Application Honesty API¶
scpn_quantum_control.applications.honesty_kits is the public BL-63 policy
surface for immutable domain claim boundaries and deterministic evidence.
scpn_quantum_control.applications.dataset_catalog owns the complementary
packaged-artifact privacy audit.
These APIs validate software metadata and versioned local artifacts only. They do not certify domain fidelity, clinical use, facility prediction, operational control, hardware performance, or quantum advantage. For the guided workflow and extension rules, read Domain Application Honesty Kits.
Honesty-kit policy records¶
scpn_quantum_control.applications.honesty_kits ¶
Fail-closed claim and data boundaries for domain-facing applications.
The objects in this module do not certify domain validity. They make the opposite boundary explicit: each kit identifies the small software route that is supported, the data origin admitted by that route, and the claims that remain forbidden. The built-in registry covers the BL-63 power-grid, Josephson, EEG-like, and ITER-inspired application families.
All returned records are immutable and JSON-ready. The audit functions are local and deterministic; they do not read credentials, contact providers, submit hardware work, or inspect private datasets.
ApplicationSupportStatus ¶
Bases: str, Enum
Public support grade for a domain-facing application route.
BOUNDED_RESEARCH means that the named software path is tested for its
documented small benchmark, while SIMULATION_ONLY requires generated
inputs and forbids measured-domain interpretation.
ApplicationDataOrigin ¶
Bases: str, Enum
Admitted input provenance for an honesty kit.
DomainApplicationHonestyKit
dataclass
¶
Immutable claim boundary for one domain-facing application family.
Parameters¶
kit_id
Stable machine identifier for the kit.
domain_tag
Non-promotional domain label used in reports and user interfaces.
title
Human-readable kit name.
support_status
Whether the route is a bounded research benchmark or simulation-only.
data_origin
Provenance class admitted by this kit.
synthetic_only
True when measured or curated domain data must not enter the route.
dataset_ids
Packaged public catalogue identifiers governed by the kit. An empty
tuple means that the route generates its inputs in code.
source_modules
Import paths implementing the bounded route.
allowed_uses
Positive, narrowly worded descriptions of supported software use.
caveats
Scientific and operational limitations that callers must preserve.
claims_forbidden
Explicit claims that this kit never authorises.
forecasting_tags
BL-37 simulation-only tags that may be cross-referenced. These tags
do not convert a synthetic forecast into domain evidence.
Notes¶
Construction validates the internal policy relationships. In particular, a synthetic-only kit cannot declare curated input data or packaged dataset identifiers, and every kit must retain at least one forbidden claim.
ApplicationHonestyAuditReport
dataclass
¶
Deterministic aggregate of honesty kits and dataset privacy checks.
Parameters¶
kits
Validated built-in honesty-kit records.
dataset_privacy
Catalogue privacy audit rows. Each row has already loaded and
validated the corresponding packaged QPUDataArtifact.
as_dict ¶
Return the canonical JSON evidence payload including its digest.
list_domain_application_honesty_kits ¶
Return all built-in BL-63 honesty kits in stable registry order.
get_domain_application_honesty_kit ¶
get_domain_application_honesty_kit_for_dataset ¶
Return the unique kit governing a packaged dataset identifier.
Synthetic-only kits intentionally have no packaged dataset identifiers and therefore cannot be resolved through this function.
Raises¶
KeyError
If no built-in kit governs dataset_id.
RuntimeError
If registry corruption assigns the same dataset to multiple kits.
build_application_honesty_audit_report ¶
Build deterministic local evidence for every kit and catalogue row.
Returns¶
ApplicationHonestyAuditReport Immutable report with a canonical content digest.
Notes¶
The audit loads only versioned packaged application artifacts. It performs no network access and never opens a user-supplied or private dataset.
Packaged-dataset privacy records¶
scpn_quantum_control.applications.dataset_catalog.ApplicationBenchmarkDescriptor
dataclass
¶
Metadata and privacy boundary for a packaged benchmark artifact.
contains_personal_data describes the packaged file, not every possible
external input accepted by a third-party plugin. The built-in catalogue
is intentionally restricted to curated public constants and small matrices
with no raw participant, clinical, SCADA, or proprietary facility records.
path
property
¶
Absolute path to the packaged artifact.
scpn_quantum_control.applications.dataset_catalog.ApplicationBenchmarkPrivacyAudit
dataclass
¶
One successful packaged-dataset privacy audit row.
Parameters¶
dataset_id
Stable packaged dataset identifier.
source_mode
Validated artifact provenance mode.
privacy_classification
Descriptor classification for the packaged bytes.
contains_personal_data
Whether the packaged artifact contains personal data. Built-in rows
must remain False.
privacy_boundary
Exact licence/provenance note bound to the artifact metadata.
artifact_hashes
Validated SHA-256 custody hashes embedded in the artifact.
passed
Always True for returned rows; mismatches raise instead of
returning an ambiguous partial result.
as_dict()
¶
Return a JSON-ready audit row with defensive hash copying.
scpn_quantum_control.applications.dataset_catalog.audit_application_benchmark_privacy()
¶
Audit every packaged application artifact against its privacy descriptor.
Returns¶
tuple[ApplicationBenchmarkPrivacyAudit, ...] One immutable, JSON-ready success row per catalogue descriptor.
Notes¶
This audit reads only files beneath data/public_application_benchmarks.
It never traverses external paths, downloads data, or treats a curated
matrix as raw domain evidence. Any mismatch raises immediately.