Pilot dependency-free function citation declarations for PyUnitWizard

What — Completed bounded development pilot of dependency-free function citation declarations for quantity, convert, conversion_factor and standardize.

How — Original PyUnitWizard software bibliography/version is exported as inert metadata. Applications explicitly enable Ackredit's development observer; completed Pint/unyt backend attribution remains separately opted into. Added direct numerical/citation guards for conversion-factor and standardization entry. Normal import and scientific behavior retain their optional-provider boundary.

Why — The real producer and installed receiver now satisfy the pilot's criteria: original references/versions/roles, independent captures and pipeline parentage, selected backends, entry versus successful dispatch semantics, lazy activation, missing provider, and producer-blocked saved reading. This accepts experimental receiving; stable schema/API classification and release remain with upstream owners.

Validation — Exact Ackredit development producer 1a5dd4566737f6195571b4cb421af6f01647c5f6, wheel SHA-256 4c1d65477af8a09f66873c119b40c9228e6e678a79e360a2e4c0fd4644080604: seven source-CI jobs and eight Linux/macOS-arm64 × Python 3.11–3.14 installed receiving cells pass, 48 tests without skips. The immutable provider-owned summary verifies wheel/package-file/test-event identities and equals the hosted aggregate. That matrix's PyUnitWizard source is 33fec8a; current scientific source ef85201 is additionally qualified locally with the exact provider and a normally installed identified receiver wheel. Final current tests: 712 passed, one covered NaN/JSON skip; installed external modules: 27 passed, build-from-checkout guard separately exercised in full suite. Ruff/reporting gates pass. Raw seven-repeat conversion timings are retained. Historical hosted scope, current local scope and provider files mapped from public Conda archives are explicitly distinguished in the receipt.

Guard — tests/integration/test_function_citation_provider.py::test_normally_installed_consumer_outside_checkout; the same module covers each declared operation and saved-reader/optional semantics.

Status — Resolved experimental receiving pilot. Public Ackredit 0.9.0 has no function observer and remains unchanged; its separate portable backend qualification is #92. No required observer, client minimum, public extra, stable schema/API promotion, tag, package release, synchronized-guide rollout or universal instrumentation follows. Stable/public decisions remain uibcdf/ackredit#84/#87, uibcdf/molsyssuite#97 and uibcdf/moli#46.

Record — devguide/completed_proposals/function_citation_provider.md.
Evidence — devguide/evidence/function_provider_94_receiving_2026-10-04.json; devtools/receipts/function_provider_94_qualified_2026-10-04.json.

feat(attribution): declare function citations and reuse prepared cred…

Qualified source outcome (no new tag or release):

Ackredit source: d3fd892f2ee3d6de7b980658c2ccdb6f1de73f37, including the test-first author-tuple representation correction after initial implementation edde8a5.
PyUnitWizard source: 33fec8a627505a4f5426babe87e8e85438105041.
Local Python 3.14.7: Ackredit 1,636 passed without skips; PyUnitWizard 633 passed, ten optional/sibling-dependent skips, one existing deprecated-import warning. After the final provider correction, all 26 real attribution integration tests pass again. Ruff and developer-guide indexes pass. Ackredit strict Sphinx passes; PyUnitWizard docs build succeeds with four unchanged homepage heading warnings reproduced on unmodified upstream source.
Ackredit exact-head CI 37213060907: 7/7 jobs (docs, lint, Linux Python 3.11–3.14 and macOS arm64 3.14). MolSysSuite policy and publication policy pass.
PyUnitWizard exact-head ordinary CI 37211484215 and MolSysSuite policy pass. Its ordinary workflow does not install the provisional Ackredit observer, so this is not a hosted coupled-provider matrix. Coupled normally installed receiving evidence here is Linux/Python 3.14; broader receiving review remains open.
All hosted inspections use GH Run Receptor's compact llm reports and preserve run/job/manifest receipts.

Two normally installed clean-source wheels were exercised from an independent directory outside both repositories: Ackredit 0.9.0+23.gd3fd892, PyUnitWizard 0.27.0+52.g33fec8a. The producer verifies numerical parity, independent reused captures, unused-backend exclusion, original software versions, software/article roles, pipeline parentage and workflow union. A separate reader blocks PyUnitWizard/Pint/unyt imports, preserves the saved records and renders both the software DOI and unyt article DOI without new credit.

Receiving receipt summary (the full saved captures/graph are retained locally in /tmp/ackredit-puw-receiving-94/receipt.json):

{
  "python": "3.14.7",
  "ackredit": "0.9.0+23.gd3fd892",
  "pyunitwizard": "0.27.0+52.g33fec8a",
  "source_commits": {
    "ackredit": "d3fd892f2ee3d6de7b980658c2ccdb6f1de73f37",
    "pyunitwizard": "33fec8a627505a4f5426babe87e8e85438105041"
  },
  "source_sha256": {
    "ackredit/core/collector.py": "eb9bf60d3232a251f60e3e9e672b1a13d68dca0414ed027d97dc4aaa741e9db5",
    "ackredit/core/providers.py": "779b6ef1d8db5e3bcc667e33f983e47f02915d2e544e61e4408229b33e1f3202",
    "pyunitwizard/_ackredit.py": "28c162bee24efb2023386465a7527e21a7c033f1bc135557d906d4f0cb6e4f78"
  },
  "checks": [
    "normal_installation_outside_checkouts",
    "runtime_distribution_identity",
    "numerical_parity",
    "independent_reused_captures",
    "unused_backend_excluded",
    "software_article_roles",
    "original_versions",
    "pipeline_parentage",
    "bibtex_software_doi",
    "workflow_union",
    "separate_reader_without_producer_imports_or_new_credit"
  ]
}

Real warmed conversion measurements remain pinned to their measured runtime fingerprints in the before and after receipts: small backend result capture 272.34 -> 96.75 microseconds; combined public-function/backend capture 302.02 -> 114.79 microseconds. Seven raw samples are retained; initial preparation/imports are excluded and ordinary timing drift is reported. These are local total-time changes, not universal performance guarantees.

Ackredit and PyUnitWizard are also installed editable, without dependency resolution, in molsyssuite@uibcdf_3.14; runtime/distribution versions agree, pip check passes and all nine pre-existing Ackredit human files are byte-for-byte preserved.

Both observe_calls and prepare_credit remain provisional under uibcdf/ackredit#84/#87, with the real producer tracked in #94. Review stays open in uibcdf/molsyssuite#97 and uibcdf/moli#46. Public Ackredit 0.9.0, its immutable Conda file and the canonical integration guide remain unchanged. PyUnitWizard retains the released register/track fallback. No new required host dependency or suite-wide adoption obligation is introduced.

Preparing the next development qualification checkpoint for Ackredit #84/#87 and PyUnitWizard #94, within the existing provider/receiving review in MolSysSuite #97 and MOLI #46.

Ackredit will own a manual installed receiving workflow. One builder will produce a single immutable wheel bundle for its exact source revision, PyUnitWizard 33fec8a627505a4f5426babe87e8e85438105041, and a separate released-API baseline from Ackredit 0.9.0's original producer 598abf993a2409c025de5e912acd7eb45a257ebd. Eight Linux/macOS arm64 × Python 3.11–3.14 consumers will install the same candidate files outside the source trees.

The mandatory gate will check real Pint/unyt computations, explicit units, reused references, original versions, software/article roles, executed public-function/backend parentage, no-op/failed-entry boundaries, provider absence and a fresh producer-blocked saved reader. The original 0.9.0 source wheel will test the genuine optional-provider fallback in a separate process; it is not the immutable public Conda archive or a new release.

Bundle/installed-resource hashes, per-cell receipts, JUnit and Pytest Receptor events will be retained. The aggregate gate will use the supported Pytest Receptor reader and reject a partial matrix, skipped/deselected tests, incomplete artifacts or different candidate identities. This complements the full Ackredit suite and does not change ordinary consumer dependencies.

