AU Reversibility Guardrails (In-Camera)¶
Non-binding. Advisory only. Originally written when the engine was US-first; now superseded by the AU-first RTOpacks pivot.
Status¶
Non-binding. Advisory only.
This document records architectural guardrails intended to preserve optional future compatibility with regulated vocational education markets (e.g. Australia), without embedding compliance logic into the UCCA Engine.
Nothing in this document creates product commitments, feature requirements, or implementation obligations.
Purpose¶
UCCA Engine is being designed and delivered for a US-first, learner-centric context.
However, given prior experience operating in heavily regulated vocational systems, it is strategically prudent to ensure that core architectural decisions do not foreclose future reinterpretation of engine outputs for compliance-driven markets, should that ever be required.
This document captures what must not be lost in the engine's design in order to preserve that optionality.
Explicit Non-Goals¶
UCCA Engine does not aim to:
- Model Australian regulatory frameworks (ASQA, AQF, RTO standards)
- Generate compliance artefacts (TAS, TAMs, RPL tools, policies)
- Enforce volume-of-learning heuristics
- Support audit workflows
- Replace SMS, LMS, or compliance consulting ecosystems
Any future interaction with regulated markets is expected to occur via downstream interpretation layers, not core engine functionality.
Core Guardrails¶
1. Preserve Unit Atomicity¶
- Units of learning remain first-class, indivisible objects
- No consolidation or blending that obscures unit boundaries
- Any aggregation must remain reversible back to unit-level truth
Rationale: Regulated markets make determinations at the unit level, even when delivery or assessment is clustered.
2. Preserve Provenance and Version Identity¶
Engine outputs must remain identifiable by:
- generation timestamp
- source inputs
- transformation rulesets
- version identifiers
Rationale: Audits are retrospective. The ability to explain when and how content was produced is more important than the content itself.
3. Contextualisation Must Be Additive, Not Destructive¶
- Contextual overlays must not overwrite original intent
- Base content must remain recoverable
- No irreversible mutation of source material
Rationale: Compliance narratives rely on demonstrating authorship and decision logic, not replacing truth with local customisations.
4. Preserve Decision Surfaces¶
The engine should avoid collapsing decisions into opaque prose where possible.
Where judgement exists (scope, sufficiency, emphasis), it should remain reconstructable even if not explicitly formalised.
Rationale: Regulatory scrutiny focuses on "why this was sufficient" rather than "what this contains".
5. Avoid Regulatory Semantics in the Core¶
The engine must not internally encode:
- jurisdiction-specific terminology
- compliance constructs
- audit logic
- regulatory thresholds
Rationale: Compliance regimes are volatile, interpretive, and external. Embedding them would contaminate the abstraction and reduce portability.
Strategic Framing¶
Future interaction with regulated markets, if any, should be treated as:
- downstream reinterpretation
- sidecar tooling
- external adapters or renderers
— not core engine evolution.
This preserves both architectural integrity and strategic optionality.
Review Notes¶
This document is intended to be reviewed only when:
- considering entry into a regulated market, or
- evaluating a proposed change that might reduce architectural reversibility.
It should not be used to justify near-term feature work.
End of Document
Version History¶
| Version | Date | Change | Author |
|---|---|---|---|
| 1.0 | 2026-03-11 | Archived from engine/ucca-engine/docs/architecture/AU_Reversibility_Guardrails.md | Claude Code |