Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

SCPN Studio Platform

The domain-neutral SDK at the foundation of the SCPN Studio ecosystem: a federation of standalone scientific studios (neuromorphic, fusion, control, quantum, magneto-inertial, phase dynamics) that share one platform, one evidence standard, and one look, composed by a federating Hub shell.

This package is the open-core foundation. Each project builds its studio vertical on it; the managed multi-tenant Hub is the separate paid layer. The boundary is permanent and documented in docs/guides/open_core_boundary.md: the claim lattice and every verification path stay open-core forever; the Hub sells operation, never trust semantics.

What makes it a category by itself

Provenance-first, as a product surface rather than only an engineering practice. Every number a studio surfaces is an evidence bundle that:

  • is a PROV-O graph, packaged as an RO-Crate profile (interoperable with the scientific data ecosystem, not a proprietary blob);
  • carries an empirical evidence level (0–3) and an orthogonal evidence_kind — measured, curated, or formally-proven — so a machine-checked proof is never rendered as if it were a measurement;
  • is signed as an in-toto attestation, with a subject_digest so a proof is voided the moment its subject drifts;
  • declares its claim boundary on a nine-state lattice (reference-validated, bounded-model, bounded-support, validation-gap, external-dependency-blocked, roadmap, toolchain-gated, refuted, voided-by-upstream-retraction) — the Hub never upgrades a bounded, roadmap, refuted, or cascade-voided number to "validated";
  • records how to reproduce it (regenerated_by + host), not just attest it;
  • chains cross-studio derivation by content digest, so the provenance graph survives coordination-log rotation.

Package surface

Subpackage Responsibility
evidence The studio.*.v1 evidence bundle, attestation, and claim-boundary lattice.
manifest Content-addressed, language-agnostic, deterministic capability manifest.
reactor_portfolio Strict reactor descriptor, source/map/authority validation, and deterministic producer parity.
verbs Verb taxonomy + per-verb attribute contract (safety tier, fidelity, timing, side-effect, proof).
identity Tenant-aware opaque identity + sessions, shared by both transport profiles.
jobs Bounded, fail-closed job/pipeline workers + the real-time timing contract.

Functional pipeline contract

Keepers can select their implemented manifest eras through load_manifest_envelope(..., supported_eras=("v2",)) and aggregate_federation(..., supported_eras=("v2",)), or the aggregation CLI's repeatable --supported-era option. Omission keeps v1; explicit selection does not change evidence admission or permit mixed-era execution. See the manifest aggregation guide.

Each Verb can now declare consumes symmetrically with produces. A consumes entry means a hard runtime dependency on an upstream studio.*.v1 schema or interface identifier; soft/demo couplings stay out of this field and remain ordinary provenance or optional inputs. manifest.resolve_pipeline() turns the visible manifest set into content-addressed pipeline edges, surfaces unresolved_upstreams, and reports incompatible-era matches as hard resolution conflicts.

Opaque native projections

EvidenceBundle(..., native_detail=mapping) retains the supplied projection in to_dict() and its RO-Crate payload. None omits it. This generic SDK does not decode or qualify native claims, so renders_as_validated is false whenever the producer supplies a projection. The wire gate likewise withholds validated presentation whenever native_detail is present, including an explicit null from another producer; all common-envelope refusals remain enforced.

The Hub preserves declared outer values and shows the original detail at its boundary. A supplied native success flag, trusted publication seal or common admission decision cannot supply native qualification. Unsupported outer wire versions remain refused.

Only the exact top-level wire key native_detail is recognised. Aliases, case variants and nested fields are ordinary unknown metadata; they do not declare a supported native projection or confer native qualification. Producers must use the canonical key.

Producer mappings are snapshotted as finite JSON data with string keys and no cycles. Later caller mutations and edits to returned JSON cannot rewrite the bundle. Hashing remains available; opaque detail is excluded from its hash.

Recovered dynamics provenance

Recovered latent-dynamics claims carry an explicit numeric_provenance.recovered_dynamics envelope. It records the recovery method, latent and observed dimensions, independent view count, observation-noise model, identifiability scope, symbolic function basis, optional rollout horizon, and held-out trajectory/dynamics scores. DYSCO-style claims are therefore bounded as recoveries within an affine gauge and a declared basis, never as direct access to the original latent coordinates.

Cascade retraction contract

evidence.RetractionRecord carries the signed studio.retraction.v1 unit used when an upstream claim must be withdrawn. The policy is stale-safe by default: cascade_retraction() walks content-addressed derived_from evidence edges and marks every downstream consumer as voided-by-upstream-retraction. A supersedes_with digest records a replacement candidate but never auto-heals existing consumers; downstream studios must re-derive against the replacement and emit a new evidence bundle.