No sibling source changes, API promotion, canonical-guide rollout, tag or package publication is planned in this checkpoint. The provisional APIs and cross-component review remain open. Exact-source execution and receipt links will follow after local verification and the authorized Ackredit direct push.

The installed receiving checkpoint is complete, with no tag or package publication.

Qualified source and exact candidate

Ackredit producer: fc00a6cf1e2426ab7d7662fa3e9a1e3b09472306.
Normally installed candidate: ackredit-0.9.0+25.gfc00a6c-py3-none-any.whl, SHA-256 6d71199ab22aac8c00918ba4cd8ea050639109403556613161074668edc9408d.
Real producer: PyUnitWizard 33fec8a627505a4f5426babe87e8e85438105041, pyunitwizard-0.27.0+52.g33fec8a-py3-none-any.whl, SHA-256 9797dc35439b601e5a7c71bed963b49afab47a1bfd9cecb2ad9e348b2593dfba.

Executed evidence

Receiving run 37217520167: all eight Linux/macOS arm64 × Python 3.11–3.14 cells and the final aggregate pass; 40 tests, zero skips/deselections, ten retained artifacts.
Exact-producer full CI 37217509058: all seven jobs pass, including the five ordinary installed test jobs, lint and documentation. Both MolSysSuite and Conda publication policy gates also pass.
Each cell installs the same one-time wheel bundle, verifies its SHA-256 and every installed package resource, and runs outside both checkouts. The designated scientific gate requires real Pint/unyt; missing or shadowed engines fail.
The real pipeline checks numerical/unit parity, exact reference counts, unused backends, original versions, software/article roles and DOIs, reused captures, enclosing workflow and public-function/backend parentage. No-op/failed public entries earn no completed-backend credit.
Separate processes verify optional-provider absence and restore portable bibliography/graph with PyUnitWizard/Pint/unyt imports actively blocked, without new credit.
The original Ackredit 0.9.0 source is normally installed separately to exercise the genuine released-API fallback without prepare_credit; independent captures retain reused backend references. This source wheel is not the published Conda artifact.
All downloaded bundle/resource identities and complete Pytest Receptor artifacts independently verify locally through the supported reader. Remote inspections used GH Run Receptor.

Durable record and final development state

The reviewed public allow-list is retained in the immutable receipt, including per-cell original versions, reference roles, graphs, saved-payload/event hashes and fresh-process results. Raw run captures are not committed. The maintained receiving guide explains reproduction and evidence boundaries.

Ordinary CI initially exposed one remaining undeclared-NumPy preload in an old discovery test. Ackredit #88 records its reproduction and correction; the guard now actively blocks NumPy and proves the same no-sweep behavior using required modules. The archived record retains the exact-source green CI and receiving evidence. Runtime dependencies/import behavior are unchanged.

Final documentation/archive head 136b5d6fb2751659dcfd5cacd68b30e82157c881 passes CI 37218150874, MolSysSuite policy 37218151322 and publication policy 37218151290. Local final gates pass 1,655 tests without skips, Ruff, generated report indexes and strict Sphinx. The eight-cell receipt qualifies fc00a6c; the final head adds documentation/receipt/archive only and does not claim a different wheel was tested.

Ackredit remains editable in molsyssuite@uibcdf_3.14 / Python 3.14.7 as 0.9.0+26.g136b5d6.dirty, with runtime/distribution version agreement and pip check passing. The nine pre-existing human formatting/notebook/test files retain their original SHA-256. PyUnitWizard source and ordinary dependency declarations were not changed in this checkpoint.

Ackredit #84/#87 and PyUnitWizard #94 remain partial: the source integration is executed across the complete receiving matrix, while MolSysSuite #97/MOLI #46 retain review of the provisional observe_calls, prepare_credit and declaration contract. There is no API promotion, new client minimum, canonical-guide rollout or new tag. Public Ackredit 0.9.0 and its immutable Conda file remain their separate published contract.

Preparing the next Ackredit development checkpoint from 136b5d6, tracked in Ackredit #89/#90/#91.

#89 adds an explicitly selected workflow report joining the existing detached bibliography, original producer/use versions, roles/context and saved graph. It does not infer scientific success, invocation counts, chronology or complete instrumentation. Saved readers remain offline and producer-independent. Existing report formats, the portable schema and the two-argument plugin contract remain intact; the new built-in reserves workflow, so external plugins using that name must choose another. The format is absent from public Ackredit 0.9.0.

#90 repairs nested format-plugin mutation of registered bibliography, with report-time detached inputs; ordinary built-ins copy only relevant records, while plugins retain their complete-registry view. #91 replaces repeated/recursive provenance expansion with an iterative traversal: every graph edge remains, shared targets expand once with an explicit marker, cycles stay distinct and ordinary trees retain their output. Tracking/capture and scientific execution paths are unchanged.

The real PyUnitWizard pin stays 33fec8a627505a4f5426babe87e8e85438105041. Ackredit will extend its owned installed receiving gate to exercise the new report against that producer and from a producer-blocked reader, across the existing eight Linux/macOS arm64 × Python 3.11–3.14 cells. No PyUnitWizard source/dependency change or sibling source edit is planned. Exact source/bytes and reporting-only performance receipts will follow after verification.

This notice continues the consumer-impact review in MolSysSuite #97/MOLI #46 and the receiving pilot in PyUnitWizard #94. Ackredit #84/#87 remain provisional and separately reviewed. There is no canonical-guide rollout, API promotion, new client minimum, tag or public package publication in this checkpoint.

Ackredit reporting follow-up is implemented and qualified under #89/#90/#91.

The explicit development workflow format combines numbered bibliography, recorded roles, original software/article versions, contexts and saved graph. Producer-free readers neither import engines nor consult current bibliography nor add credit. Plugins cannot mutate nested registered metadata. Shared graph targets expand once with all incoming links retained, and deep graph rendering avoids recursion.

Exact source cb3e58df0a82ecc46da50f1969ddc921a097664e passes ordinary CI 37224726278 (seven jobs) and installed matrix 37228402277: eight Linux/macOS arm64 × Python 3.11–3.14 cells, 48 passed tests, zero skips/deselections, ten artifacts. Downloaded bundle/resource/event identities independently verify and the local aggregate equals the hosted summary.

Reviewed receiving evidence, reporting documentation, and bounded graph timings are retained in Ackredit. Candidate wheel ackredit-0.9.0+27.gcb3e58d-py3-none-any.whl, SHA-256 4eea193bf651c38367ba823a896022f1a5167680d901dd9e19145d7e3ddcf32b.

PyUnitWizard remains pinned to unchanged source 33fec8a627505a4f5426babe87e8e85438105041; optional absence and the actual source-built original Ackredit 0.9.0 fallback still pass. No sibling source change is requested by this checkpoint. workflow reserves a new built-in name in development; plugins using that name must choose another.

No tag/public package, provider API promotion, portable schema or canonical-guide rollout. Continue the provisional provider/receiving review in Ackredit #84/#87, MolSysSuite #97 and MOLI #46. These receipts qualify development source and do not replace public Conda or client release gates.

Independent MolSysSuite receiving review for uibcdf/molsyssuite#97 is complete for Ackredit fc00a6cf1e2426ab7d7662fa3e9a1e3b09472306 + PyUnitWizard 33fec8a627505a4f5426babe87e8e85438105041.

Run 37217520167, attempt 1: native/installed GH Run Receptor agree; all ten native artifact ZIP digests verified; provider-owned qualification_bundle.py summarize reused with public Pytest Receptor 1.2.1. Independent summary equals the uploaded aggregate: eight Linux/macOS arm64 × Python 3.11–3.14 cells, 40 passed and no skips/deselections/failures/incomplete tests. Exact wheel SHA-256: Ackredit 6d71199ab22aac8c00918ba4cd8ea050639109403556613161074668edc9408d; PyUnitWizard 9797dc35439b601e5a7c71bed963b49afab47a1bfd9cecb2ad9e348b2593dfba.

