Monitoring Diagnostics Toolkit 0.17.0 — public research alpha

Initial public release of the original equipment-monitoring toolkit.

• Eleven equipment diagnostic kinds plus Python modules for normal residuals, causal evidence, filter service-risk, physical input preparation and alarm policies.
• Detailed README tables describe each module purpose, sensor roles/units, parameters, outputs and applicability. API reference, user/model guides, retrospective development history, rejected candidates, provenance, reproduction and research handoff are included.
• Local software/physics regression: 286 passing tests; scoped upstream object integrations: 7 passing tests. Public Windows/Linux Python 3.11/3.12 CI passes. Installed-wheel README and 10-diagnostic synthetic CLI execution pass.
• The31 core source ASTs match evaluated0.16 after removing docstrings/version; this release makes no new independent measured-data accuracy claim. Reused heat-pump gains remain confounded by a0.3K condensing-temperature session difference; engine-filter faster candidates were rejected for false alarms.
• Original code/docs: MIT. Raw datasets, papers, upstream source and trained corpus profiles are excluded and retain separate terms.

Zenodo software archive: https://doi.org/10.5281/zenodo.23992253

The attached wheel and sdist are the exact audited Zenodo files. PyPI publication is handled separately by the project-scoped trusted-publisher workflow. Alpha research software; alarms do not establish root causes or field safety qualification.