# GEAIRM 0.1 Automated Conformance Runner — A12

GEAIRM-A12 provides the deterministic offline runner for the authoritative A10 positive and A11 negative / semantic-boundary testcase manifests.

## Authoritative discovery

The runner **does not discover fixtures by scanning directories**. It executes only testcase entries declared by:

- `../testcases/geairm-0.1.positive-testcases.json` — 17 A10 positive fixtures;
- `../testcases/geairm-0.1.negative-testcases.json` — 36 A11 negative / boundary fixtures.

The combined A12 suite therefore contains **53 testcases**.

## Validation layers

A12 executes the machine-checkable layers declared by each testcase:

- fixture SHA-256 integrity;
- JSON syntax;
- A5 context / namespace policy where declared;
- A7 Draft 2020-12 structural schema;
- A8 Profile / Extension schema;
- RDF/Turtle parsing;
- A6 SHACL constraints where declared.

Semantic-assessment declarations are preserved in the report but are **not machine executed**.

> **Conformance boundary:** a runner PASS means that the observed machine behavior matches the testcase manifests. It does not establish complete GEAIRM semantic conformance, AIPS Product qualification, governance approval, architecture approval, or any hybrid/assessment-only conclusion.

## A6 SHACL execution profile

The current A6 artifact uses a deliberately conservative SHACL subset:

- `sh:NodeShape` only;
- `sh:targetNode` and `sh:targetClass` targets;
- `sh:SPARQLConstraint` with `sh:select`.

A12 executes that exact profile using RDFLib. It is **not a generic SHACL implementation**. The runner validates the A6 shape profile before execution and fails closed if unsupported SHACL constructs appear in a future revision. This prevents a future A6 expansion from being silently ignored.

Current frozen A6 profile checked by A12:

- 24 NodeShapes;
- 34 SPARQL constraints.

## Requirements

- Python 3.10+
- `jsonschema` 4.x
- `rdflib` 7.x

The runner requires no network access at execution time. See `requirements.txt`.

## Usage

From the repository root:

```bash
python static/spec/aips/geairm/0.1/conformance/geairm-0.1-conformance-runner.py \
  --report static/spec/aips/geairm/0.1/conformance/geairm-0.1.conformance-report.json
```

Run only one authoritative suite when required:

```bash
python static/spec/aips/geairm/0.1/conformance/geairm-0.1-conformance-runner.py --suite positive
python static/spec/aips/geairm/0.1/conformance/geairm-0.1-conformance-runner.py --suite negative
```

The runner derives the repository root from its canonical repository location. `--root <path>` may be supplied explicitly for CI or extracted-package execution.

## Exit codes

| Exit | Meaning |
|---|---|
| `0` | All executable expectations matched. |
| `1` | At least one testcase or executable-layer expectation mismatched. |
| `2` | Runner, configuration, dependency, unsupported-A6-profile, or other execution error. |

## Reference A12 execution

The checked-in deterministic reference report is `geairm-0.1.conformance-report.json`.

Reference result:

- 53/53 testcase expectations matched;
- 208/208 executable-layer expectations matched;
- 41 semantic-assessment declarations preserved;
- 0 semantic assessments machine executed;
- 0 complete semantic-conformance claims made.

The reference report contains no wall-clock timestamp so identical inputs and dependency behavior produce stable report content.