The concrete installed pilot criterion is satisfied for these development bytes. Entry observation and completed-backend credit retain distinct semantics; lazy declared PEP 562 exports and released-API fallback were reviewed. No stable API, public artifact promotion, client minimum or guide rollout follows automatically. Original 0.9.0 Conda bytes are untouched; the baseline source wheel is separate.

The later Ackredit #89/#90/#91 reporting checkpoint remains pending its own immutable source/bytes and extended installed report/reader evidence. We will keep it separate from the qualified producer. Stable-contract/release handoff and the maintainer's decision remain pending under #97; local receiving adoption belongs to #94 and direct MOLI scope to uibcdf/moli#46. No scientific suites were rerun centrally.

Central immutable review: b23c774958d1fcf9bf1360a671e822ee283be80e. Maintained record, primary review receipt. Offline governance guard passes. Hosted central validation 37228497030 passes on that exact head: 331 central tests, governance and publisher controls, plus default coverage publication.

Ackredit #92 records a reproduced pre-stabilization bug on de209caddd618042183884239ffb395e3d21a615: public registration with tuple authors is incorrectly refused by observe_calls when a provider declares the same JSON bibliography using a list (E012 before activation).

The Ackredit-owned repair compares portable values during activation, preserves accepted raw registered data, and keeps a detached raw snapshot for repeated-call replacement/deletion checks. Portable captures stay normalized; no JSON serialization moves to the per-call path. Actual conflicting or non-portable metadata remains refused before mutation, and later recording gaps remain diagnosed without changing science.

The existing real PyUnitWizard pin 33fec8a627505a4f5426babe87e8e85438105041 is unchanged. Its Ackredit-owned installed report test will pre-register valid tuple metadata, and the exact revised bundle will run across Linux/macOS arm64 × Python 3.11–3.14. This is a new qualification for changed runtime bytes, not an extension of previous evidence by assumption.

This finding also informs the concrete proposed stability/release handoff for Ackredit #84/#87. No API promotion, shared adoption rule, client minimum, synchronized-guide edit, tag or public publication is authorized by these checks. Existing public 0.9.0/fallback remains separate. Follow-up source/receipt links will be delivered here and in Ackredit #92 and PyUnitWizard #94.

The Ackredit-owned stable-contract/release handoff is now concrete and reviewable: proposal at immutable documentation head 40f930a.

Recommendation: retain the bounded observe_calls, dependency-free ackredit.provider@1 declarations and explicit prepare_credit in the next planned minor release after acceptance. The table fixes original bibliography/roles, actual function entry versus host-chosen completed dispatch, context-local restoration, optional boundaries, offline portable reading and catalog failure behavior. Existing limitations remain explicit; no universal profiler or mandatory host adoption is proposed. A released schema compatibility commitment and stable classification are proposed for a maintainer decision, not silently adopted. The release sequence retains an exact staged file, installed matrix, publication and public-channel verification as separate gates; no next tag is selected.

Pre-stabilization review found and resolved Ackredit #92: tuple-author public registration was falsely rejected against an equivalent JSON provider list. Activation now compares portable values but preserves raw registrations; repeated calls compare detached raw snapshots without added bibliography normalization. Genuine conflicts and later replacement/deletion/nested mutation are still diagnosed; science is preserved.

New runtime producer: 1a5dd4566737f6195571b4cb421af6f01647c5f6.
Real producer: unchanged PyUnitWizard 33fec8a627505a4f5426babe87e8e85438105041.
Candidate: ackredit-0.9.0+29.g1a5dd45-py3-none-any.whl, SHA-256 4c1d65477af8a09f66873c119b40c9228e6e678a79e360a2e4c0fd4644080604.

Exact-source CI 37230213187: seven jobs pass, plus both policy lanes. Local Python 3.14.7: 1,698 passed without skips, Ruff, indexes and strict Sphinx.
Exact installed matrix 37230225286: eight Linux/macOS arm64 × Python 3.11–3.14 cells, 48 passed, zero skips/deselections, ten artifacts and passing aggregate. The real report now exercises pre-existing tuple registration, original normalized metadata, pipeline/roles and a fresh producer-blocked offline reader. Optional absence and the genuine original 0.9.0 source-wheel fallback still pass.
Downloaded wheel/resource identities and complete Pytest Receptor events independently verify with the provider-owned supported summary; it equals the hosted aggregate. Reviewed receiving receipt. Remote inspection uses GH Run Receptor.

This changed-runtime qualification is distinct from the independently reviewed fc00a6c pilot and later cb3e58d reporting checkpoint. Documentation head 40f930a is not substituted for the wheel producer. Please review the proposed bounded guarantees and released-schema commitment under Ackredit #84/#87, MolSysSuite #97 and MOLI #46; receiving citation/adoption remains PyUnitWizard #94's responsibility.

No API promotion, public package, consumer minimum, synchronized-guide rollout, sibling source edit or tag. Public Ackredit 0.9.0 remains unchanged. The proposal is ready for the maintainer decision rather than keeping the concrete receiving requirement described as unimplemented.

Ackredit #93 prepares the authorized 0.10.0 checkpoint from current main 40f930a. This is provisional delivery, not acceptance of the stable proposal: observe_calls, prepare_credit and declaration interpretation remain provisional; #84/#87 and the shared contract/adoption review stay open. The public 0.9.0 portable minimum and its immutable archive remain unchanged.

Affected inventory: ACKREDIT_GUIDE consumers PyUnitWizard, MolSysMT, MolSysViewer, TopoMT, PharmacophoreMT and ElastNetMT (the canonical guide stays at its existing released portable scope); Sabueso is a separate direct MOLI receiver with its own mandatory >=0.9.0 decision. No automatic observation, additional core dependency, client source edit or new consumer minimum is required. Existing optional boundaries and the original 0.9.0 fallback are retained.

The candidate will pass declared source CI/full matrix, one staged noarch build, eight exact installed cells, and eight real PyUnitWizard receiving cells against that same filename/SHA-256 before promotion. The receiving workflow reuses the pinned MolSysSuite exact-file installer/verifier (c3e2b9b3dabf3d1c65349c389a23048957bea21a) before and after science; only PyUnitWizard 33fec8a and the released 0.9.0 fallback are wheel-built in this profile. Candidate wheels cannot substitute for Conda qualification. Final source, file digest, runs, public clean installation and original-producer tag will be handed off after verification.

Owners: Ackredit #93 delivery; Ackredit #84/#87 product stability; MolSysSuite #97 shared review and member impact; MOLI #46 direct-component boundary; PyUnitWizard #94 receiving pilot; Sabueso #108/#110 its own scientific release. No general shared-action adoption or package withdrawal is introduced.

The bounded development receiving pilot is complete, published directly to main in f34b111baee0eacf7c6c97f62e7143fcfdc2b9e7, as authorized. This accepts local experimental citation declarations and optional observation; it does not stabilize or publicly distribute the upstream function-provider contract.

Added two guards for conversion_factor and standardize, complementing quantity/convert: correct numerical results, actual reached public function, original software DOI/version and executed_software role. Scientific implementation and dependency floors are unchanged.

Exact provider: Ackredit 1a5dd4566737f6195571b4cb421af6f01647c5f6, ackredit-0.9.0+29.g1a5dd45-py3-none-any.whl, SHA-256 4c1d65477af8a09f66873c119b40c9228e6e678a79e360a2e4c0fd4644080604. Seven-job provider CI and eight-cell installed receiving matrix pass; 48 tests, no skips/deselections. The immutable provider-owned qualification_bundle.py summarize verifies exact wheel/package-file identities and complete pytest events, and its aggregate equals the hosted aggregate.

