Production Readiness Boundary

SCPN Control is not certified facility-control software. It is a research and control-integration library with production-oriented engineering gates around installation, API contracts, bounded validation, provenance, tests, and fail-closed claim admission.

This page separates the current library-readiness status from the much stricter requirements for tokamak plant operation.

Readiness summary

Scope	Current status	Meaning
Library packaging and APIs	Production-oriented	Public Python package, documented APIs, optional dependencies, generated capability inventory, and compatibility checks are maintained.
Local software quality gates	Production-oriented	Module-specific tests, lint/type checks, pre-commit gates, generated traceability checks, and security scans are part of normal development.
Bounded physics/control evidence	Bounded research evidence	Repository benchmarks, deterministic fixtures, formal checks, and checksum/provenance gates support limited claims only inside declared domains.
Facility physics validation	Not complete	Many physics surfaces still require measured-shot, external-code, or documented public reference artifacts before quantitative or facility claims are admissible.
Real plant deployment	Not ready	No commissioned PCS deployment, no CODAC/EPICS plant acceptance, no qualified WebSocket control-stream runtime artifact, no hardware interlock qualification, no operator procedure package, and no safety certification evidence pack.
Safety certification	Not ready	Certification would require hazard analysis, requirements traceability, independent V&V, configuration management, audit trails, and device-specific acceptance evidence.

Claim status levels

Level	Allowed wording	Required evidence	Current use in this repository
Bounded	Bounded model, bounded regression, deterministic fixture, local acceptance	Unit/module tests, deterministic benchmark reports, declared assumptions, provenance metadata, and fail-closed APIs	Most controller facades and many physics adapters are here.
Reference-validated	Validated against named documented/public/reference artefacts	Immutable artifacts, SHA-256 digests, unit contracts, declared tolerances, and passing strict validators	Available only for surfaces with matching reference artifacts.
External-code validated	Matched to GENE, TGLF, CGYRO, EFIT/P-EFIT, VMEC, TRANSP/TSC, or equivalent	Same case definitions, code/version provenance, numerical tolerances, and reproducible reports	Required for quantitative solver claims; not assumed by default.
Measured-facility validated	Matched to measured discharges or facility replay data	Shot IDs, diagnostics, acquisition manifests, facility data policy, checksums, and error metrics	Required for facility-control claims.
Deployment-certified	Safe for commissioned plant operation	Facility safety case, interlock qualification, operator procedures, latency envelope, cyber review, independent V&V, and acceptance sign-off	Not claimed.

What is production-oriented now

• API and package surfaces are treated as production software contracts.
• Optional external dependencies are guarded and fail closed where practical.
• Claim-admission helpers reject unsupported facility claims unless reference artifacts pass strict validators.
• Physics traceability records validity domains, public-claim status, required actions, evidence paths, and external-validation blockers.
• Module-specific tests are preferred over generic coverage buckets.
• Benchmark reports are persisted only as bounded evidence unless their matching validator admits the stronger claim.

What is not production-grade for plant operation

• No real tokamak actuator authority is claimed.
• No safety interlock or machine-protection acceptance is claimed.
• No ITER CODAC, EPICS, MARTe, or site PCS commissioning evidence is claimed.
• No operator training, alarm response, or procedure package is included.
• No measured-facility validation is implied by synthetic, deterministic, or repository-local benchmarks.
• No external-code agreement is implied unless the exact artifact gate for that surface passes.

Reading the evidence correctly

Use validation/physics_traceability.json and docs/physics_traceability.md as the authority for public claim boundaries. A benchmark report means the bounded path is reproducible. It does not by itself upgrade a model to measured-facility validation or deployment certification.

Before presenting a stronger claim for any surface, require all of the following:

1. The traceability entry allows the claim level.
2. The strict validator for that surface passes on the relevant artifacts.
3. The artifact declares source provenance, model identity, units, checksums, case count, and tolerances.
4. The report states the same validity domain as the public documentation.
5. The change is reflected in the changelog, validation docs, and generated traceability report.

Decision tree for users

• If you need a Python package for research, bounded controller experiments, formal checks, and reproducible validation reports, SCPN Control is usable today inside the documented claim boundaries.
• If you need publication-grade quantitative physics claims, use only surfaces whose strict validators admit the required public, external-code, or measured artefacts.
• If you need target-hardware or HIL claims, require target-specific timing, interlock, backpressure, replay, and hardware metadata reports.
• If you need plant deployment, treat this repository as supporting evidence infrastructure only; facility-specific safety engineering remains required.

Current concise classification

SCPN Control is a production-oriented research and control-integration library. It is not production-grade software for real tokamak operation or certified industrial deployment.