Verifiable numeric references

The seal package signs WS-1 honesty envelopes and now carries the committed reference values required by tolerance and stochastic recompute claims. Small reference series are embedded as canonical little-endian float64 bytes under reference_value.storage = "inline" with a sha256: digest over those bytes. Large arrays use reference_value.storage = "artifact" and sign a side-artifact descriptor containing artifact_uri, media type application/vnd.anulum.float64-array.v1, dtype = "float64", byte_order = "little", shape, content_length, and canonical_value_digest.

The SDK validates the signed descriptor. Inline payloads must fit within 16 KiB and their digest must match the embedded bytes. reference_artifact_descriptor, seal and verify do not retrieve side-artifact bytes. The caller must retrieve them and check their length and digest against the signed descriptor before comparing numeric outputs. A valid seal authenticates the metadata; it does not establish artifact retrieval or numerical agreement.

Development

See the development guide for the complete Python environment, hash locks, Node/pnpm toolchain and local gates.

Status

Early development. The v1 contract is locked; the platform is being extracted from the reference SNN Studio implementation and hardened to the contract.

Licence

Dual-licensed under the GNU AGPL v3 with a commercial option. For commercial licensing enquiries, contact protoscience@anulum.li.

Validity-domain intervals

A numeric validity range is a closed interval: low and high must be finite numbers and low <= high. Both endpoints are included, and a singleton such as [3, 3] is valid. Booleans, NaN, Infinity, JSON numeric overflow and reversed bounds are refused. A range names its parameter and may declare its unit.

The SDK checks this when constructing ValidityRange. Python and TypeScript federation gates check supplied domains again on untrusted wire input, including stored era-v1 evidence; the browser parser refuses malformed intervals before returning a renderable bundle. The browser also rechecks numeric intervals before granting validated presentation, so later changes to a parsed range cannot retain validated tone. Existing proof, refutation, freshness and native-detail boundaries continue to apply.

A qualitative domain may use a non-empty note without numeric ranges. Do not encode unbounded or unknown scope as Infinity. Well-formed interval metadata is only a declared applicability boundary; it does not establish scientific validity, source trust, native qualification or permission to control a device.

Text and data mining reservation

The rightsholder, Miroslav Šotek (ANULUM), expressly reserves the use of the works in this repository for text and data mining, including the collection of this content for, and its use in, training, fine-tuning or evaluating machine learning and artificial intelligence models, in accordance with Article 4(3) of Directive (EU) 2019/790 and the national provisions transposing it.

No permission for such use has been granted beyond what the licence in LICENSE grants. This reservation does not change that licence: uses the licence permits remain permitted on its terms, including its attribution and share-alike conditions.

Reservation made on 2026-10-19. Requests for permission: contact the rightsholder through the contact listed in this repository.

Metadata

Release files for scpn-studio-platform 0.11.3.dev0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for scpn-studio-platform 0.11.3.dev0
File Size Uploaded Provenance Actions
scpn_studio_platform-0.11.3.dev0.tar.gz 239.8 kB With provenance Details for scpn_studio_platform-0.11.3.dev0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for scpn-studio-platform 0.11.3.dev0
File Interpreter ABI Platform Actions
scpn_studio_platform-0.11.3.dev0-py3-none-any.whl Python 3 none any Details for scpn_studio_platform-0.11.3.dev0-py3-none-any.whl

Total release size: 441.4 kB

Release files / scpn_studio_platform-0.11.3.dev0.tar.gz

Download URL scpn_studio_platform-0.11.3.dev0.tar.gz
Size 239.8 kB
Tags Source
SHA-256 checksum
How to use checksums
067dc4f7fcff7e15ff63002d806825d6f70f49799c00386bf3214de52b58ccfe
BLAKE2b-256 checksum
How to use checksums
b6b65f57ab80e7cd6dffa5299822156bd7f470d0b135badb21928c9dc6fe19e1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / scpn_studio_platform-0.11.3.dev0-py3-none-any.whl

Download URL scpn_studio_platform-0.11.3.dev0-py3-none-any.whl
Size 201.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
d6cbb1c6dd7f5ba60ba0590c933c2e0e1b401800bffe2a4ca8df2c75fd8c6162
BLAKE2b-256 checksum
How to use checksums
d77e1da1db5d00fcff02460ea92786dbcf016aba8af84482dd2c129676b6bf8f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Security for scpn-studio-platform 0.11.3.dev0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for scpn_studio_platform-0.11.3.dev0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page