The hosted receiver is original 33fec8a; current scientific receiver ef85201 was additionally built and normally installed from wheel SHA-256 ed10e748aa5a9b8ae95f7d498ff122b9d23e746efbbe3c1a14c63d3cd6145c96. Both installed package trees match all wheel files outside the checkout. Final Python 3.14 receiving results with public ArgDigest 0.14: pip check passes; 712 tests pass, one existing covered NaN/JSON skip; 27 installed external guards pass with the build-from-checkout guard separately passing in the source suite. Ruff and nine reporting tests pass. With public Ackredit 0.9.0 the focused suite passes all 17 portable backend guards and explicitly skips the 11 function-observer cases unavailable in that provider.

Seven-repeat current pipeline timing: ordinary small conversion 44.29 us, opted-in backend capture 86.98 us, combined function/backend capture 107.86 us. Raw samples are retained; activation/preparation excluded, no hardware performance guarantee. Optional explicit activation remains unchanged.

Guard: tests/integration/test_function_citation_provider.py::test_normally_installed_consumer_outside_checkout checks actual installed producer versions, separate provider-absence science, repeated reference captures and pipeline parentage. The module additionally protects lazy activation, selected backends, entry/failure semantics, original article roles, prepared-credit reuse/replacement diagnostics and producer-blocked saved reading.

Final archived record and reviewed receipt distinguish original hosted pair, current local receiver/test inputs, normally installed wheels, mapped public Conda provider files and diagnostic/compatibility limits. No Windows or complete current hosted scientific-matrix claim is made from the earlier pilot matrix.

Stable-contract/release decisions remain uibcdf/ackredit#84/#87, uibcdf/molsyssuite#97 and uibcdf/moli#46. Public 0.9.0 bytes, canonical guides, mandatory dependencies and release decisions are unchanged. The accepted portable backend pilot is independently closed under #92.

Final unskipped head f34b111 also passes CI 37236634394 and suite policy 37236634877, verified through GH Run Receptor.

MolSysSuite #97 is preparing the central reconciliation of the changed Ackredit provider and the settled PyUnitWizard #94 experimental pilot.

Independent review verifies Ackredit source 1a5dd4566737f6195571b4cb421af6f01647c5f6, development wheel SHA-256 4c1d65477af8a09f66873c119b40c9228e6e678a79e360a2e4c0fd4644080604, native runs 37230213187/37230225286, all ten artifact ZIP digests and the immutable owning summary. Eight Linux/macOS-arm64 Python 3.11–3.14 cells execute six tests each: 48 total, zero skips/deselections/incomplete events. The independent aggregate equals hosted and owner identities. Raw source-wheel identities remain separate from public 0.9.0 Conda bytes and upcoming provisional 0.10.0.

The real report/producer-blocked reader, tuple-author registration and original references/versions/roles are reviewed. Invocation remains distinct from completed backend dispatch. PyUnitWizard f34b111 resolves its experimental receiving pilot with two numerical/citation guard cases, unchanged runtime and dependency floors. Current local 712/1 and 27 installed/one deliberate deselection remain owner evidence; independently inspected final ordinary CI 37236634394 has 699 passed/14 skips, distinct from the experimental-provider local scope. Final policy 37236634877 passes. No new test is dispatched centrally.

Central record will be devguide/pending_proposals/review_ackredit_function_citation_providers.md; new receipt devguide/rollouts/ackredit_function_receiving_reconciliation_97_20261004.json. MolSysSuite #97 remains partial for the maintainer's bounded stable-contract/schema/API-classification decision. The reviewed proposal is immutable Ackredit 40f930a852e3fe4517ba0a6d32d7dabc8624d285:devguide/function_provider_contract_review.md. Ackredit #93 may continue its separately authorized provisional release gates; no central stabilization approval is inferred or required for that provisional delivery.

No provider/member source, dependency floor, canonical guide or package bytes change in this central reconciliation. No mandatory observation or general rollout follows. We will return the immutable central commit and its inspected CI here; direct MOLI authority and consumer citation decisions remain with their respective owners.

Central #97 reconciliation is published in immutable commit 5184e9f2081e31aea90736a0401441972deb329e.

The changed provider 1a5dd4566737f6195571b4cb421af6f01647c5f6 and wheel SHA-256 4c1d65477af8a09f66873c119b40c9228e6e678a79e360a2e4c0fd4644080604 are independently reviewed. All ten artifact ZIP digests from receiving run 37230225286 match native metadata. The immutable owning summary with public Pytest Receptor 1.2.1 verifies the eight Linux/macOS-arm64 Python 3.11–3.14 cells: six tests per cell, 48 total, zero skips/deselections/incomplete events. Independent/native aggregates and owner package identities agree. No candidate was imported or component test dispatched centrally.

PyUnitWizard #94's closure at f34b111baee0eacf7c6c97f62e7143fcfdc2b9e7 is now recorded as settled experimental receiving. Conversion-factor/standardization guard assertions cover numerical outcomes and original citations, preserving existing runtime/dependency metadata. Hosted source 33fec8a, current scientific/local receiver ef85201 and final test/receipt source f34b111 retain separate identities. Source-CI skips, zero-skip receiving cells and owner-local 712/1 plus 27 installed/one deselection are kept distinct; final native ordinary CI 37236634394 independently has 699 passed/14 skipped.

Maintained record; new primary receipt. Exact-head central CI passes 337 tests, offline governance/Conda controls and separate coverage publication. Existing #102 private-member audit debt is unchanged and is not certified by this result.

#97 stays partial for the principal maintainer's bounded stable-contract/schema/API decision and direct-MOLI/provider handoff. The proposal at Ackredit 40f930a852e3fe4517ba0a6d32d7dabc8624d285:devguide/function_provider_contract_review.md is reviewable; it is not silently accepted. Ackredit #93's separately authorized provisional 0.10.0 release remains usable and needs its own exact Conda file/installed/public evidence. Source-wheel qualification, stable guarantees and public package delivery remain different claims.

No mandatory observer, client minimum, synchronized guide rollout, public artifact mutation or scientific change follows from this review. The public 0.9.0 portable contract and original Conda bytes remain unchanged. No owner issue is closed centrally.

Central #92/#97 prepares a read-only provisional Ackredit 0.10.0 publication reconciliation. Maintainer decision: observe_calls, prepare_credit and ackredit.provider@1 remain provisional; later gates/releases do not automatically stabilize them.

Original producer 16c356d54f245db8dd7fd6aaab72200df9a96e7d; noarch/ackredit-0.10.0-py_0.tar.bz2; SHA-256 2ed4841af32eaee603574b185a644fc16c9473497ad15a732788d6630b9cedc3. Existing native staging 37237182299, installed 37237527349, real receiving 37237527141 and promotion 37237913352 all succeed. The central shared operator independently returns public-verified, staging/main, no next command/mutation; downloaded public archive bytes match the digest. Annotated 0.10.0 tag resolves to the original producer.

All ten receiving ZIP digests match native metadata; immutable owning qualification_bundle.py summarize with public Pytest Receptor 1.2.1 reconstructs the uploaded aggregate. Eight Linux/macOS-arm64 Python 3.11–3.14 cells have six tests each, 48 total/no skips; the same Conda identity is checked before/after science. Producer/fallback wheels are separately identified. No component tests, install, dispatch, promotion, rebuild, overwrite or retagging is performed centrally.

Important retained limitation: direct archive inspection confirms Ackredit #94, CFF version 0.9.0/date 2026-10-03 inside distribution 0.10.0. The shared publisher preserved inputs correctly; semantic preparation/gate repair remains with Ackredit #93/#94. We record the owner's additive 0.10.1 recovery and wait for its exact file/installed/receiving/public handoff before selecting the corrected checkpoint. Current portable >=0.9.0 usage remains valid. No new component floor/guide/shared gate is introduced.

#92 bounded episode: four observed successful runs, aligned staging/installed/promotion subset three; creation/update envelope 12m18s, including queues/parallel receiving and excluding source CI/later correction. This is not measured human effort or a controlled improvement. Please retain whether the noarch_release operator was used before dispatch, actual manual steps and any already measured active time under devguide/noarch_release_operations.md; do not reconstruct invented timing estimates. Independent central operator use after publication does not establish ordinary end-to-end adoption.

