v1.5.6¶
Release date: 2026-05-18
v1.5.6 is the Jetson research-loop patch release.
It closes the gap between the work added after v1.5.5 and what a user can install from PyPI. The most visible fix is simple: the documented Jetson flags now ship in the released wheel too.
Highlights¶
1. Jetson research mode is now actually public¶
The installable CLI now includes:
iints jetson endurance start --wall-clock
iints jetson endurance start --finalize-research
iints jetson endurance finalize-research
That means a clean released install can:
- run a true wall-clock 24-hour Jetson study
- export a research dataset bundle
- train post-run local models
- write a closed-loop evaluation report
2. Local AI research is now a full loop¶
The SDK now includes:
- source-aware predictor dataset blending
- controller teacher-dataset generation
- an auditable linear imitation baseline
- a stronger PyTorch neural controller
- held-out closed-loop comparison against the clinical baseline
- one-command Jetson research finalization
The important shift is that the SDK no longer stops at “a model trained.” It now also produces the first evidence layer needed to compare that model under unseen simulator scenarios.
3. Quickstart and demo flows are easier to trust¶
This patch also carries the work added after v1.5.5:
- self-contained quickstart patient files
- safer physiology defaults and run-preview diagnostics
- clearer early-termination summaries
- a more polished live-demo presenter flow for calls and expo use
- physiology reference docs plus EUCYS-ready handouts
Validation¶
Before release, the updated SDK passed:
python3 -m pytest -q411 passed, 4 skippedflake8 .mypy src/iints/mkdocs build --strict
Upgrade¶
Full workstation:
python -m pip install -U "iints-sdk-python35[full,mdmp]==1.5.6"
Edge / Raspberry Pi / Jetson:
python -m pip install -U "iints-sdk-python35[edge,mdmp]==1.5.6"
Practical takeaway¶
If you installed v1.5.5 and saw No such option '--wall-clock' or No such option '--finalize-research', your environment was not broken; those flags landed after the v1.5.5 tag. Upgrade to v1.5.6 and the released package now matches the documented Jetson research workflow.