, offline bundle, edge study, and long-study paths.

Upgrade¶

Full workstation:

python -m pip install -U "iints-sdk-python35[full,mdmp]==1.5.4"

Edge / Raspberry Pi:

python -m pip install -U "iints-sdk-python35[edge,mdmp]==1.5.4"

Practical takeaway¶

v1.5.4 makes the edge SDK much more credible as both:

a public-facing demo system for Raspberry Pi and UNO Q
a serious long-horizon research platform that can run, preserve, resume, and export studies from the edge itself

That combination is what makes this release important: it improves reliability and presentation at the same time, without losing the scientific workflow.