Qualification scope: native source matrix has 1702 passed/seven skips per cell; installed selection has 1635 passed/eight skips per cell; real receiving has six/zero per cell. All counts remain distinct. Please clarify the source/installed collection difference in the owner release receipt; it is not treated as a proven defect or hidden under a full-suite equivalence claim.

Maintained records: devguide/pending_proposals/qualify_and_simplify_shared_release_pipeline.md and review_ackredit_function_citation_providers.md. New receipt: devguide/rollouts/ackredit_010_public_pipeline_92_97_20261004.json. Offline governance passes. #92/#97 remain partial for their explicit remaining evidence/decisions. Immutable central commit and actual CI will be returned here after publication.

Published immutable central commit 05f866ab7f17af6b046e89befa014460d5d13160 records the principal maintainer's decision: observe_calls, prepare_credit and ackredit.provider@1 remain provisional. The settled PyUnitWizard #94 experimental pilot remains accepted; no stable compatibility promise, mandatory observer, new minimum or guide rollout follows. A later corrected release will not automatically stabilize them. Revisit only by explicit later decision and the owning Ackredit #84/#87, MolSysSuite #97 and direct MOLI #46 handoff.

#92 ordinary release review verifies original Ackredit 0.10.0 producer 16c356d54f245db8dd7fd6aaab72200df9a96e7d and public archive SHA-256 2ed4841af32eaee603574b185a644fc16c9473497ad15a732788d6630b9cedc3. Shared operator returns public-verified/no command/no mutation; downloaded public bytes and original annotated tag agree. Native stage/installed/receiving/promotion steps succeed; all ten receiving ZIPs and provider-owned independent aggregate match, eight cells/six tests each/48 total/no skips and exact Conda identity before/after science.

Retained limitation: direct public archive inspection confirms CFF 0.9.0/date 2026-10-03 inside package 0.10.0. Semantic preparation/gate repair belongs to Ackredit #94/#93; shared byte preservation is correct. Additive corrected 0.10.1 source dd500842b6085111e01e62cfc243f68406eb8cc7 and SHA-256 26e75a0780ad4e6abc2de55df90b29b4a2aa4e510d6b50fa54a5812ad929228e have new owner-reported installed/receiving successes. Those are not qualified by the prior 0.10.0 central checkpoint: wait for corrected exact-file/public/tag/clean-install handoff and its separate review before selecting it. No existing archive/tag is changed.

The bounded four-run 0.10.0 creation/update envelope is 12m18s; source and installed totals retain separate scopes (1702/7 versus 1635/8 per cell). #92 still needs actual ordinary end-to-end operator use/manual-step/active-effort evidence and owner clarification of collection scope; no retrospective time estimate or controlled-effort improvement is invented. The original source/installed/scientific results are inspected, not rerun centrally.

Exact-head CI: governance, 337 tests (OK) and Conda controls pass in attempt 1. Separate coverage publisher fails in attempts 1 and 2 while downloading Codecov (curl 35/TLS handshake; absent signature files thereafter). Attempt 2 reruns only the failed publisher using retained XML. Overall run stays failed; coverage recovery is tracked in MolSysSuite #69. No scientific or central test suite was repeated for that retry.

Publication/decision receipt; #97 maintained record; #92 maintained record. #92/#97 remain partial for their explicit remaining evidence/deferred decisions; owner-local issue completion stays with each owner.

Ackredit 0.10.1 is published and independently verified on the public uibcdf Conda channel. The immutable tag identifies the original producer dd500842b6085111e01e62cfc243f68406eb8cc7.

Exact file: ackredit-0.10.1-py_0.tar.bz2.
SHA-256: 26e75a0780ad4e6abc2de55df90b29b4a2aa4e510d6b50fa54a5812ad929228e.
Source CI and full source matrix pass at that producer; both policy lanes pass.
Staging built one noarch file. Installed qualification passes all eight Linux x86-64/macOS arm64 × Python 3.11–3.14 cells, including the full installed suite and provenance recheck.
Real PyUnitWizard receiving matrix passes 48 mandatory tests across those eight cells, with zero skips/deselections. Shared verification before and after science binds every cell to the same Conda file. All ten native scientific artifact ZIP digests and contents independently verify; the reconstructed aggregate equals the hosted summary. The producer remains 33fec8a627505a4f5426babe87e8e85438105041.
Promotion adds the public main label to those same bytes. The shared verifier independently confirms the main label and public solver index.
A fresh-cache public-channel installation on Linux/Python 3.14.7, outside both repositories, passes installed origins/resources, matching CFF/runtime/distribution 0.10.1, portable capture/reuse and detached reader, provisional observation/prepared credit/contextual report, CLI and pip check.
Public Sabueso 0.12.0 (py_1) passes its 56 unchanged receiving tests from its original public producer and its offline example: source traces, three result bibliographies, workflow union and saved readers. Both libraries are normally installed public Conda distributions. This receiving result is separate from a new Sabueso candidate's own qualification.

The reviewed delivery receipt, installation instructions and release notes retain the identities and bounded evidence. Ackredit #93 delivers the checkpoint; #94 adds plan/CFF and installed citation-version guards and corrects the stale self-citation in additive 0.10.1. Original 0.10.0 bytes and tag are unchanged; this was Ackredit preparation/qualification, not a shared publisher fault.

Availability is separate from stable promotion and client adoption. observe_calls, prepare_credit and declaration interpretation remain provisional under Ackredit #84/#87, MolSysSuite #97 and MOLI #46. No automatic hooks, new core dependency, shared adoption requirement or consumer minimum is introduced. Clients using only the released ackredit.attribution@1 contract can keep ackredit>=0.9.0; the original public 0.9.0 evidence remains valid. Owners should review explicit entry versus completed-backend semantics, original software/article versions, failure/absence and saved-reader behavior before requiring provisional capabilities or certifying their next release. The canonical stable integration guide is unchanged; a broader synchronized-guide rollout waits for that review.

Central #97 independently verifies corrected public Ackredit 0.10.1 in 098aa38d84f2c8c4f226c0a9ee94e57016113d50: original producer dd500842b6085111e01e62cfc243f68406eb8cc7; exact noarch/ackredit-0.10.1-py_0.tar.bz2; SHA-256 26e75a0780ad4e6abc2de55df90b29b4a2aa4e510d6b50fa54a5812ad929228e. Public bytes, original annotated tag and corrected CFF 0.10.1 agree. Installed and real PyUnitWizard receiving matrices pass on Linux/macOS arm64 and Python 3.11–3.14; ten native ZIPs independently verify and the immutable owning summary reconstructs 48 tests/no skips with exact Conda identity before/after science. Owner Ackredit #93/#94 are closed; prior 0.10.0 bytes/tag remain intact with their recorded limitation.

The principal maintainer explicitly retains provisional observe_calls, prepare_credit and ackredit.provider@1. This completes a corrected public evidence checkpoint, not stable adoption or a consumer version-floor increase. Existing portable >=0.9.0 usage and the accepted bounded PyUnitWizard #94 experimental pilot remain valid. Direct MOLI #46 contract review remains deferred; no mandatory rollout follows. No consumer test, installation, rebuild or publication is dispatched centrally.

Independent receipt. Exact-head CI, attempt 1: governance, 337 tests/OK, Conda controls and coverage XML retention pass. Separate coverage publisher fails during Codecov download with curl 35/TLS handshake, followed by missing signature files. Overall run remains failed. The same transport failure is already tracked in MolSysSuite #69; no blind retry or relaxed verification is added.

The accepted decision to keep observe_calls, prepare_credit and ackredit.provider@1 provisional is now recorded explicitly in Ackredit's maintained contract review, with separate exit criteria for each surface. This respects the principal-maintainer decision retained centrally at 05f866ab7f17af6b046e89befa014460d5d13160; successful publication does not silently stabilize them.

