Skip to content

SPDX-License-Identifier: AGPL-3.0-or-later

Commercial license available

© Concepts 1996-2026 Miroslav Sotek. All rights reserved.

© Code 2020-2026 Miroslav Sotek. All rights reserved.

ORCID: 0009-0009-3560-0851

Contact: www.anulum.li | protoscience@anulum.li

scpn-quantum-control - Pipeline Runtime Contract

Pipeline Runtime Contract

Every component that claims to participate in the SCPN stack must expose a runnable pipeline path. Importable modules are not enough: each pipeline needs a command, a typed contract, validation, and an auditable output artifact.

Three-repository runtime model

Repository Runtime role Contract
SC-NeuroCore Source-facing data bridge and neural/SCPN payload provider. Provides scpn_neurocore.bridge loaders and labelled source artifacts.
Phase Orchestrator Domain-to-oscillator compiler. Emits validated QPU data artifacts from binding specs, domain packs, or calibrated streams.
SCPN Quantum Control QPU compiler, executor, and measurement analyser. Consumes QPU data artifacts and produces hardware/simulator result artifacts.

The data flow is:

source data -> SC-NeuroCore bridge -> Phase Orchestrator compiler
            -> QPU data artifact -> Quantum Control circuit/runtime
            -> counts/result artifact -> analysis/publication gate

Quantum Control must consume the final artifact. It must not invent campaign matrices as a fallback path.

Runnable pipeline definition

A pipeline is runnable only when all of the following are true:

  1. It has a stable entry point: console command, script, or documented Python function.
  2. It accepts explicit inputs: source path, replay id, artifact path, or configuration object.
  3. It emits a typed artifact with schema version, source provenance, numeric hashes, and reproducibility metadata.
  4. It has a smoke test that exercises the entry point without external paid services.
  5. It has at least one integration test covering the handoff to the next repository.
  6. It fails loudly when source data or optional hardware is unavailable.
  7. Synthetic data requires explicit opt-in and is labelled in metadata.

Purpose-tuned instances

Multiple instances may run at the same time when their state is isolated. This is useful because the same stack can serve different purposes:

Instance type Example purpose Isolation key
Smoke instance Fast contract check with fixture artifacts. Temporary output directory and fixture replay id.
Replay instance Re-run a recorded source/campaign exactly. Replay id and immutable artifact hash.
Calibration instance Tune extraction, coupling, or mitigation parameters. Calibration id and parameter manifest.
Hardware instance Submit QPU jobs under cost and queue controls. Backend, job id, shot budget, and result directory.
Analysis instance Recompute observables from saved counts. Result artifact hash and analysis version.

The opening is significant: we can run a connectome replay, a plasma calibration, and a hardware retrieval loop concurrently without them mutating each other's state. Each instance becomes an auditable experiment rather than a terminal session.

Minimum command surface

The target command surface is:

sc-neurocore bridge export --source SOURCE --mode MODE --out artifact.json
spo qpu-artifact emit --domain-pack PATH --mode MODE --out artifact.json
python -m scpn_quantum_control ... --artifact artifact.json --out result.json

Until those exact commands exist, each repository must document the current equivalent Python function or script and its limitations.

Publication gate

Publication-grade runs require:

  • source_mode in recorded, replay, curated, or derived
  • a source_timestamp or replay_id
  • stable SHA-256 hashes for K_nm, omega, and optional theta0
  • no silent local random fallback
  • no hard-coded observable targets
  • preserved QPU job ids or simulator seeds
  • result artifacts that state whether hardware, simulator, mitigation, and proxy observables were used

Smoke fixtures are valuable for pipeline health, but they are not publication evidence.

Current readiness snapshot

As of 2026-05-12:

  • SC-NeuroCore exposes the canonical source-facing bridge namespace scpn_neurocore.bridge. The prior unseparated package spelling is no longer the active tracked package.
  • scpn_neurocore.bridge.load_live_stream returns an artifact object with K_nm and omega, and bridge-only imports do not eagerly load datastream codec dependencies.
  • Phase Orchestrator has a QPU data artifact emitter/validator; focused artifact tests pass locally.
  • Quantum Control imports the canonical namespace in the frontier parameter generator and live-loop script. Focused generator tests consume SC-NeuroCore-style artifact objects.

The next integration target is a one-command dry run:

SC-NeuroCore fixture/replay -> Phase Orchestrator artifact
-> Quantum Control simulator -> result artifact

That dry run should become the permanent inter-repository health check.