ASE-SDK v1.1.1 — Reference Validation Release

This release marks the first complete and validated public version of ASE-SDK.

ASE-SDK implements the Admissible Step Engine (ASE): a deterministic, stateless safety layer that gates proposed state-update steps (ΔS) such that the resulting next state remains within a formally defined admissible domain.

What ASE-SDK is

• A generic safety envelope over state-update dynamics
• Applicable to control loops, optimization pipelines, and learning systems
• Deterministic, fail-closed, and explicitly bounded by construction

What ASE-SDK is not

• Not an optimizer
• Not a learning algorithm
• Not a controller or policy

Validation scope

This release includes executable validation evidence demonstrating the use of ASE as a safety envelope for ML-style training loops:

• Parameter-level envelopes (θ)
• Extended training-state envelopes (θ + EMA + step)
• Full optimizer-state envelopes (θ + m + v + EMA + step, Adam-like)

In all cases, admissibility is enforced on the next system state, rather than on raw gradients or internal optimizer signals.

Reproducibility

All validation results are reproducible via deterministic builds and test runs. Internal simulations are provided for validation purposes and are not part of the public API.

Documentation

• README.md — API contract, design intent, and integration boundaries
• VALIDATION.md — executable validation evidence and test references