The review now distinguishes decisions that remain from evidence already available: schema interpretation/compatibility, observer entry/await/restoration and warning-gap semantics, and inert prepared-callable/current-session/host-owned completion semantics. Each can be accepted, amended or deferred independently. PyUnitWizard #94's experimental receiving review is settled; stable provider acceptance, shared/platform obligations, canonical-guide distribution, actual client adoption and a delivering release remain separate states. No additional engine/platform or repeated identical matrix is made a new requirement.

tests/test_provider_lifecycle.py adds five supplementary guards:

Cancelled and scientifically failed concurrent tasks retain only entered-method references until the host actually invokes completed-backend credit. Surviving observer leases remain usable and original exports restore.
One prepared callable serves four isolated thread sessions, each with eight independent reused captures.
Explicit warning-as-error refuses the scientific call and restores exports and the parent scope; a fresh activation remains usable.
An unawaited coroutine credits nothing.
Delayed prepared credit cannot mutate an expired inherited result capture; its later result and enclosing workflow receive the credit.

All five pass against the unchanged public 0.10.1 Conda file on Linux/Python 3.14.7 outside source, using the unchanged pinned shared installed_noarch.run_tests operation (c3e2b9b3dabf3d1c65349c389a23048957bea21a) for same-interpreter import provenance before and after tests. The supplementary receipt binds the exact archive SHA-256, runtime/test-module hashes and complete/integrity-valid Pytest Receptor outcome. It retains two assert-rewrite warnings from libraries imported before the administrative pytest call, separately from five passing/no-skip tests. Jinja2 was added only as a qualification tool in that isolated environment; core requirements are unchanged and pip check passes.

These same five guards pass in the routine molsyssuite@uibcdf_3.14 development environment. The isolated source suite passes 1,721 tests, Ruff check/format, current devguide indexes and strict Sphinx. This is additional lifecycle coverage, not another real-producer/hosted full matrix, performance result, runtime change, stable promise, tag or public-file mutation. The existing exact-file release/real-producer evidence remains separate.

The unskipped published head 16ddeb83d35584c5e79b85942d0131f758d9ef67 also passes CI 37274224721 (seven jobs), suite policy 37274225313 and publication policy 37274225437, inspected with GH Run Receptor. The primary editable installation is refreshed in Python 3.14.7 as 0.10.1+2.g16ddeb8.dirty, with matching runtime/distribution metadata and successful pip check; all nine pre-existing human files are byte-identical.

Contract review and exit criteria; supplementary public-runtime receipt; lifecycle guards. Ackredit #84/#87 remain partial for the explicit later product decision; MolSysSuite #97 and direct MOLI #46 retain their ownership. The stable portable 0.9.0 floor, consumer guides and member sources are unchanged.

0.11.0 selected and producer inputs published; exact-file gates next

The maintainer authorized completing the next release. Ackredit #107 selects 0.11.0 and original candidate source 85deae594e65b2fd443d6ca9a7347eb2bda537e1. The reviewed staged noarch plan, matching root/packaged self-citation version/date, required composition/evidence resources and release scope are committed on main.

This checkpoint delivers the accepted bounded stable-provider contracts under Ackredit #84/#87 plus reviewed #95–#106 metadata fidelity, repeated-use performance, portable CLI/composition/explanation and opt-in evidence capabilities. The evidence APIs remain provisional. Existing public 0.10.1 and portable >=0.9.0 compatibility are unchanged; 0.11.0 is not yet publicly available or qualified. Client adoption/release ownership and central guide synchronization remain separate.

Complete local source qualification passes 2,078 tests without skips in Python 3.14.7, Ruff, report indexes, strict fresh Sphinx and the adopted shared noarch recipe inspector. Local receipt. Human BibTeX/LaTeX/notebook work is excluded and preserved.

The exact producer's ordinary CI/policy lanes and full eight-cell source matrix are now running. After they pass, build one Conda archive once, inspect/stage it and retain the original filename/SHA-256. Require full eight-cell exact-file installed tests plus real PyUnitWizard receiving and independent offline readers (including original 0.9.0 fallback). Only then promote that same file, verify public labels/index and a clean public installation, register the original-producer tag and deliver version/file/digest/evidence.

Coordination: uibcdf/ackredit#107, uibcdf/molsyssuite#97 and uibcdf/moli#46. Receiving owners: #94 and uibcdf/sabueso#108. This is advance release notice, not an adoption requirement or certification of a consumer release. Any new shared publisher defect will be reported with its original run/receipts.

Ackredit 0.11.0 is publicly delivered. The accepted bounded prepare_credit, observe_calls and ackredit.provider@1 compatibility promise now starts at ackredit>=0.11.0. The portable-only minimum remains >=0.9.0; evidence representation, opt-in collection and integrated reporting remain provisional. Consumer adoption and consumer release qualification remain owner-local.

Original producer and immutable tag: 85deae594e65b2fd443d6ca9a7347eb2bda537e1 / 0.11.0.
Exact file: ackredit-0.11.0-py_0.tar.bz2.
SHA-256: df8963ca2d286f50b19eb778e95c54c5ebb79c12fb55a6504e7c23daf5717d4f.
Public archive.
Eight-cell full installed matrix.
Eight-cell real receiving matrix: 72 mandatory tests without skips/deselections, pinned PyUnitWizard, original 0.9.0 fallback and fresh independent readers. All original ZIP identities and the downloaded aggregate verify independently.
Exact-file promotion: same original archive, verified public main label/index. No rebuild, replacement or reupload.
Clean public Linux/Python 3.14.7 installation passes actual downloaded archive bytes, Conda URL/digest/provenance, citation/runtime/distribution identity, provider/prepared and portable readers, composition, provisional evidence CLI/library agreement and pip check.
Public delivery receipt retains exact original identities, executed source gates, installed/receiving/public claims and limits.

Closeout aae9ad98c89d162b946a65fa01603707db02359a passes all seven CI jobs, suite policy and publication policy, inspected with GH Run Receptor. Local closeout passes 262 contract/documentation tests without skips, Ruff, indexes and strict Sphinx; executable ASTs are unchanged from the qualified producer apart from docstrings. All nine primary human files retain their original bytes.

Ackredit #84/#87/#107 records are archived with the delivery outcome. Canonical ACKREDIT_GUIDE.md now states the delivered 0.11.0 stable-provider floor. Please coordinate its consumer synchronization through the existing registry/owner route; no sibling copy or implementation was repaired locally. The standalone declaration validator/author tooling, separate evidence-contract acceptance and general 1.0 decision remain roadmap work.

The shared development environment currently has unrelated scientific dependency conflicts reported in MolSysSuite #82. The clean public environment passes dependency closure; these are distinct observations, not a consumer release certification.

Provider-author tooling notice under uibcdf/molsyssuite#97: uibcdf/ackredit#111 adds a provisional development API in uibcdf/ackredit#112, exact source f81dea77151855a725c836539caca2473f3c8fb2.

validate_provider(module) -> dict reuses the observer's declaration parser and returns a detached ackredit.provider@1 declaration with merged module/function metadata. It does not register bibliography, record uses, install Ackredit wrappers or query a DOI. Explicit lazy export loaders retain their own side effects; current-registry conflicts and science/citation-quality verification are out of scope. E012 wording becomes operation-neutral while its code and refusal facts remain unchanged.

Public 0.11.0 and baseline 71f7f1984de1a3925c32d29074784bb553e89bfc have no standalone export. Existing accepted observer/provider meanings remain intact. There is no required consumer migration, changed runtime minimum or automatic adoption. The provider's signature/result construction is provisional; this notice does not claim a new public artifact. Existing released observation/portable paths remain the fallback.

This member is a candidate producer-author user based on its existing Ackredit pilot, not a verified caller of the new operation. Its owner decides whether to use the validator after qualified source/version adoption. Wider registered guide consumers are potential audiences, with no standalone usage claimed. The concise author guide/example is subsequent work.

Local Python 3.14.7 passes 2,122 tests without skips/warnings, including 24 standalone guards and the normally installed producer/reader case, through Pytest Receptor; Ruff 0.16.5 and strict docs pass. Exact-head hosted gates are inspected with GH Run Receptor. Canonical guide synchronization, source adoption, API promotion and public package delivery remain separate evidence.

The maintainer explicitly approved all three bounded recorder-evidence surfaces under Ackredit #114, following review source f29d0f170c7eb758e85d9a583bda13c7cf6958a4. The accepted source decision and canonical guidance are ready for publication after that baseline.

Accepted: detached AttributionEvidence/schema-1 interpretation and positional original occurrences; provider-observer-only opt-in capture collection with active same-session positive facts and diagnostic/scientific failure limits; explicitly requested integrated workflow/CLI presentation preserving default and inert reader behavior. Accepted contract and canonical guide implement the exact bounded scope.

The forward public evidence promise awaits a separately selected, qualified future release. Original public 0.11.0 remains provisionally classified for evidence; its stable-provider minimum >=0.11.0 and portable-only minimum >=0.9.0 are unchanged. Standalone validate_provider remains provisional. There is no new artifact, runtime/schema change, mandatory adoption, broad-recorder prerequisite or general 1.0 acceptance.

Local Python 3.14.7: 381 selected contract/documentation/API/reporting cases pass without skips/warnings through Pytest Receptor; Ruff 0.16.5, indexes, current suite repository conformance and strict Sphinx pass. Both changed runtime modules retain identical ASTs after removing docstrings. The exact published source/CI controls will be linked in #114 after inspection with GH Run Receptor.

Registered canonical-guide consumers are PyUnitWizard, MolSysMT, MolSysViewer, TopoMT, PharmacophoreMT and ElastNetMT. Existing owner issues receive this notice; Sabueso's existing receiving issue receives the same release-boundary notice separately. MolSysSuite owns guide-copy synchronization through its registry. Do not repair consumer copies locally or infer runtime adoption from synchronized text. Client owners retain implementation, required-version and release qualification decisions.

The maintainer has explicitly authorized Ackredit 0.12.0 for pre-1.0 stabilization under Ackredit #127, following #126's retained scope/adoption pause. This is the actual delivery of completed work before consumers adopt; general public 1.x stability remains deferred.

The old public baseline is immutable 0.11.0, producer 85deae594e65b2fd443d6ca9a7347eb2bda537e1, ackredit-0.11.0-py_0.tar.bz2, SHA-256 df8963ca2d286f50b19eb778e95c54c5ebb79c12fb55a6504e7c23daf5717d4f. The candidate is new 0.12.0/build 0, staged noarch; its source/file/digest will be bound by actual production. It includes the already reviewed evidence/validator promises, standalone author validation/example, citation-key/name/book-kind repairs, DOI presentation and deferred feature imports. Python/platform support, runtime dependency floors, schemas and optional-host behavior remain unchanged.

After successful exact-source/full-installed/real-PyUnitWizard-receiving/public gates, the newly accepted evidence/validator promises will have public minimum >=0.12.0. Existing portable-only >=0.9.0 and stable provider/prepared >=0.11.0 minima remain available; no consumer minimum is automatically raised. Public 0.11.0 lacks standalone validation and retains provisional evidence contracts. Candidate guide text names 0.12.0 as selected but still unqualified; it is not yet a public installation route. The original 0.11.0 file is not rebuilt or relabelled.

PyUnitWizard #94 and Sabueso #108 are existing receiving clients; registered guide/runtime consumers retain their own adoption and release decisions. Existing MolSysSuite #97/MOLI #46 coordination owns canonical-guide synchronization and cross-component boundaries. No sibling source or guide copy is edited here. The 16-component status audit found remote lag and independent work; all were preserved. Source/controlled receiving, habitual dogfooding, guide rollout and client delivery remain separately observed outcomes.

This notice precedes candidate/public publication. Final producer/file/digest, actual gate/clean-public evidence and the canonical-guide commit/hash will follow. After delivery, retain the feature-development pause and respond to concrete owning feedback rather than extending scope.

Ackredit 0.12.0 delivered for pre-1.0 stabilization

This is the after-publication handoff for Ackredit #127, following the earlier before-delivery notice. The public uibcdf Conda channel now provides Ackredit 0.12.0, file ackredit-0.12.0-py_0.tar.bz2, SHA-256 160b452c2b9de3b44bc6c6e2f4bd8c47e44b1d2779b620f63048231708a1d4aa. Canonical tag 0.12.0 remains at original producer 6f4dbf39996a7185b8aaff7c52b9daeb167a100a.

All required exact-source gates, eight full installed cells, 72 mandatory real PyUnitWizard receiving tests without skips/deselections and independent native-artifact/aggregate verification pass against that same file. Promotion 37599450602 adds the public label without rebuilding. Independent public labels/solver index and a fresh public Linux/Python 3.14 installation verify the original download SHA, all 70 shipped package files, runtime/distribution/self-citation identity, dependency closure, detached readers/CLI, evidence/validation and actual optional absence. The candidate installation also passes 158 selected contracts without skips and bounded real BibTeX/Pandoc, two BibLaTeX/Biber styles and JabRef 5.15 receiving, retaining tool warnings and fixture limits.

The public receipt and candidate-installed receipt retain original identities, native evidence and practical limits. Initial task-cache index lag was resolved with a read-only cache refresh/full public index; promotion was not repeated.

The updated owned canonical integration guide is standards/ACKREDIT_GUIDE.md at 887b81c525ff9b285150b2d801c557321b4d48fa, SHA-256 6353892d552eab79973cecec5f8e3c8c31e146416e1cb481786e21cd3fcbf99d. This guide-publication identity is separate from original archive/tag source. Consumer copies are synchronized by the central registry/owning consumer route, never repaired locally.

Public minima remain distinct: portable attribution >=0.9.0, stable provider/observer/prepared credit >=0.11.0, and the separately accepted bounded recorder-evidence/explicit reporting and standalone validate_provider promises >=0.12.0. Public 0.11.0 retains its original evidence classification and lacks the standalone validator. General public 1.x commitment remains deferred until a separately authorized, qualified 1.0.0.

Python 3.11–3.14, Linux x86-64/macOS arm64, the four runtime dependency floors, saved schemas and host optional-operation pattern are unchanged. Qualification uses public SMonitor 0.19.0, ArgDigest 0.15.0 and DepDigest 0.13.0; no host must enable observation/evidence automatically. Consumer source, scientific schemas, guide adoption and released dependency closure retain their owners. This provider release does not certify client releases, arbitrary publication tools or habitual dogfooding.

The accepted pause in proactive feature development now begins so MolSysSuite/MOLI components can adopt this installable checkpoint and supply habitual-workflow feedback. Resume focused Ackredit work for concrete owning feedback, a demonstrated defect or an explicit maintainer request. Acknowledgements remain deferred under #124 until a real owned case exists; no last-pre-1.0 schedule is imposed.

CFF/CSL compatibility handoff from uibcdf/molsyssuite#103 and provider uibcdf/ackredit#98.

Delivered change: CFF-declared entities/persons retain their identity in CSL authors/editors; comma-containing institutions are not invented people. Explicit personal suffixes/particles survive. Original human-readable names and identifiers remain saved metadata. Original-name hints apply only while the corresponding list still matches; explicit replacements and typed preferred-work ownership remain authoritative. Existing portable schema, optional loading and generic human-readable string paths are preserved. No new dependency or mandatory client minimum follows from this correction.

Exact releases independently reviewed centrally:

First corrected public file: ackredit-0.11.0-py_0.tar.bz2, producer 85deae594e65b2fd443d6ca9a7347eb2bda537e1, SHA-256 df8963ca2d286f50b19eb778e95c54c5ebb79c12fb55a6504e7c23daf5717d4f. Original installed matrix 37429662858; same-byte promotion 37430816845. All four repaired runtime files and the CFF regression module are byte-identical to fix 46b2635b17f47e03e18ab576016ace94b7e6461b.
Current owner-delivered stabilization: ackredit-0.12.0-py_0.tar.bz2, producer 6f4dbf39996a7185b8aaff7c52b9daeb167a100a, SHA-256 160b452c2b9de3b44bc6c6e2f4bd8c47e44b1d2779b620f63048231708a1d4aa. Original installed matrix 37588182382; same-byte promotion 37599450602. The CFF regression module is unchanged; two exporter/name files have later bibliographic fixes, so source ancestry is not claimed as byte equality with the original fix. Full 0.12.0 contract/adoption work remains separately owned in Deliver 0.12.0 for pre-1.0 stabilization and consumer adoption ackredit#127 and Review provisional dependency-free Ackredit function citation providers molsyssuite#97.

For each release, the read-only shared operator independently reacquires original producer receipts/digests, exact-file Linux/macOS-arm64 Python 3.11–3.14 installed jobs and executed required steps, public main label and solver index. All four required original source gates are successful. Public archive download digest and shipped repaired-module hashes match the candidate source. Original native promotion receipt ZIP/digest and verified staging-to-main effect agree on those same bytes. No build, upload, promotion, component suite or artifact substitution was performed centrally. Owner-reported real PyUnitWizard receiving outcomes remain separately attributed; they do not qualify every registered guide client.

Owner review when this path is next touched: determine whether your actual CFF inputs/exported CSL use affected entities, single-part or comma-containing personal names, suffixes/particles or preferred works. If applicable, retain original bibliography/versions, explicit replacement behavior, producer-free saved reading and meaningful host-owned diagnostics; qualify your own selected dependency/file normally before claiming consumer compatibility. Otherwise record no affected use or a tracked deferral. Review the current public release when selecting an upgrade; 0.11.0 here records first delivery, not a requested downgrade/pin. This notice does not mandate adoption, guide synchronization or a new scientific suite. Use the registered synchronizer for any future canonical guide refresh. Normal direct-push/CI policies and existing scientific deferrals continue.

Consumer identity: uibcdf/pyunitwizard at inspected 4606aa18dfecfeda5b5cb1b38fa8ea2ad3db4faa; identified from the registered ACKREDIT_GUIDE.md consumer list.

This reuses your existing owning attribution/guide issue for notice. Its previous closed/open outcome is preserved; no previously closed implementation is reopened or recertified. Please keep any later concrete migration or receiving defect in its own owning issue linked to uibcdf/molsyssuite#103.

Accepted optional contract review — advance delivery notice (MolSysSuite #97)

The principal maintainer has accepted registration of Ackredit's bounded public promises for optional use: portable ackredit.attribution@1 from 0.9.0; prepare_credit, observe_calls and ackredit.provider@1 interpretation from 0.11.0; bounded recorder evidence and standalone validate_provider from 0.12.0. General public 1.x remains a separate future qualification. Existing 0.10.x provisional evidence retains its original scope.

The central review independently reconstructed original receiving run 37588186298: original source 6f4dbf39996a7185b8aaff7c52b9daeb167a100a, all ten native ZIP digests/source bindings, eight Linux/macOS arm64 × Python 3.11–3.14 cells, 72 passed/no skips. This certifies that bounded original pipeline, not every client's current source.

We will deliver the provider's current canonical ACKREDIT_GUIDE.md from b319e7f868301d01a3325a37a1937da9267aa737, SHA-256 6353892d552eab79973cecec5f8e3c8c31e146416e1cb481786e21cd3fcbf99d, through the registered synchronizer. The six consumers are PyUnitWizard, MolSysMT, MolSysViewer, TopoMT, PharmacophoreMT and ElastNetMT. Delivery uses isolated clones, guide-only commits and exact-head administrative gates; science/browser suites remain deferred where already authorized. Administrative checks do not clear scientific full-suite debt.

Client action: review the optional capability only when needed. No dependency upgrade or runtime change is requested. Activation must be explicit; observed entry, declaration or validation does not prove successful science or complete instrumentation. The application owns citations, supported call paths, failure/warning policy, original saved records, actual receiving tests and any claimed released fallback. The existing portable >=0.9.0 route remains available. Guide delivery, functional adoption and tested artifacts will be reported separately.

Detailed central evidence: immutable review record. Exact guide commit and gate handoffs will follow in this same issue. Existing closed pilots remain closed; this notice adds no new runtime-adoption obligation.

docs(ackredit): deliver optional published contracts (#94; MolSysSuite …

Canonical guide delivered — MolSysSuite #97

Guide-only commit 4e4eefad58f12778f31e9ec3f81f53193750d567 is published on main. ACKREDIT_GUIDE.md is byte-identical to provider source b319e7f868301d01a3325a37a1937da9267aa737, SHA-256 6353892d552eab79973cecec5f8e3c8c31e146416e1cb481786e21cd3fcbf99d, distributed with the registered synchronizer. The six registered copies all pass the synchronization check.

Applicable administrative gates:

.github/workflows/molsyssuite-policy.yml: exact commit, native attempt/jobs and required executed steps independently verified.

Local report/index/conformance checks pass for the delivery's applicable policy. Internal guide-only [skip ci] used the authorized manual administrative route. Science/browser/full-suite debt remains with the component's maintainers and existing recovery/deferral records; these results do not clear it.

Accepted optional public capability floors: portable attribution >=0.9.0; prepare_credit, observe_calls, ackredit.provider@1 interpretation >=0.11.0; bounded recorder evidence and validate_provider >=0.12.0. General public 1.x remains separately pending. The central policy records these bounded choices; no dependency upgrade or runtime integration has been performed here. Existing closed pilots remain closed.

Next action when relevant: choose only the needed optional feature; review explicit application activation, call/exclusion scope, invocation versus completed science, citation/failure/warning handling, original saved records, actual receiving qualification and any claimed released fallback. Runtime adoption and scientific references remain your team's decision. The original eight-cell/72-test Ackredit pipeline is qualified for its pinned producer; it does not qualify this component's current source.

Coordination: uibcdf/molsyssuite#97; direct MOLI review: uibcdf/moli#46. The central durable receipt will record source/public/guide/adoption evidence separately.

Navigation Menu

Repository navigation

Pilot dependency-free function citation declarations for PyUnitWizard #94

2 remaining items

0.11.0 selected and producer inputs published; exact-file gates next

Ackredit 0.12.0 delivered for pre-1.0 stabilization

Accepted optional contract review — advance delivery notice (MolSysSuite #97)

Canonical guide delivered — MolSysSuite #97

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Pilot dependency-free function citation declarations for PyUnitWizard #94

Description

Activity

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

LMMV commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

dprada commented on Oct 4, 2026

LMMV commented on Oct 4, 2026

LMMV commented on Oct 4, 2026

2 remaining items

LMMV commented on Oct 5, 2026

LMMV commented on Oct 5, 2026

dprada commented on Oct 5, 2026

LMMV commented on Oct 5, 2026

dprada commented on Oct 5, 2026

dprada commented on Oct 6, 2026

0.11.0 selected and producer inputs published; exact-file gates next

dprada commented on Oct 6, 2026

LMMV commented on Oct 6, 2026

LMMV commented on Oct 6, 2026

LMMV commented on Oct 7, 2026

LMMV commented on Oct 7, 2026

Ackredit 0.12.0 delivered for pre-1.0 stabilization

LMMV commented on Oct 8, 2026

LMMV commented on Oct 8, 2026

Accepted optional contract review — advance delivery notice (MolSysSuite #97)

LMMV commented on Oct 8, 2026

Canonical guide delivered — MolSysSuite #97

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions